mirror of
https://github.com/tinygrad/tinygrad.git
synced 2026-08-16 04:38:27 +00:00
Compare commits
6
Commits
| Author | SHA1 | Date | |
|---|---|---|---|
 geohot
|
e4d8e03954 | ||
|
geohot
|
09574a096a | ||
|
geohot
|
b89c233917 | ||
|
geohot
|
00d847afdf | ||
|
geohot
|
8d75eed0a4 | ||
|
geohot
|
e59cbf78bd |
@@ -54,7 +54,7 @@ runs:
|
||||
steps:
|
||||
- name: Set up Python ${{ inputs.python-version }}
|
||||
id: setup-python
|
||||
uses: actions/setup-python@v6
|
||||
uses: actions/setup-python@v5
|
||||
with:
|
||||
python-version: ${{ inputs.python-version }}
|
||||
|
||||
@@ -70,7 +70,7 @@ runs:
|
||||
- name: Cache Python packages
|
||||
if: github.event_name != 'pull_request'
|
||||
id: restore-venv
|
||||
uses: actions/cache@v5
|
||||
uses: actions/cache@v4
|
||||
with:
|
||||
path: ${{ github.workspace }}/.venv
|
||||
key: venv-${{ runner.os }}-${{ runner.arch }}-python-${{ steps.setup-python.outputs.python-version }}-${{ inputs.deps }}-${{ inputs.pydeps }}-${{ env.CACHE_VERSION }}
|
||||
@@ -85,7 +85,7 @@ runs:
|
||||
key: downloads-${{ github.job }}-${{ inputs.key }}-${{ env.CACHE_VERSION }}
|
||||
- name: Cache downloads
|
||||
if: inputs.key != '' && github.event_name != 'pull_request'
|
||||
uses: actions/cache@v5
|
||||
uses: actions/cache@v4
|
||||
with:
|
||||
path: ${{ runner.os == 'Linux' && '~/.cache/tinygrad/downloads/' || '~/Library/Caches/tinygrad/downloads/' }}
|
||||
key: downloads-${{ github.job }}-${{ inputs.key }}-${{ env.CACHE_VERSION }}
|
||||
@@ -202,7 +202,7 @@ runs:
|
||||
key: ${{ runner.os }}-${{ runner.arch }}-apt-${{ steps.apt-pkgs.outputs.hash }}-${{ env.CACHE_VERSION }}
|
||||
- name: Cache apt
|
||||
if: runner.os == 'Linux' && (inputs.opencl == 'true' || inputs.amd == 'true' || inputs.cuda == 'true' || inputs.webgpu == 'true' || inputs.llvm == 'true') && github.event_name != 'pull_request'
|
||||
uses: actions/cache@v5
|
||||
uses: actions/cache@v4
|
||||
with:
|
||||
path: /var/cache/apt/archives/
|
||||
key: ${{ runner.os }}-${{ runner.arch }}-apt-${{ steps.apt-pkgs.outputs.hash }}-${{ env.CACHE_VERSION }}
|
||||
@@ -269,7 +269,7 @@ runs:
|
||||
- name: Cache gpuocelot
|
||||
if: inputs.ocelot == 'true' && github.event_name != 'pull_request'
|
||||
id: cache-build
|
||||
uses: actions/cache@v5
|
||||
uses: actions/cache@v4
|
||||
env:
|
||||
cache-name: cache-gpuocelot-build-1
|
||||
with:
|
||||
@@ -287,7 +287,6 @@ runs:
|
||||
|
||||
CMAKE_ARGS="-Wno-dev -G Ninja -DOCELOT_BUILD_TOOLS=OFF -DCMAKE_BUILD_ALWAYS=0 -DBUILD_TESTS_CUDA=OFF -DCMAKE_POLICY_VERSION_MINIMUM=3.5"
|
||||
if [[ "${{ runner.os }}" == "macOS" ]]; then
|
||||
sudo xcode-select -s /Applications/Xcode_16.2.app/Contents/Developer
|
||||
CMAKE_ARGS="$CMAKE_ARGS -DBoost_INCLUDE_DIR=$(brew --prefix boost)/include -DBoost_LIBRARY_DIR=$(brew --prefix boost)/lib"
|
||||
fi
|
||||
|
||||
|
||||
@@ -28,7 +28,7 @@ jobs:
|
||||
timeout-minutes: 15
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -56,8 +56,6 @@ jobs:
|
||||
python3 -c "from tinygrad.runtime.autogen import libusb"
|
||||
python3 -c "from tinygrad.runtime.autogen import mesa"
|
||||
python3 -c "from tinygrad.runtime.autogen import avcodec"
|
||||
python3 -c "from tinygrad.runtime.autogen import llvm_qcom"
|
||||
python3 -c "from tinygrad.runtime.autogen import mlx5"
|
||||
REGEN=1 python3 -c "from tinygrad.runtime.autogen import libclang"
|
||||
- name: Check for differences
|
||||
run: |
|
||||
@@ -69,7 +67,7 @@ jobs:
|
||||
fi
|
||||
- name: Upload patch artifact
|
||||
if: failure()
|
||||
uses: actions/upload-artifact@v7
|
||||
uses: actions/upload-artifact@v4
|
||||
with:
|
||||
name: autogen-ubuntu-patch
|
||||
path: autogen-ubuntu.patch
|
||||
@@ -80,7 +78,7 @@ jobs:
|
||||
timeout-minutes: 15
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -100,7 +98,7 @@ jobs:
|
||||
fi
|
||||
- name: Upload patch artifact
|
||||
if: failure()
|
||||
uses: actions/upload-artifact@v7
|
||||
uses: actions/upload-artifact@v4
|
||||
with:
|
||||
name: autogen-macos-patch
|
||||
path: autogen-macos.patch
|
||||
@@ -111,7 +109,7 @@ jobs:
|
||||
timeout-minutes: 15
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -139,7 +137,7 @@ jobs:
|
||||
fi
|
||||
- name: Upload patch artifact
|
||||
if: failure()
|
||||
uses: actions/upload-artifact@v7
|
||||
uses: actions/upload-artifact@v4
|
||||
with:
|
||||
name: autogen-comgr2-patch
|
||||
path: autogen-comgr2.patch
|
||||
|
||||
+152
-200
@@ -32,7 +32,7 @@ jobs:
|
||||
if: github.repository_owner == 'tinygrad'
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
# brew install uv
|
||||
- name: setup python environment
|
||||
run: |
|
||||
@@ -49,40 +49,9 @@ jobs:
|
||||
source /tmp/tinygrad_pytest_ci/bin/activate
|
||||
pytest -nauto --durations=20
|
||||
- name: openpilot compile3 0.10.1 driving_vision
|
||||
run: FLOAT16=1 DEV=CL IMAGE=1 python3.11 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/720392c9a5b986981fdbed1bb8c47a6c5573a50e/selfdrive/modeld/models/driving_vision.onnx
|
||||
|
||||
testframeworkpytest:
|
||||
name: framework pytest
|
||||
env:
|
||||
CI: ""
|
||||
CAPTURE_PROCESS_REPLAY: "0"
|
||||
runs-on: [self-hosted, framework]
|
||||
timeout-minutes: 10
|
||||
defaults:
|
||||
run:
|
||||
shell: bash -e -o pipefail {0}
|
||||
if: github.repository_owner == 'tinygrad'
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
- name: setup python environment
|
||||
run: |
|
||||
rm -rf /tmp/tinygrad_pytest_ci
|
||||
uv venv /tmp/tinygrad_pytest_ci
|
||||
source /tmp/tinygrad_pytest_ci/bin/activate
|
||||
uv pip install .[testing]
|
||||
- name: setup other stuff
|
||||
run: |
|
||||
mkdir -p extra/remu/target/release/
|
||||
ln -s ~/tinygrad/extra/remu/target/release/libremu.so extra/remu/target/release/libremu.so
|
||||
- name: setup staging db
|
||||
run: |
|
||||
echo "CACHEDB=/tmp/pytest-db-ci.db" >> $GITHUB_ENV
|
||||
rm -f /tmp/pytest-db-ci*
|
||||
- name: Run pytest -nauto
|
||||
run: |
|
||||
source /tmp/tinygrad_pytest_ci/bin/activate
|
||||
pytest -nauto --durations=20
|
||||
run: FLOAT16=1 CL=1 IMAGE=2 python3.11 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/720392c9a5b986981fdbed1bb8c47a6c5573a50e/selfdrive/modeld/models/driving_vision.onnx
|
||||
- name: IMAGE=1 openpilot compile3 0.10.1 driving_vision
|
||||
run: FLOAT16=1 CL=1 IMAGE=1 python3.11 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/720392c9a5b986981fdbed1bb8c47a6c5573a50e/selfdrive/modeld/models/driving_vision.onnx
|
||||
|
||||
testmacbenchmark:
|
||||
name: Mac Benchmark
|
||||
@@ -97,7 +66,7 @@ jobs:
|
||||
if: github.repository_owner == 'tinygrad'
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Symlink models and datasets
|
||||
run: |
|
||||
mkdir -p weights
|
||||
@@ -127,17 +96,17 @@ jobs:
|
||||
- name: Run SDXL
|
||||
run: BENCHMARK_LOG=stable_diffusion_xl ASSERT_MIN_STEP_TIME=5000 CAPTURE_PROCESS_REPLAY=0 JIT=1 python3.11 examples/sdxl.py --seed 0 --noshow --timing
|
||||
- name: Run model inference benchmark
|
||||
run: DEV=METAL NOCLANG=1 python3.11 test/external/external_model_benchmark.py
|
||||
run: METAL=1 NOCLANG=1 python3.11 test/external/external_model_benchmark.py
|
||||
- name: Test speed vs torch
|
||||
run: BIG=2 MPS=1 python3.11 test/speed/external_test_speed_v_torch.py
|
||||
- name: Test tensor cores
|
||||
run: DEV=METAL python3.11 test/opt/test_tensor_cores.py
|
||||
run: METAL=1 python3.11 test/opt/test_tensor_cores.py
|
||||
- name: Test AMX tensor cores
|
||||
run: |
|
||||
DEBUG=2 DEV=CPU AMX=1 python3.11 test/opt/test_tensor_cores.py
|
||||
DEBUG=2 DEV=CPU:LLVM AMX=1 python3.11 test/opt/test_tensor_cores.py
|
||||
DEBUG=2 DEV=CPU AMX=1 python3.11 test/opt/test_gen_float4.py TestFloat4.test_float4_multidim_amx TestFloat4.test_float4_multidim_unaligned_load_amx
|
||||
DEBUG=2 DEV=CPU:LLVM AMX=1 python3.11 test/opt/test_gen_float4.py TestFloat4.test_float4_multidim_amx TestFloat4.test_float4_multidim_unaligned_load_amx
|
||||
DEBUG=2 CPU=1 CPU_LLVM=0 AMX=1 python3.11 test/opt/test_tensor_cores.py
|
||||
DEBUG=2 CPU=1 CPU_LLVM=1 AMX=1 python3.11 test/opt/test_tensor_cores.py
|
||||
DEBUG=2 CPU=1 CPU_LLVM=0 AMX=1 python3.11 test/opt/test_gen_float4.py TestFloat4.test_float4_multidim_amx TestFloat4.test_float4_multidim_unaligned_load_amx
|
||||
DEBUG=2 CPU=1 CPU_LLVM=1 AMX=1 python3.11 test/opt/test_gen_float4.py TestFloat4.test_float4_multidim_amx TestFloat4.test_float4_multidim_unaligned_load_amx
|
||||
- name: Run Tensor Core GEMM (float)
|
||||
run: DEBUG=2 SHOULD_USE_TC=1 python3.11 extra/gemm/simple_matmul.py
|
||||
- name: Run Tensor Core GEMM (half)
|
||||
@@ -145,7 +114,7 @@ jobs:
|
||||
- name: Run Tensor Core GEMM (bfloat16)
|
||||
run: DEBUG=2 SHOULD_USE_TC=1 BFLOAT16=1 python3.11 extra/gemm/simple_matmul.py
|
||||
- name: Fuzz Padded Tensor Core GEMM
|
||||
run: DEV=METAL M_START=6 M_STOP=10 M_STEP=1 N_START=6 N_STOP=10 N_STEP=1 K_START=6 K_STOP=24 K_STEP=1 TC_OPT=2 DEBUG=2 python3.11 ./extra/gemm/fuzz_matmul.py
|
||||
run: METAL=1 M_START=6 M_STOP=10 M_STEP=1 N_START=6 N_STOP=10 N_STEP=1 K_START=6 K_STOP=24 K_STEP=1 TC_OPT=2 DEBUG=2 python3.11 ./extra/gemm/fuzz_matmul.py
|
||||
- name: Run LLaMA
|
||||
run: |
|
||||
BENCHMARK_LOG=llama_nojit JIT=0 python3.11 examples/llama.py --gen 1 --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
@@ -187,7 +156,7 @@ jobs:
|
||||
# TODO: too slow
|
||||
# - name: Run 10 CIFAR training steps w winograd
|
||||
# run: BENCHMARK_LOG=cifar_10steps_wino JIT=1 ASSERT_MIN_STEP_TIME=150 WINO=1 STEPS=10 python3.11 examples/hlb_cifar10.py
|
||||
- uses: actions/upload-artifact@v7
|
||||
- uses: actions/upload-artifact@v4
|
||||
with:
|
||||
name: Speed (Mac)
|
||||
path: |
|
||||
@@ -207,7 +176,7 @@ jobs:
|
||||
if: github.repository_owner == 'tinygrad'
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: setup staging db
|
||||
if: github.ref == 'refs/heads/update_benchmark_staging'
|
||||
run: |
|
||||
@@ -218,19 +187,17 @@ jobs:
|
||||
PYTHONPATH=. ./extra/hcq/hcq_smi.py amd kill_pids
|
||||
PYTHONPATH=. ./extra/hcq/hcq_smi.py nv kill_pids
|
||||
- name: UsbGPU boot time
|
||||
run: sudo -E PYTHONPATH=. GMMU=0 DEBUG=2 AM_RESET=1 DEV=USB+AMD time python3.11 test/test_tiny.py TestTiny.test_plus
|
||||
run: sudo -E PYTHONPATH=. DEBUG=2 AM_RESET=1 AMD=1 AMD_IFACE=USB time python3.11 test/test_tiny.py TestTiny.test_plus
|
||||
- name: UsbGPU tiny tests
|
||||
run: sudo -E PYTHONPATH=. GMMU=0 DEV=USB+AMD python3.11 test/test_tiny.py
|
||||
run: sudo -E PYTHONPATH=. AMD=1 AMD_IFACE=USB python3.11 test/test_tiny.py
|
||||
- name: UsbGPU copy speeds
|
||||
run: sudo -E PYTHONPATH=. GMMU=0 DEV=USB+AMD python3.11 test/external/external_test_usb_asm24.py TestDevCopySpeeds
|
||||
run: sudo -E PYTHONPATH=. AMD=1 AMD_IFACE=USB python3.11 test/external/external_test_usb_asm24.py TestDevCopySpeeds
|
||||
#- name: UsbGPU openpilot test
|
||||
# run: sudo -E PYTHONPATH=. GMMU=0 DEV=USB+AMD GRAPH_ONE_KERNEL=1 python3.11 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/9118973ed03c1ae1d40cf69a29507ec2cc78efd7/selfdrive/modeld/models/supercombo.onnx
|
||||
- name: UsbGPU (USB4/TB) install script
|
||||
run: PYTHONPATH=. sh extra/setup_tinygpu_osx.sh
|
||||
# run: sudo -E PYTHONPATH=. AMD=1 AMD_IFACE=USB GRAPH_ONE_KERNEL=1 python3.11 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/9118973ed03c1ae1d40cf69a29507ec2cc78efd7/selfdrive/modeld/models/supercombo.onnx
|
||||
- name: UsbGPU (USB4/TB) boot time
|
||||
run: PYTHONPATH=. DEBUG=3 DEV=PCI+NV:NAK time python3.11 test/test_tiny.py TestTiny.test_plus
|
||||
run: PYTHONPATH=. DEBUG=3 NV=1 NV_IFACE=PCI NV_NAK=1 time python3.11 test/test_tiny.py TestTiny.test_plus
|
||||
- name: UsbGPU (USB4/TB) tiny tests
|
||||
run: PYTHONPATH=. DEV=PCI+NV:NAK python3.11 test/test_tiny.py
|
||||
run: PYTHONPATH=. NV=1 NV_IFACE=PCI NV_NAK=1 python3.11 test/test_tiny.py
|
||||
|
||||
testnvidiabenchmark:
|
||||
name: tinybox green Benchmark
|
||||
@@ -242,7 +209,7 @@ jobs:
|
||||
if: github.repository_owner == 'tinygrad'
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Print nvidia-smi
|
||||
run: nvidia-smi
|
||||
- name: Symlink models and datasets
|
||||
@@ -262,67 +229,67 @@ jobs:
|
||||
- name: reset process replay
|
||||
run: test/external/process_replay/reset.py
|
||||
- name: Run model inference benchmark
|
||||
run: DEV=NV CAPTURE_PROCESS_REPLAY=0 NOCLANG=1 python3 test/external/external_model_benchmark.py
|
||||
run: NV=1 CAPTURE_PROCESS_REPLAY=0 NOCLANG=1 python3 test/external/external_model_benchmark.py
|
||||
- name: Test speed vs torch
|
||||
run: DEV=NV CAPTURE_PROCESS_REPLAY=0 HALF=1 BIG=2 TORCHCUDA=1 python3 test/speed/external_test_speed_v_torch.py
|
||||
run: NV=1 CAPTURE_PROCESS_REPLAY=0 HALF=1 BIG=2 TORCHCUDA=1 python3 test/speed/external_test_speed_v_torch.py
|
||||
- name: Test speed vs theoretical
|
||||
run: DEV=NV IGNORE_BEAM_CACHE=1 CCACHE=0 BEAM_DEBUG=1 DEBUG=1 python -m pytest -rA test/external/speed_v_theoretical.py --durations=20
|
||||
run: NV=1 IGNORE_BEAM_CACHE=1 CCACHE=0 BEAM_DEBUG=1 DEBUG=1 python -m pytest -rA test/external/speed_v_theoretical.py --durations=20
|
||||
- name: Test benchmark allreduce
|
||||
run: DEV=NV python test/external/external_benchmark_multitensor_allreduce.py
|
||||
run: NV=1 python test/external/external_benchmark_multitensor_allreduce.py
|
||||
- name: Test tensor cores
|
||||
run: |
|
||||
DEV=NV ALLOW_TF32=1 python3 test/opt/test_tensor_cores.py
|
||||
DEV=NV:PTX ALLOW_TF32=1 python3 test/opt/test_tensor_cores.py
|
||||
NV=1 ALLOW_TF32=1 python3 test/opt/test_tensor_cores.py
|
||||
NV=1 NV_PTX=1 ALLOW_TF32=1 python3 test/opt/test_tensor_cores.py
|
||||
- name: Run Tensor Core GEMM (CUDA)
|
||||
run: |
|
||||
DEV=CUDA SHOULD_USE_TC=1 HALF=1 DEBUG=2 python3 extra/gemm/simple_matmul.py
|
||||
DEV=CUDA SHOULD_USE_TC=1 BFLOAT16=1 DEBUG=2 python3 extra/gemm/simple_matmul.py
|
||||
DEV=CUDA SHOULD_USE_TC=1 ALLOW_TF32=1 DEBUG=2 ATOL=2e-2 python3 extra/gemm/simple_matmul.py
|
||||
DEV=CUDA SHOULD_USE_TC=1 FP8E4M3=1 DEBUG=2 python3 extra/gemm/simple_matmul.py
|
||||
CUDA=1 SHOULD_USE_TC=1 HALF=1 DEBUG=2 python3 extra/gemm/simple_matmul.py
|
||||
CUDA=1 SHOULD_USE_TC=1 BFLOAT16=1 DEBUG=2 python3 extra/gemm/simple_matmul.py
|
||||
CUDA=1 SHOULD_USE_TC=1 ALLOW_TF32=1 DEBUG=2 ATOL=2e-2 python3 extra/gemm/simple_matmul.py
|
||||
CUDA=1 SHOULD_USE_TC=1 FP8E4M3=1 DEBUG=2 python3 extra/gemm/simple_matmul.py
|
||||
- name: Run Tensor Core GEMM (PTX)
|
||||
run: DEV=NV:PTX SHOULD_USE_TC=1 HALF=1 DEBUG=2 python3 extra/gemm/simple_matmul.py
|
||||
run: NV=1 NV_PTX=1 SHOULD_USE_TC=1 HALF=1 DEBUG=2 python3 extra/gemm/simple_matmul.py
|
||||
- name: Run Tensor Core GEMM (NV)
|
||||
run: DEV=NV SHOULD_USE_TC=1 HALF=1 DEBUG=2 python3 extra/gemm/simple_matmul.py
|
||||
- name: Test DEV=NV
|
||||
run: DEBUG=2 DEV=NV python -m pytest -rA test/test_tiny.py
|
||||
- name: Test DEV=CUDA
|
||||
run: DEBUG=2 DEV=CUDA python -m pytest -rA test/test_tiny.py
|
||||
run: NV=1 SHOULD_USE_TC=1 HALF=1 DEBUG=2 python3 extra/gemm/simple_matmul.py
|
||||
- name: Test NV=1
|
||||
run: DEBUG=2 NV=1 python -m pytest -rA test/test_tiny.py
|
||||
- name: Test CUDA=1
|
||||
run: DEBUG=2 CUDA=1 python -m pytest -rA test/test_tiny.py
|
||||
- name: Run Stable Diffusion
|
||||
run: BENCHMARK_LOG=stable_diffusion DEV=NV python3 examples/stable_diffusion.py --fp16 --seed 0 --noshow --timing
|
||||
run: BENCHMARK_LOG=stable_diffusion NV=1 python3 examples/stable_diffusion.py --fp16 --seed 0 --noshow --timing
|
||||
# TODO: too slow
|
||||
# - name: Run SDXL
|
||||
# run: BENCHMARK_LOG=stable_diffusion_xl ASSERT_MIN_STEP_TIME=2000 CAPTURE_PROCESS_REPLAY=0 DEV=NV CAPTURE_PROCESS_REPLAY=0 python3 examples/sdxl.py --seed 0 --noshow --timing
|
||||
# run: BENCHMARK_LOG=stable_diffusion_xl ASSERT_MIN_STEP_TIME=2000 CAPTURE_PROCESS_REPLAY=0 NV=1 CAPTURE_PROCESS_REPLAY=0 python3 examples/sdxl.py --seed 0 --noshow --timing
|
||||
- name: Run LLaMA
|
||||
run: |
|
||||
BENCHMARK_LOG=llama_nojit DEV=NV JIT=0 python3 examples/llama.py --gen 1 --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
BENCHMARK_LOG=llama DEV=NV JIT=1 python3 examples/llama.py --gen 1 --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
BENCHMARK_LOG=llama_nojit NV=1 JIT=0 python3 examples/llama.py --gen 1 --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
BENCHMARK_LOG=llama NV=1 JIT=1 python3 examples/llama.py --gen 1 --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
- name: Run LLaMA with BEAM
|
||||
run: BENCHMARK_LOG=llama_beam DEV=NV JITBEAM=2 IGNORE_BEAM_CACHE=1 python3 examples/llama.py --gen 1 --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
run: BENCHMARK_LOG=llama_beam NV=1 JITBEAM=2 IGNORE_BEAM_CACHE=1 python3 examples/llama.py --gen 1 --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
# - name: Run LLaMA 7B on 4 GPUs
|
||||
# run: DEV=NV CAPTURE_PROCESS_REPLAY=0 python3 examples/llama.py --gen 1 --size 7B --shard 4 --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
# run: NV=1 CAPTURE_PROCESS_REPLAY=0 python3 examples/llama.py --gen 1 --size 7B --shard 4 --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
# - name: Run LLaMA 7B on 6 GPUs
|
||||
# run: DEV=NV CAPTURE_PROCESS_REPLAY=0 python3 examples/llama.py --gen 1 --size 7B --shard 6 --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
# run: NV=1 CAPTURE_PROCESS_REPLAY=0 python3 examples/llama.py --gen 1 --size 7B --shard 6 --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
- name: Run LLaMA-3 8B BEAM
|
||||
run: BENCHMARK_LOG=llama3_beam DEV=NV JITBEAM=2 IGNORE_BEAM_CACHE=1 python3 examples/llama3.py --size 8B --model weights/LLaMA-3/8B-SF-DPO/ --benchmark --temperature 0
|
||||
run: BENCHMARK_LOG=llama3_beam NV=1 JITBEAM=2 IGNORE_BEAM_CACHE=1 python3 examples/llama3.py --size 8B --model weights/LLaMA-3/8B-SF-DPO/ --benchmark --temperature 0
|
||||
- name: Run LLaMA-3 8B on 4 GPUs with BEAM
|
||||
run: BENCHMARK_LOG=llama3_beam_4gpu DEV=NV JITBEAM=2 IGNORE_BEAM_CACHE=1 CAPTURE_PROCESS_REPLAY=0 python3 examples/llama3.py --size 8B --shard 4 --model weights/LLaMA-3/8B-SF-DPO/ --benchmark --temperature 0
|
||||
run: BENCHMARK_LOG=llama3_beam_4gpu NV=1 JITBEAM=2 IGNORE_BEAM_CACHE=1 CAPTURE_PROCESS_REPLAY=0 python3 examples/llama3.py --size 8B --shard 4 --model weights/LLaMA-3/8B-SF-DPO/ --benchmark --temperature 0
|
||||
- name: Run quantized LLaMA3
|
||||
run: BENCHMARK_LOG=llama3_fp8 python3 examples/llama3.py --size 8B --model weights/LLaMA-3/8B-SF-DPO/ --temperature 0 --benchmark --quantize fp8
|
||||
# - name: Run LLaMA-3 8B on 6 GPUs
|
||||
# run: DEV=NV CAPTURE_PROCESS_REPLAY=0 python3 examples/llama3.py --size 8B --shard 6 --model weights/LLaMA-3/8B-SF-DPO/ --benchmark --temperature 0
|
||||
# run: NV=1 CAPTURE_PROCESS_REPLAY=0 python3 examples/llama3.py --size 8B --shard 6 --model weights/LLaMA-3/8B-SF-DPO/ --benchmark --temperature 0
|
||||
# - name: Run LLaMA-2 70B
|
||||
# run: DEV=NV CAPTURE_PROCESS_REPLAY=0 MAX_CONTEXT=256 python3 examples/llama.py --gen 2 --size 70B --shard 6 --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
# run: NV=1 CAPTURE_PROCESS_REPLAY=0 MAX_CONTEXT=256 python3 examples/llama.py --gen 2 --size 70B --shard 6 --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
- name: Run Mixtral 8x7B
|
||||
run: time BENCHMARK_LOG=mixtral DEV=NV CAPTURE_PROCESS_REPLAY=0 python3 examples/mixtral.py --temperature 0 --count 10 --timing
|
||||
run: time BENCHMARK_LOG=mixtral NV=1 CAPTURE_PROCESS_REPLAY=0 python3 examples/mixtral.py --temperature 0 --count 10 --timing
|
||||
- name: Run GPT2
|
||||
run: |
|
||||
BENCHMARK_LOG=gpt2_nojit DEV=NV JIT=0 python3 examples/gpt2.py --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
BENCHMARK_LOG=gpt2 DEV=NV JIT=1 ASSERT_MIN_STEP_TIME=4 python3 examples/gpt2.py --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
BENCHMARK_LOG=gpt2_nojit NV=1 JIT=0 python3 examples/gpt2.py --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
BENCHMARK_LOG=gpt2 NV=1 JIT=1 ASSERT_MIN_STEP_TIME=4 python3 examples/gpt2.py --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
- name: Run GPT2 w HALF
|
||||
run: BENCHMARK_LOG=gpt2_half DEV=NV HALF=1 ASSERT_MIN_STEP_TIME=6 python3 examples/gpt2.py --count 10 --temperature 0 --timing
|
||||
run: BENCHMARK_LOG=gpt2_half NV=1 HALF=1 ASSERT_MIN_STEP_TIME=6 python3 examples/gpt2.py --count 10 --temperature 0 --timing
|
||||
- name: Run GPT2 w HALF/BEAM
|
||||
run: BENCHMARK_LOG=gpt2_half_beam DEV=NV HALF=1 JITBEAM=2 IGNORE_BEAM_CACHE=1 python3 examples/gpt2.py --count 10 --temperature 0 --timing
|
||||
- uses: actions/upload-artifact@v7
|
||||
run: BENCHMARK_LOG=gpt2_half_beam NV=1 HALF=1 JITBEAM=2 IGNORE_BEAM_CACHE=1 python3 examples/gpt2.py --count 10 --temperature 0 --timing
|
||||
- uses: actions/upload-artifact@v4
|
||||
with:
|
||||
name: Speed (NVIDIA)
|
||||
path: |
|
||||
@@ -340,7 +307,7 @@ jobs:
|
||||
if: github.repository_owner == 'tinygrad'
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Symlink models and datasets
|
||||
run: |
|
||||
mkdir -p weights
|
||||
@@ -360,35 +327,35 @@ jobs:
|
||||
run: test/external/process_replay/reset.py
|
||||
# TODO: too slow
|
||||
# - name: Fuzz Padded Tensor Core GEMM (NV)
|
||||
# run: DEV=NV M_START=12 M_STOP=20 M_STEP=1 N_START=6 N_STOP=10 N_STEP=1 K_START=28 K_STOP=36 K_STEP=1 HALF=1 TC_OPT=2 python3 ./extra/gemm/fuzz_matmul.py
|
||||
# run: NV=1 M_START=12 M_STOP=20 M_STEP=1 N_START=6 N_STOP=10 N_STEP=1 K_START=28 K_STOP=36 K_STEP=1 HALF=1 TC_OPT=2 python3 ./extra/gemm/fuzz_matmul.py
|
||||
# TODO: too slow
|
||||
# - name: Fuzz Padded Tensor Core GEMM (PTX)
|
||||
# run: DEV=NV:PTX M_START=12 M_STOP=20 M_STEP=1 N_START=6 N_STOP=10 N_STEP=1 K_START=28 K_STOP=36 K_STEP=1 HALF=1 TC_OPT=2 python3 ./extra/gemm/fuzz_matmul.py
|
||||
# run: NV=1 NV_PTX=1 M_START=12 M_STOP=20 M_STEP=1 N_START=6 N_STOP=10 N_STEP=1 K_START=28 K_STOP=36 K_STEP=1 HALF=1 TC_OPT=2 python3 ./extra/gemm/fuzz_matmul.py
|
||||
- name: HEVC Decode Benchmark
|
||||
run: VALIDATE=1 MAX_FRAMES=100 ASSERT_FPS=1400 JITBEAM=1 DEV=NV PYTHONPATH=. python3 extra/hevc/decode.py
|
||||
run: VALIDATE=1 MAX_FRAMES=100 ASSERT_FPS=1400 JITBEAM=1 NV=1 PYTHONPATH=. python3 extra/hevc/decode.py
|
||||
- name: Train MNIST
|
||||
run: time PYTHONPATH=. DEV=NV TARGET_EVAL_ACC_PCT=96.0 python3 examples/beautiful_mnist.py
|
||||
run: time PYTHONPATH=. NV=1 TARGET_EVAL_ACC_PCT=96.0 python3 examples/beautiful_mnist.py
|
||||
- name: Run 10 CIFAR training steps
|
||||
run: BENCHMARK_LOG=cifar_10steps ASSERT_MIN_STEP_TIME=120 DEV=NV STEPS=10 python3 examples/hlb_cifar10.py
|
||||
run: BENCHMARK_LOG=cifar_10steps ASSERT_MIN_STEP_TIME=120 NV=1 STEPS=10 python3 examples/hlb_cifar10.py
|
||||
- name: Run 10 CIFAR training steps w HALF
|
||||
run: BENCHMARK_LOG=cifar_10steps_half ASSERT_MIN_STEP_TIME=120 DEV=NV STEPS=10 DEFAULT_FLOAT=HALF python3 examples/hlb_cifar10.py
|
||||
run: BENCHMARK_LOG=cifar_10steps_half ASSERT_MIN_STEP_TIME=120 NV=1 STEPS=10 DEFAULT_FLOAT=HALF python3 examples/hlb_cifar10.py
|
||||
- name: Run 10 CIFAR training steps w BF16
|
||||
run: BENCHMARK_LOG=cifar_10steps_bf16 ASSERT_MIN_STEP_TIME=120 DEV=NV STEPS=10 DEFAULT_FLOAT=BFLOAT16 python3 examples/hlb_cifar10.py
|
||||
run: BENCHMARK_LOG=cifar_10steps_bf16 ASSERT_MIN_STEP_TIME=120 NV=1 STEPS=10 DEFAULT_FLOAT=BFLOAT16 python3 examples/hlb_cifar10.py
|
||||
# - name: Run 10 CIFAR training steps w winograd
|
||||
# run: BENCHMARK_LOG=cifar_10steps_half_wino ASSERT_MIN_STEP_TIME=350 DEV=NV WINO=1 STEPS=10 DEFAULT_FLOAT=HALF python3 examples/hlb_cifar10.py
|
||||
# run: BENCHMARK_LOG=cifar_10steps_half_wino ASSERT_MIN_STEP_TIME=350 NV=1 WINO=1 STEPS=10 DEFAULT_FLOAT=HALF python3 examples/hlb_cifar10.py
|
||||
- name: Run full CIFAR training w 1 GPU
|
||||
run: time BENCHMARK_LOG=cifar DEV=NV DEFAULT_FLOAT=HALF STEPS=1000 TARGET_EVAL_ACC_PCT=93.0 python3 examples/hlb_cifar10.py
|
||||
run: time BENCHMARK_LOG=cifar NV=1 DEFAULT_FLOAT=HALF STEPS=1000 TARGET_EVAL_ACC_PCT=93.0 python3 examples/hlb_cifar10.py
|
||||
- name: Run full CIFAR training steps w 6 GPUS
|
||||
run: time BENCHMARK_LOG=cifar_6gpu CAPTURE_PROCESS_REPLAY=0 DEV=NV DEFAULT_FLOAT=HALF STEPS=350 BS=1536 GPUS=6 TARGET_EVAL_ACC_PCT=93.0 python3 examples/hlb_cifar10.py
|
||||
run: time BENCHMARK_LOG=cifar_6gpu CAPTURE_PROCESS_REPLAY=0 NV=1 DEFAULT_FLOAT=HALF STEPS=350 BS=1536 GPUS=6 TARGET_EVAL_ACC_PCT=93.0 python3 examples/hlb_cifar10.py
|
||||
- name: Run MLPerf resnet eval on training data
|
||||
run: time BENCHMARK_LOG=resnet_eval DEV=NV MODEL=resnet python3 examples/mlperf/model_eval.py
|
||||
run: time BENCHMARK_LOG=resnet_eval NV=1 MODEL=resnet python3 examples/mlperf/model_eval.py
|
||||
- name: Run 10 MLPerf ResNet50 training steps (1 gpu)
|
||||
run: BENCHMARK_LOG=resnet_10steps DEV=NV DEFAULT_FLOAT=HALF BENCHMARK=10 BS=256 GPUS=1 MODEL=resnet python3 examples/mlperf/model_train.py
|
||||
run: BENCHMARK_LOG=resnet_10steps NV=1 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=256 GPUS=1 MODEL=resnet python3 examples/mlperf/model_train.py
|
||||
- name: Run 10 MLPerf ResNet50 training steps (6 gpu)
|
||||
run: BENCHMARK_LOG=resnet_10steps_6gpu DEV=NV CAPTURE_PROCESS_REPLAY=0 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=1536 GPUS=6 MODEL=resnet python3 examples/mlperf/model_train.py
|
||||
run: BENCHMARK_LOG=resnet_10steps_6gpu NV=1 CAPTURE_PROCESS_REPLAY=0 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=1536 GPUS=6 MODEL=resnet python3 examples/mlperf/model_train.py
|
||||
- name: Run 10 MLPerf Bert training steps (6 gpu)
|
||||
# TODO: remove BERT_LAYERS once scheduler is fast
|
||||
run: BENCHMARK_LOG=bert_10steps_6gpu DEV=NV CAPTURE_PROCESS_REPLAY=0 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=72 GPUS=6 BERT_LAYERS=2 MODEL=bert python3 examples/mlperf/model_train.py
|
||||
run: BENCHMARK_LOG=bert_10steps_6gpu NV=1 CAPTURE_PROCESS_REPLAY=0 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=72 GPUS=6 BERT_LAYERS=2 MODEL=bert python3 examples/mlperf/model_train.py
|
||||
- name: Run process replay tests
|
||||
run: cp test/external/process_replay/process_replay.py ./process_replay.py && git fetch origin master && git -c advice.detachedHead=false checkout origin/master && PYTHONPATH=. python3 process_replay.py
|
||||
|
||||
@@ -402,7 +369,7 @@ jobs:
|
||||
if: github.repository_owner == 'tinygrad'
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setcap to python
|
||||
run: ./extra/amdpci/setup_python_cap.sh
|
||||
- name: Remove amd modules
|
||||
@@ -444,18 +411,18 @@ jobs:
|
||||
# python3 -c "import torch; print(torch.__version__)"
|
||||
# LD_PRELOAD="/opt/rocm/lib/libhsa-runtime64.so" HSA=1 BIG=2 TORCHCUDA=1 python3 test/speed/external_test_speed_v_torch.py
|
||||
- name: Test speed vs theoretical
|
||||
run: DEV=AMD IGNORE_BEAM_CACHE=1 CCACHE=0 BEAM_DEBUG=1 DEBUG=1 python -m pytest -rA test/external/speed_v_theoretical.py --durations=20
|
||||
- name: Test tensor cores (no LLVM)
|
||||
run: DEV=AMD python3 test/opt/test_tensor_cores.py
|
||||
run: AMD=1 IGNORE_BEAM_CACHE=1 CCACHE=0 BEAM_DEBUG=1 DEBUG=1 python -m pytest -rA test/external/speed_v_theoretical.py --durations=20
|
||||
- name: Test tensor cores AMD_LLVM=0
|
||||
run: AMD=1 AMD_LLVM=0 python3 test/opt/test_tensor_cores.py
|
||||
# TODO: this is flaky
|
||||
# - name: Test tensor cores AMD:LLVM
|
||||
# run: DEV=AMD:LLVM python3 test/opt/test_tensor_cores.py
|
||||
# - name: Test tensor cores AMD_LLVM=1
|
||||
# run: AMD=1 AMD_LLVM=1 python3 test/opt/test_tensor_cores.py
|
||||
- name: Run Tensor Core GEMM (AMD)
|
||||
run: |
|
||||
DEV=AMD SHOULD_USE_TC=1 BFLOAT16=1 DEBUG=2 python3 extra/gemm/simple_matmul.py
|
||||
DEV=AMD SHOULD_USE_TC=1 HALF=1 DEBUG=2 ATOL=2e-2 python3 extra/gemm/simple_matmul.py
|
||||
- name: Test DEV=AMD
|
||||
run: DEBUG=2 DEV=AMD python -m pytest -rA test/test_tiny.py
|
||||
AMD=1 SHOULD_USE_TC=1 BFLOAT16=1 DEBUG=2 python3 extra/gemm/simple_matmul.py
|
||||
AMD=1 SHOULD_USE_TC=1 HALF=1 DEBUG=2 ATOL=2e-2 python3 extra/gemm/simple_matmul.py
|
||||
- name: Test AMD=1
|
||||
run: DEBUG=2 AMD=1 python -m pytest -rA test/test_tiny.py
|
||||
#- name: Test HIP=1
|
||||
# run: DEBUG=2 HIP=1 python -m pytest -rA test/test_tiny.py
|
||||
# TODO: AMD compiler bug causes this to fail
|
||||
@@ -464,43 +431,43 @@ jobs:
|
||||
#- name: Remove amdgpu
|
||||
# run: sleep 10 && sudo rmmod amdgpu # sleep a bit to let the driver unload the prev pid.
|
||||
- name: Test AM cold start time
|
||||
run: time DEV=AMD AM_RESET=1 python3 test/test_tiny.py TestTiny.test_plus
|
||||
run: time AMD=1 AM_RESET=1 python3 test/test_tiny.py TestTiny.test_plus
|
||||
- name: Test AM warm start time
|
||||
run: time DEV=AMD python3 test/test_tiny.py TestTiny.test_plus
|
||||
run: time AMD=1 python3 test/test_tiny.py TestTiny.test_plus
|
||||
- name: Run Stable Diffusion
|
||||
run: BENCHMARK_LOG=stable_diffusion ASSERT_MIN_STEP_TIME=550 DEV=AMD python3 examples/stable_diffusion.py --fp16 --seed 0 --noshow --timing
|
||||
run: BENCHMARK_LOG=stable_diffusion ASSERT_MIN_STEP_TIME=550 AMD=1 python3 examples/stable_diffusion.py --fp16 --seed 0 --noshow --timing
|
||||
- name: Run SDXL
|
||||
run: BENCHMARK_LOG=stable_diffusion_xl ASSERT_MIN_STEP_TIME=3200 CAPTURE_PROCESS_REPLAY=0 DEV=AMD python3 examples/sdxl.py --seed 0 --noshow --timing
|
||||
run: BENCHMARK_LOG=stable_diffusion_xl ASSERT_MIN_STEP_TIME=3200 CAPTURE_PROCESS_REPLAY=0 AMD=1 python3 examples/sdxl.py --seed 0 --noshow --timing
|
||||
- name: Run LLaMA 7B
|
||||
run: |
|
||||
BENCHMARK_LOG=llama_nojit DEV=AMD JIT=0 python3 examples/llama.py --gen 1 --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
BENCHMARK_LOG=llama DEV=AMD JIT=1 python3 examples/llama.py --gen 1 --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
BENCHMARK_LOG=llama_nojit AMD=1 JIT=0 python3 examples/llama.py --gen 1 --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
BENCHMARK_LOG=llama AMD=1 JIT=1 python3 examples/llama.py --gen 1 --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
- name: Run LLaMA 7B with BEAM
|
||||
run: BENCHMARK_LOG=llama_beam DEV=AMD JITBEAM=2 IGNORE_BEAM_CACHE=1 python3 examples/llama.py --gen 1 --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
run: BENCHMARK_LOG=llama_beam AMD=1 JITBEAM=2 IGNORE_BEAM_CACHE=1 python3 examples/llama.py --gen 1 --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
# - name: Run LLaMA 7B on 4 GPUs
|
||||
# run: DEV=AMD CAPTURE_PROCESS_REPLAY=0 python3 examples/llama.py --gen 1 --size 7B --shard 4 --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
# run: AMD=1 CAPTURE_PROCESS_REPLAY=0 python3 examples/llama.py --gen 1 --size 7B --shard 4 --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
# - name: Run LLaMA 7B on 6 GPUs
|
||||
# run: DEV=AMD CAPTURE_PROCESS_REPLAY=0 python3 examples/llama.py --gen 1 --size 7B --shard 6 --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
# run: AMD=1 CAPTURE_PROCESS_REPLAY=0 python3 examples/llama.py --gen 1 --size 7B --shard 6 --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
- name: Run LLaMA-3 8B BEAM
|
||||
run: BENCHMARK_LOG=llama3_beam DEV=AMD JITBEAM=2 IGNORE_BEAM_CACHE=1 python3 examples/llama3.py --size 8B --model weights/LLaMA-3/8B-SF-DPO/ --benchmark --temperature 0
|
||||
run: BENCHMARK_LOG=llama3_beam AMD=1 JITBEAM=2 IGNORE_BEAM_CACHE=1 python3 examples/llama3.py --size 8B --model weights/LLaMA-3/8B-SF-DPO/ --benchmark --temperature 0
|
||||
- name: Run LLaMA-3 8B on 4 GPUs with BEAM
|
||||
run: BENCHMARK_LOG=llama3_beam_4gpu DEV=AMD JITBEAM=2 IGNORE_BEAM_CACHE=1 CAPTURE_PROCESS_REPLAY=0 python3 examples/llama3.py --size 8B --shard 4 --model weights/LLaMA-3/8B-SF-DPO/ --benchmark --temperature 0
|
||||
run: BENCHMARK_LOG=llama3_beam_4gpu AMD=1 JITBEAM=2 IGNORE_BEAM_CACHE=1 CAPTURE_PROCESS_REPLAY=0 python3 examples/llama3.py --size 8B --shard 4 --model weights/LLaMA-3/8B-SF-DPO/ --benchmark --temperature 0
|
||||
# - name: Run LLaMA-3 8B on 6 GPUs
|
||||
# run: DEV=AMD CAPTURE_PROCESS_REPLAY=0 python3 examples/llama3.py --size 8B --shard 6 --model weights/LLaMA-3/8B-SF-DPO/ --benchmark --temperature 0
|
||||
# run: AMD=1 CAPTURE_PROCESS_REPLAY=0 python3 examples/llama3.py --size 8B --shard 6 --model weights/LLaMA-3/8B-SF-DPO/ --benchmark --temperature 0
|
||||
#- name: Restore amdgpu
|
||||
# run: sudo modprobe amdgpu
|
||||
# - name: Run LLaMA-2 70B
|
||||
# run: DEV=AMD CAPTURE_PROCESS_REPLAY=0 python3 examples/llama.py --gen 2 --size 70B --shard 6 --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
# run: AMD=1 CAPTURE_PROCESS_REPLAY=0 python3 examples/llama.py --gen 2 --size 70B --shard 6 --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
- name: Run Mixtral 8x7B
|
||||
run: time BENCHMARK_LOG=mixtral DEV=AMD python3 examples/mixtral.py --temperature 0 --count 10 --timing
|
||||
run: time BENCHMARK_LOG=mixtral AMD=1 python3 examples/mixtral.py --temperature 0 --count 10 --timing
|
||||
- name: Run GPT2
|
||||
run: |
|
||||
BENCHMARK_LOG=gpt2_nojit DEV=AMD JIT=0 python3 examples/gpt2.py --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
BENCHMARK_LOG=gpt2 DEV=AMD JIT=1 ASSERT_MIN_STEP_TIME=5 python3 examples/gpt2.py --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
BENCHMARK_LOG=gpt2_nojit AMD=1 JIT=0 python3 examples/gpt2.py --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
BENCHMARK_LOG=gpt2 AMD=1 JIT=1 ASSERT_MIN_STEP_TIME=5 python3 examples/gpt2.py --prompt "Hello." --count 10 --temperature 0 --timing
|
||||
- name: Run GPT2 w HALF
|
||||
run: BENCHMARK_LOG=gpt2_half DEV=AMD HALF=1 ASSERT_MIN_STEP_TIME=5 python3 examples/gpt2.py --count 10 --temperature 0 --timing
|
||||
run: BENCHMARK_LOG=gpt2_half AMD=1 HALF=1 ASSERT_MIN_STEP_TIME=5 python3 examples/gpt2.py --count 10 --temperature 0 --timing
|
||||
- name: Run GPT2 w HALF/BEAM
|
||||
run: BENCHMARK_LOG=gpt2_half_beam DEV=AMD HALF=1 JITBEAM=2 IGNORE_BEAM_CACHE=1 python3 examples/gpt2.py --count 10 --temperature 0 --timing
|
||||
run: BENCHMARK_LOG=gpt2_half_beam AMD=1 HALF=1 JITBEAM=2 IGNORE_BEAM_CACHE=1 python3 examples/gpt2.py --count 10 --temperature 0 --timing
|
||||
- name: Run process replay tests
|
||||
run: cp test/external/process_replay/process_replay.py ./process_replay.py && git fetch origin master && git -c advice.detachedHead=false checkout origin/master && PYTHONPATH=. python3 process_replay.py
|
||||
|
||||
@@ -514,7 +481,7 @@ jobs:
|
||||
if: github.repository_owner == 'tinygrad'
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setcap to python
|
||||
run: ./extra/amdpci/setup_python_cap.sh
|
||||
- name: Remove amd modules
|
||||
@@ -539,20 +506,20 @@ jobs:
|
||||
- name: reset process replay
|
||||
run: test/external/process_replay/reset.py
|
||||
- name: Train MNIST
|
||||
run: time PYTHONPATH=. DEV=AMD TARGET_EVAL_ACC_PCT=96.0 python3 examples/beautiful_mnist.py
|
||||
run: time PYTHONPATH=. AMD=1 TARGET_EVAL_ACC_PCT=96.0 python3 examples/beautiful_mnist.py
|
||||
- name: Run 10 CIFAR training steps
|
||||
run: BENCHMARK_LOG=cifar_10steps ASSERT_MIN_STEP_TIME=200 DEV=AMD STEPS=10 python3 examples/hlb_cifar10.py
|
||||
run: BENCHMARK_LOG=cifar_10steps ASSERT_MIN_STEP_TIME=200 AMD=1 STEPS=10 python3 examples/hlb_cifar10.py
|
||||
- name: Run 10 CIFAR training steps w HALF
|
||||
run: BENCHMARK_LOG=cifar_10steps_half ASSERT_MIN_STEP_TIME=230 DEV=AMD STEPS=10 DEFAULT_FLOAT=HALF python3 examples/hlb_cifar10.py
|
||||
run: BENCHMARK_LOG=cifar_10steps_half ASSERT_MIN_STEP_TIME=230 AMD=1 STEPS=10 DEFAULT_FLOAT=HALF python3 examples/hlb_cifar10.py
|
||||
# - name: Run 10 CIFAR training steps w BF16
|
||||
# run: BENCHMARK_LOG=cifar_10steps_bf16 ASSERT_MIN_STEP_TIME=288 DEV=AMD STEPS=10 DEFAULT_FLOAT=BFLOAT16 python3 examples/hlb_cifar10.py
|
||||
# run: BENCHMARK_LOG=cifar_10steps_bf16 ASSERT_MIN_STEP_TIME=288 AMD=1 STEPS=10 DEFAULT_FLOAT=BFLOAT16 python3 examples/hlb_cifar10.py
|
||||
# TODO: too slow
|
||||
# - name: Run 10 CIFAR training steps w winograd
|
||||
# run: BENCHMARK_LOG=cifar_10steps_half_wino ASSERT_MIN_STEP_TIME=66 DEV=AMD WINO=1 STEPS=10 DEFAULT_FLOAT=HALF python3 examples/hlb_cifar10.py
|
||||
# run: BENCHMARK_LOG=cifar_10steps_half_wino ASSERT_MIN_STEP_TIME=66 AMD=1 WINO=1 STEPS=10 DEFAULT_FLOAT=HALF python3 examples/hlb_cifar10.py
|
||||
- name: Run full CIFAR training w 1 GPU
|
||||
run: time BENCHMARK_LOG=cifar DEV=AMD DEFAULT_FLOAT=HALF STEPS=1000 TARGET_EVAL_ACC_PCT=93.0 python3 examples/hlb_cifar10.py
|
||||
run: time BENCHMARK_LOG=cifar AMD=1 DEFAULT_FLOAT=HALF STEPS=1000 TARGET_EVAL_ACC_PCT=93.0 python3 examples/hlb_cifar10.py
|
||||
- name: Run full CIFAR training steps w 6 GPUS
|
||||
run: time BENCHMARK_LOG=cifar_6gpu DEV=AMD DEFAULT_FLOAT=HALF STEPS=350 BS=1536 GPUS=6 TARGET_EVAL_ACC_PCT=93.0 python3 examples/hlb_cifar10.py
|
||||
run: time BENCHMARK_LOG=cifar_6gpu AMD=1 DEFAULT_FLOAT=HALF STEPS=350 BS=1536 GPUS=6 TARGET_EVAL_ACC_PCT=93.0 python3 examples/hlb_cifar10.py
|
||||
# TODO: broken on some of the machines
|
||||
#- name: Test full tinyfs load
|
||||
# run: TINYFS_ENDPOINT=10.0.52.11:6767 PYTHONPATH=. python extra/tinyfs/fetch_file.py --hash d734f5e3be9f1e9d863bfaa4fc6c1ef2 --len 175866113 --dest mapping.json --check
|
||||
@@ -569,7 +536,7 @@ jobs:
|
||||
if: github.repository_owner == 'tinygrad'
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setcap to python
|
||||
run: ./extra/amdpci/setup_python_cap.sh
|
||||
- name: Remove amd modules
|
||||
@@ -594,14 +561,14 @@ jobs:
|
||||
- name: reset process replay
|
||||
run: test/external/process_replay/reset.py
|
||||
- name: Run MLPerf resnet eval
|
||||
run: time BENCHMARK_LOG=resnet_eval DEV=AMD MODEL=resnet python3 examples/mlperf/model_eval.py
|
||||
run: time BENCHMARK_LOG=resnet_eval AMD=1 MODEL=resnet python3 examples/mlperf/model_eval.py
|
||||
- name: Run 10 MLPerf ResNet50 training steps (1 gpu)
|
||||
run: BENCHMARK_LOG=resnet_10steps DEV=AMD DEFAULT_FLOAT=HALF BENCHMARK=10 BS=256 GPUS=1 MODEL=resnet python3 examples/mlperf/model_train.py
|
||||
run: BENCHMARK_LOG=resnet_10steps AMD=1 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=256 GPUS=1 MODEL=resnet python3 examples/mlperf/model_train.py
|
||||
- name: Run 10 MLPerf ResNet50 training steps (6 gpu)
|
||||
run: BENCHMARK_LOG=resnet_10steps_6gpu DEV=AMD CAPTURE_PROCESS_REPLAY=0 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=1536 GPUS=6 MODEL=resnet python3 examples/mlperf/model_train.py
|
||||
run: BENCHMARK_LOG=resnet_10steps_6gpu AMD=1 CAPTURE_PROCESS_REPLAY=0 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=1536 GPUS=6 MODEL=resnet python3 examples/mlperf/model_train.py
|
||||
- name: Run 10 MLPerf Bert training steps (6 gpu)
|
||||
# TODO: remove BERT_LAYERS once scheduler is fast
|
||||
run: BENCHMARK_LOG=bert_10steps_6gpu DEV=AMD CAPTURE_PROCESS_REPLAY=0 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=72 GPUS=6 BERT_LAYERS=2 MODEL=bert python3 examples/mlperf/model_train.py
|
||||
run: BENCHMARK_LOG=bert_10steps_6gpu AMD=1 CAPTURE_PROCESS_REPLAY=0 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=72 GPUS=6 BERT_LAYERS=2 MODEL=bert python3 examples/mlperf/model_train.py
|
||||
- name: Run process replay tests
|
||||
run: cp test/external/process_replay/process_replay.py ./process_replay.py && git fetch origin master && git -c advice.detachedHead=false checkout origin/master && PYTHONPATH=. python3 process_replay.py
|
||||
|
||||
@@ -615,7 +582,7 @@ jobs:
|
||||
if: github.repository_owner == 'tinygrad'
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: setup staging db
|
||||
if: github.ref == 'refs/heads/update_benchmark_staging'
|
||||
run: |
|
||||
@@ -623,30 +590,30 @@ jobs:
|
||||
rm -f /tmp/staging.db /tmp/staging.db-shm /tmp/staging.db-wal
|
||||
- name: reset process replay
|
||||
run: test/external/process_replay/reset.py
|
||||
- name: openpilot compile3 0.11.0 driving_vision
|
||||
run: BENCHMARK_LOG=openpilot_0_11_0_vision PYTHONPATH="." ASSERT_MIN_STEP_TIME=17 DEV=QCOM FLOAT16=1 IMAGE=1 NOLOCALS=1 taskset -c 4-7 python3 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/v0.11.0/selfdrive/modeld/models/driving_vision.onnx
|
||||
- name: IR3 openpilot compile3 0.11.0 driving_vision
|
||||
run: BENCHMARK_LOG=ir3_openpilot_0_11_0_vision PYTHONPATH="." ASSERT_MIN_STEP_TIME=17 DEV=QCOM:IR3 FLOAT16=1 IMAGE=1 NOLOCALS=1 taskset -c 4-7 python3 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/v0.11.0/selfdrive/modeld/models/driving_vision.onnx
|
||||
- name: openpilot compile3 0.11.0 driving_policy
|
||||
run: BENCHMARK_LOG=openpilot_0_11_0_policy PYTHONPATH="." ASSERT_MIN_STEP_TIME=3 DEV=QCOM FLOAT16=1 IMAGE=1 NOLOCALS=1 taskset -c 4-7 python3 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/v0.11.0/selfdrive/modeld/models/driving_policy.onnx
|
||||
- name: openpilot compile3 0.11.0 dmonitoring
|
||||
run: BENCHMARK_LOG=openpilot_0_11_0_dmonitoring PYTHONPATH="." ASSERT_MIN_STEP_TIME=11 DEV=QCOM FLOAT16=1 IMAGE=1 NOLOCALS=1 taskset -c 4-7 python3 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/v0.11.0/selfdrive/modeld/models/dmonitoring_model.onnx
|
||||
- name: openpilot compile3 0.10.0 driving_policy
|
||||
run: BENCHMARK_LOG=openpilot_0_10_0_policy PYTHONPATH="." ASSERT_MIN_STEP_TIME=3 DEV=QCOM FLOAT16=1 IMAGE=2 NOLOCALS=1 taskset -c 4-7 python3 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/v0.10.0/selfdrive/modeld/models/driving_policy.onnx
|
||||
- name: openpilot compile3 0.10.0 dmonitoring
|
||||
run: BENCHMARK_LOG=openpilot_0_10_0_dmonitoring PYTHONPATH="." ASSERT_MIN_STEP_TIME=11 DEV=QCOM FLOAT16=1 IMAGE=2 NOLOCALS=1 taskset -c 4-7 python3 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/v0.10.0/selfdrive/modeld/models/dmonitoring_model.onnx
|
||||
- name: DEBUG=2 openpilot compile3 0.10.1 driving_vision
|
||||
run: PYTHONPATH="." DEBUG=2 DEV=QCOM FLOAT16=1 IMAGE=2 NOLOCALS=1 taskset -c 4-7 python3 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/720392c9a5b986981fdbed1bb8c47a6c5573a50e/selfdrive/modeld/models/driving_vision.onnx
|
||||
- name: DEBUG=2 IMAGE=1 openpilot compile3 0.10.1 driving_vision
|
||||
run: PYTHONPATH="." DEBUG=2 DEV=QCOM FLOAT16=1 IMAGE=1 NOLOCALS=1 taskset -c 4-7 python3 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/720392c9a5b986981fdbed1bb8c47a6c5573a50e/selfdrive/modeld/models/driving_vision.onnx
|
||||
- name: IMAGE=1 openpilot compile3 0.10.1 driving_vision
|
||||
run: BENCHMARK_LOG=image_1_openpilot_0_10_1_vision PYTHONPATH="." DEV=QCOM FLOAT16=1 IMAGE=1 NOLOCALS=1 taskset -c 4-7 python3 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/720392c9a5b986981fdbed1bb8c47a6c5573a50e/selfdrive/modeld/models/driving_vision.onnx
|
||||
- name: openpilot compile3 0.10.1 driving_vision
|
||||
run: BENCHMARK_LOG=openpilot_0_10_1_vision PYTHONPATH="." ASSERT_MIN_STEP_TIME=17 DEV=QCOM FLOAT16=1 IMAGE=1 NOLOCALS=1 taskset -c 4-7 python3 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/720392c9a5b986981fdbed1bb8c47a6c5573a50e/selfdrive/modeld/models/driving_vision.onnx
|
||||
run: BENCHMARK_LOG=openpilot_0_10_1_vision PYTHONPATH="." ASSERT_MIN_STEP_TIME=17 DEV=QCOM FLOAT16=1 IMAGE=2 NOLOCALS=1 taskset -c 4-7 python3 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/720392c9a5b986981fdbed1bb8c47a6c5573a50e/selfdrive/modeld/models/driving_vision.onnx
|
||||
- name: openpilot compile3 0.10.1 driving_policy
|
||||
run: BENCHMARK_LOG=openpilot_0_10_1_policy PYTHONPATH="." ASSERT_MIN_STEP_TIME=3 DEV=QCOM FLOAT16=1 IMAGE=1 NOLOCALS=1 taskset -c 4-7 python3 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/720392c9a5b986981fdbed1bb8c47a6c5573a50e/selfdrive/modeld/models/driving_policy.onnx
|
||||
run: BENCHMARK_LOG=openpilot_0_10_1_policy PYTHONPATH="." ASSERT_MIN_STEP_TIME=3 DEV=QCOM FLOAT16=1 IMAGE=2 NOLOCALS=1 taskset -c 4-7 python3 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/720392c9a5b986981fdbed1bb8c47a6c5573a50e/selfdrive/modeld/models/driving_policy.onnx
|
||||
- name: openpilot compile3 0.10.1 dmonitoring
|
||||
run: BENCHMARK_LOG=openpilot_0_10_1_dmonitoring PYTHONPATH="." ASSERT_MIN_STEP_TIME=11 DEV=QCOM FLOAT16=1 IMAGE=1 NOLOCALS=1 taskset -c 4-7 python3 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/720392c9a5b986981fdbed1bb8c47a6c5573a50e/selfdrive/modeld/models/dmonitoring_model.onnx
|
||||
run: BENCHMARK_LOG=openpilot_0_10_1_dmonitoring PYTHONPATH="." ASSERT_MIN_STEP_TIME=11 DEV=QCOM FLOAT16=1 IMAGE=2 NOLOCALS=1 taskset -c 4-7 python3 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/720392c9a5b986981fdbed1bb8c47a6c5573a50e/selfdrive/modeld/models/dmonitoring_model.onnx
|
||||
- name: benchmark MobileNetV2 on DSP
|
||||
run: |
|
||||
# generate quantized weights
|
||||
ln -s /data/home/tiny/tinygrad/extra/datasets/imagenet extra/datasets/imagenet
|
||||
ln -s /data/home/tiny/tinygrad/testsig-*.so .
|
||||
PYTHONPATH=. CC=clang-19 DEV=CPU QUANT=1 CNT=0 python3 examples/test_onnx_imagenet.py https://github.com/xamcat/mobcat-samples/raw/refs/heads/master/onnx_runtime/InferencingSample/InferencingSample/mobilenetv2-7.onnx /tmp/model.quant.onnx
|
||||
PYTHONPATH=. CC=clang-19 CPU=1 CPU_LLVM=0 QUANT=1 CNT=0 python3 examples/test_onnx_imagenet.py https://github.com/xamcat/mobcat-samples/raw/refs/heads/master/onnx_runtime/InferencingSample/InferencingSample/mobilenetv2-7.onnx /tmp/model.quant.onnx
|
||||
# benchmark on DSP with NOOPT=1, the devectorizer has issues
|
||||
PYTHONPATH=. CC=clang-19 DEV=DSP NOOPT=1 CNT=2 DEBUG=2 python3 examples/test_onnx_imagenet.py /tmp/model.quant.onnx
|
||||
PYTHONPATH=. CC=clang-19 DSP=1 NOOPT=1 CNT=2 DEBUG=2 python3 examples/test_onnx_imagenet.py /tmp/model.quant.onnx
|
||||
- name: Run process replay tests
|
||||
run: cp test/external/process_replay/process_replay.py ./process_replay.py && git fetch origin master && git -c advice.detachedHead=false checkout origin/master && PYTHONPATH=. python3 process_replay.py
|
||||
|
||||
@@ -660,16 +627,16 @@ jobs:
|
||||
if: github.repository_owner == 'tinygrad'
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: setup staging db
|
||||
if: github.ref == 'refs/heads/update_benchmark_staging'
|
||||
run: |
|
||||
echo "CACHEDB=/tmp/staging.db" >> $GITHUB_ENV
|
||||
rm -f /tmp/staging.db /tmp/staging.db-shm /tmp/staging.db-wal
|
||||
- name: openpilot compile3 0.10.1 driving_vision
|
||||
run: BENCHMARK_LOG=usbgpu_openpilot_0_10_1_vision PYTHONPATH="." GMMU=0 DEV=USB+AMD:LLVM ASSERT_MIN_STEP_TIME=50 python3 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/720392c9a5b986981fdbed1bb8c47a6c5573a50e/selfdrive/modeld/models/driving_vision.onnx
|
||||
run: BENCHMARK_LOG=usbgpu_openpilot_0_10_1_vision PYTHONPATH="." DEV=AMD AMD_LLVM=1 AMD_IFACE=USB ASSERT_MIN_STEP_TIME=50 python3 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/720392c9a5b986981fdbed1bb8c47a6c5573a50e/selfdrive/modeld/models/driving_vision.onnx
|
||||
- name: openpilot load_pickle 0.10.1 driving_vision
|
||||
run: BENCHMARK_LOG=usbgpu_openpilot_0_10_1_vision_load_pickle PYTHONPATH="." GMMU=0 DEV=USB+AMD ASSERT_MIN_LOAD_TIME=15 python3 examples/openpilot/load_pickle.py
|
||||
run: BENCHMARK_LOG=usbgpu_openpilot_0_10_1_vision_load_pickle PYTHONPATH="." DEV=AMD AMD_IFACE=USB ASSERT_MIN_LOAD_TIME=15 python3 examples/openpilot/load_pickle.py
|
||||
|
||||
testreddriverbenchmark:
|
||||
name: AM Benchmark
|
||||
@@ -681,7 +648,7 @@ jobs:
|
||||
if: github.repository_owner == 'tinygrad'
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setcap to python
|
||||
run: ./extra/amdpci/setup_python_cap.sh
|
||||
- name: Remove amd modules
|
||||
@@ -706,40 +673,32 @@ jobs:
|
||||
- name: reset process replay
|
||||
run: test/external/process_replay/reset.py
|
||||
- name: Test driver cold start time
|
||||
run: time DEBUG=3 DEV=AMD AM_RESET=1 python3 test/test_tiny.py TestTiny.test_plus
|
||||
run: time DEBUG=3 AMD=1 AM_RESET=1 python3 test/test_tiny.py TestTiny.test_plus
|
||||
- name: Test driver warm start time
|
||||
run: time DEBUG=3 DEV=AMD python3 test/test_tiny.py TestTiny.test_plus
|
||||
run: time DEBUG=3 AMD=1 python3 test/test_tiny.py TestTiny.test_plus
|
||||
# Fails on 9070
|
||||
# - name: Test tensor cores
|
||||
# run: |
|
||||
# DEV=AMD python3 test/test_linearizer.py test/opt/test_tensor_cores.py
|
||||
# DEV=AMD:LLVM python3 test/test_linearizer.py test/opt/test_tensor_cores.py
|
||||
# DEV=AMD SHOULD_USE_TC=1 BFLOAT16=1 DEBUG=2 python3 extra/gemm/simple_matmul.py
|
||||
# AMD=1 AMD_LLVM=0 python3 test/test_linearizer.py test/opt/test_tensor_cores.py
|
||||
# AMD=1 AMD_LLVM=1 python3 test/test_linearizer.py test/opt/test_tensor_cores.py
|
||||
# AMD=1 SHOULD_USE_TC=1 BFLOAT16=1 DEBUG=2 python3 extra/gemm/simple_matmul.py
|
||||
- name: Run Tensor Core GEMM (AMD)
|
||||
run: DEV=AMD SHOULD_USE_TC=1 HALF=1 DEBUG=2 ATOL=2e-2 python3 extra/gemm/simple_matmul.py
|
||||
- name: Test DEV=AMD
|
||||
run: DEBUG=2 DEV=AMD python -m pytest -rA test/test_tiny.py
|
||||
run: AMD=1 SHOULD_USE_TC=1 HALF=1 DEBUG=2 ATOL=2e-2 python3 extra/gemm/simple_matmul.py
|
||||
- name: Test AMD=1
|
||||
run: DEBUG=2 AMD=1 python -m pytest -rA test/test_tiny.py
|
||||
- name: Test DISK copy time
|
||||
run: DEV=AMD TESTFILE=/raid/downloads/llama3-8b-sfr/model-00001-of-00004.safetensors python3 test/external/external_benchmark_disk_raw.py
|
||||
run: AMD=1 TESTFILE=/raid/downloads/llama3-8b-sfr/model-00001-of-00004.safetensors python3 test/external/external_benchmark_disk_raw.py
|
||||
- name: Test CPU copy time
|
||||
run: |
|
||||
DEV=AMD GRAPH_ONE_KERNEL=1 PYTHONPATH=. NSZ=8192 python3 test/speed/external_test_copy_speed.py TestCopySpeed.testCopyDefaulttoCPUJit
|
||||
DEV=AMD GRAPH_ONE_KERNEL=1 PYTHONPATH=. NSZ=8192 python3 test/speed/external_test_copy_speed.py TestCopySpeed.testCopyCPUtoDefaultJit
|
||||
AMD=1 GRAPH_ONE_KERNEL=1 PYTHONPATH=. NSZ=8192 python3 test/speed/external_test_copy_speed.py TestCopySpeed.testCopyDefaulttoCPUJit
|
||||
AMD=1 GRAPH_ONE_KERNEL=1 PYTHONPATH=. NSZ=8192 python3 test/speed/external_test_copy_speed.py TestCopySpeed.testCopyCPUtoDefaultJit
|
||||
- name: Run full CIFAR training w 1 GPU
|
||||
run: time BENCHMARK_LOG=cifar DEV=AMD DEFAULT_FLOAT=HALF STEPS=1000 TARGET_EVAL_ACC_PCT=93.0 python3 examples/hlb_cifar10.py
|
||||
run: time BENCHMARK_LOG=cifar AMD=1 DEFAULT_FLOAT=HALF STEPS=1000 TARGET_EVAL_ACC_PCT=93.0 python3 examples/hlb_cifar10.py
|
||||
# - name: Run 10 MLPerf ResNet50 training steps (1 gpu)
|
||||
# run: BENCHMARK_LOG=resnet_10steps DEV=AMD MNISTMOCK=1 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=256 GPUS=1 MODEL=resnet python3 examples/mlperf/model_train.py
|
||||
# run: BENCHMARK_LOG=resnet_10steps AMD=1 MNISTMOCK=1 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=256 GPUS=1 MODEL=resnet python3 examples/mlperf/model_train.py
|
||||
- name: Run 10 MLPerf Bert training steps (1 gpu)
|
||||
# TODO: remove BERT_LAYERS once scheduler is fast
|
||||
run: BENCHMARK_LOG=bert_10steps DEV=AMD CAPTURE_PROCESS_REPLAY=0 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=66 GPUS=1 BERT_LAYERS=2 MODEL=bert python3 examples/mlperf/model_train.py
|
||||
- name: Remote
|
||||
run: |
|
||||
pkill -f 'extra/remote/serve.py' || true
|
||||
PYTHONPATH=. python3 extra/remote/serve.py 6482 &
|
||||
sleep 1
|
||||
DEBUG=2 PYTHONPATH=. REMOTE=127.0.0.1:6482 AM_RESET=1 DEV=PCI+AMD python3 test/test_tiny.py
|
||||
DEBUG=2 PYTHONPATH=. REMOTE=127.0.0.1:6482 AM_RESET=1 DEV=PCI+AMD AMD_AQL=1 python3 test/test_tiny.py
|
||||
pkill -f 'extra/remote/serve.py' || true
|
||||
run: BENCHMARK_LOG=bert_10steps AMD=1 CAPTURE_PROCESS_REPLAY=0 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=66 GPUS=1 BERT_LAYERS=2 MODEL=bert python3 examples/mlperf/model_train.py
|
||||
- name: Run process replay tests
|
||||
run: cp test/external/process_replay/process_replay.py ./process_replay.py && git fetch origin master && git -c advice.detachedHead=false checkout origin/master && PYTHONPATH=. python3 process_replay.py
|
||||
|
||||
@@ -753,7 +712,7 @@ jobs:
|
||||
if: github.repository_owner == 'tinygrad'
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setcap to python
|
||||
run: ./extra/amdpci/setup_python_cap.sh
|
||||
- name: Remove nv modules
|
||||
@@ -778,30 +737,23 @@ jobs:
|
||||
- name: reset process replay
|
||||
run: test/external/process_replay/reset.py
|
||||
- name: Test driver start time
|
||||
run: time DEBUG=3 DEV=NV python3 test/test_tiny.py TestTiny.test_plus
|
||||
run: time DEBUG=3 NV=1 python3 test/test_tiny.py TestTiny.test_plus
|
||||
- name: Test tensor cores
|
||||
run: DEV=NV ALLOW_TF32=1 python3 test/opt/test_tensor_cores.py
|
||||
run: NV=1 ALLOW_TF32=1 python3 test/opt/test_tensor_cores.py
|
||||
- name: Test DISK copy time
|
||||
run: DEV=NV TESTFILE=/raid/downloads/llama3-8b-sfr/model-00001-of-00004.safetensors python3 test/external/external_benchmark_disk_raw.py
|
||||
run: NV=1 TESTFILE=/raid/downloads/llama3-8b-sfr/model-00001-of-00004.safetensors python3 test/external/external_benchmark_disk_raw.py
|
||||
- name: Test CPU copy time
|
||||
run: |
|
||||
DEV=NV GRAPH_ONE_KERNEL=1 PYTHONPATH=. NSZ=8192 python3 test/speed/external_test_copy_speed.py TestCopySpeed.testCopyDefaulttoCPUJit
|
||||
DEV=NV GRAPH_ONE_KERNEL=1 PYTHONPATH=. NSZ=8192 python3 test/speed/external_test_copy_speed.py TestCopySpeed.testCopyCPUtoDefaultJit
|
||||
NV=1 GRAPH_ONE_KERNEL=1 PYTHONPATH=. NSZ=8192 python3 test/speed/external_test_copy_speed.py TestCopySpeed.testCopyDefaulttoCPUJit
|
||||
NV=1 GRAPH_ONE_KERNEL=1 PYTHONPATH=. NSZ=8192 python3 test/speed/external_test_copy_speed.py TestCopySpeed.testCopyCPUtoDefaultJit
|
||||
- name: Test LLAMA-3
|
||||
run: BENCHMARK_LOG=llama3_beam DEV=NV JITBEAM=2 IGNORE_BEAM_CACHE=1 python3 examples/llama3.py --size 8B --benchmark --temperature 0
|
||||
run: BENCHMARK_LOG=llama3_beam NV=1 JITBEAM=2 IGNORE_BEAM_CACHE=1 python3 examples/llama3.py --size 8B --benchmark --temperature 0
|
||||
- name: Run full CIFAR training w 1 GPU
|
||||
run: time BENCHMARK_LOG=cifar DEV=NV DEFAULT_FLOAT=HALF STEPS=1000 TARGET_EVAL_ACC_PCT=93.0 python3 examples/hlb_cifar10.py
|
||||
run: time BENCHMARK_LOG=cifar NV=1 DEFAULT_FLOAT=HALF STEPS=1000 TARGET_EVAL_ACC_PCT=93.0 python3 examples/hlb_cifar10.py
|
||||
- name: Run 10 MLPerf ResNet50 training steps (1 gpu)
|
||||
run: BENCHMARK_LOG=resnet_10steps DEV=NV MNISTMOCK=1 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=256 GPUS=1 MODEL=resnet python3 examples/mlperf/model_train.py
|
||||
run: BENCHMARK_LOG=resnet_10steps NV=1 MNISTMOCK=1 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=256 GPUS=1 MODEL=resnet python3 examples/mlperf/model_train.py
|
||||
- name: Run 10 MLPerf Bert training steps (1 gpu)
|
||||
# TODO: remove BERT_LAYERS once scheduler is fast
|
||||
run: BENCHMARK_LOG=bert_10steps DEV=NV CAPTURE_PROCESS_REPLAY=0 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=66 GPUS=1 BERT_LAYERS=2 MODEL=bert python3 examples/mlperf/model_train.py
|
||||
- name: Remote
|
||||
run: |
|
||||
pkill -f 'extra/remote/serve.py' || true
|
||||
PYTHONPATH=. python3 extra/remote/serve.py 6483 &
|
||||
sleep 1
|
||||
DEBUG=2 PYTHONPATH=. REMOTE=127.0.0.1:6483 DEV=NV python3 test/test_tiny.py
|
||||
pkill -f 'extra/remote/serve.py' || true
|
||||
run: BENCHMARK_LOG=bert_10steps NV=1 CAPTURE_PROCESS_REPLAY=0 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=66 GPUS=1 BERT_LAYERS=2 MODEL=bert python3 examples/mlperf/model_train.py
|
||||
- name: Run process replay tests
|
||||
run: cp test/external/process_replay/process_replay.py ./process_replay.py && git fetch origin master && git -c advice.detachedHead=false checkout origin/master && PYTHONPATH=. python3 process_replay.py
|
||||
|
||||
@@ -14,7 +14,7 @@ jobs:
|
||||
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Remove amdgpu
|
||||
run: sudo rmmod amdgpu || true
|
||||
- name: Cleanup running AM processes
|
||||
@@ -22,10 +22,10 @@ jobs:
|
||||
- name: Run SDXL with new search
|
||||
# TODO: GCVM_L2_PROTECTION_FAULT_STATUS with llvm19
|
||||
run: |
|
||||
BENCHMARK_LOG=search_sdxl PYTHONPATH=. DEV=AMD JITBEAM=2 IGNORE_BEAM_CACHE=1 CCACHE=0 python examples/sdxl.py --noshow --timing --seed 0
|
||||
BENCHMARK_LOG=search_sdxl PYTHONPATH=. AMD=1 JITBEAM=2 IGNORE_BEAM_CACHE=1 CCACHE=0 python examples/sdxl.py --noshow --timing --seed 0
|
||||
- name: Run SDXL with cached search
|
||||
run: |
|
||||
BENCHMARK_LOG=search_sdxl_cached PYTHONPATH=. DEV=AMD JITBEAM=2 python examples/sdxl.py --noshow --timing --seed 0
|
||||
BENCHMARK_LOG=search_sdxl_cached PYTHONPATH=. AMD=1 JITBEAM=2 python examples/sdxl.py --noshow --timing --seed 0
|
||||
- name: Run winograd cifar with new search
|
||||
run: |
|
||||
BENCHMARK_LOG=search_wino_cifar WINO=1 DEFAULT_FLOAT=HALF JITBEAM=4 IGNORE_BEAM_CACHE=1 CCACHE=0 BS=1024 STEPS=500 python examples/hlb_cifar10.py
|
||||
|
||||
@@ -10,16 +10,16 @@ jobs:
|
||||
deploy:
|
||||
runs-on: ubuntu-latest
|
||||
steps:
|
||||
- uses: actions/checkout@v6
|
||||
- uses: actions/checkout@v4
|
||||
- name: Configure Git Credentials
|
||||
run: |
|
||||
git config user.name github-actions[bot]
|
||||
git config user.email 41898282+github-actions[bot]@users.noreply.github.com
|
||||
- uses: actions/setup-python@v6
|
||||
- uses: actions/setup-python@v5
|
||||
with:
|
||||
python-version: 3.x
|
||||
- run: echo "cache_id=$(date --utc '+%V')" >> $GITHUB_ENV
|
||||
- uses: actions/cache@v5
|
||||
- uses: actions/cache@v4
|
||||
with:
|
||||
key: mkdocs-material-${{ env.cache_id }}
|
||||
path: .cache
|
||||
|
||||
@@ -16,7 +16,7 @@ jobs:
|
||||
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Cleanup running AM processes
|
||||
run: python extra/amdpci/am_smi.py --pids --kill
|
||||
- name: Symlink datasets
|
||||
|
||||
@@ -12,9 +12,9 @@ jobs:
|
||||
deploy:
|
||||
runs-on: ubuntu-latest
|
||||
steps:
|
||||
- uses: actions/checkout@v6
|
||||
- uses: actions/checkout@v4
|
||||
- name: Set up Python
|
||||
uses: actions/setup-python@v6
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: '3.x'
|
||||
- name: Install dependencies
|
||||
|
||||
@@ -15,7 +15,7 @@ jobs:
|
||||
branchstat: ${{ steps.brstat.outputs.stat}}
|
||||
steps:
|
||||
- name: Check code from PR branch
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
with:
|
||||
repository: ${{ github.event.pull_request.head.repo.full_name }}
|
||||
ref: ${{ github.event.pull_request.head.sha }}
|
||||
@@ -46,18 +46,18 @@ jobs:
|
||||
if: needs.checkbranch.outputs.branchstat == 'false'
|
||||
steps:
|
||||
- name: Checkout code from PR branch
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
with:
|
||||
repository: ${{ github.event.pull_request.head.repo.full_name }}
|
||||
ref: ${{ github.event.pull_request.head.sha }}
|
||||
path: pr
|
||||
# the base default to tinygrad master and cannot be other fork branch for security purpose
|
||||
- name: Checkout code from tinygrad master
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
with:
|
||||
path: base
|
||||
- name: Set up Python 3.12
|
||||
uses: actions/setup-python@v6
|
||||
uses: actions/setup-python@v5
|
||||
with:
|
||||
python-version: '3.12'
|
||||
- name: Count Line Diff
|
||||
@@ -66,16 +66,18 @@ jobs:
|
||||
PR="$GITHUB_WORKSPACE/pr"
|
||||
pip install tabulate $BASE
|
||||
cp "$BASE/sz.py" .
|
||||
python sz.py "$BASE" "$PR" > loc_content.txt
|
||||
echo "loc_content<<EOF" >> "$GITHUB_ENV"
|
||||
python sz.py "$BASE" "$PR" >> "$GITHUB_ENV"
|
||||
echo "EOF" >> "$GITHUB_ENV"
|
||||
- name: Comment Code Line Diff
|
||||
continue-on-error: false
|
||||
uses: marocchino/sticky-pull-request-comment@v3
|
||||
uses: marocchino/sticky-pull-request-comment@v2
|
||||
with:
|
||||
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
|
||||
ignore_empty: true
|
||||
skip_unchanged: true
|
||||
recreate: true
|
||||
path: loc_content.txt
|
||||
message: ${{ env.loc_content }}
|
||||
|
||||
rebase:
|
||||
name: Core Library Line Difference
|
||||
@@ -87,7 +89,7 @@ jobs:
|
||||
steps:
|
||||
- name: Comment Rebase
|
||||
continue-on-error: false
|
||||
uses: marocchino/sticky-pull-request-comment@v3
|
||||
uses: marocchino/sticky-pull-request-comment@v2
|
||||
with:
|
||||
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
|
||||
skip_unchanged: true
|
||||
|
||||
+169
-178
@@ -21,7 +21,7 @@ jobs:
|
||||
timeout-minutes: 20
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -29,9 +29,9 @@ jobs:
|
||||
deps: testing_unit
|
||||
llvm: 'true'
|
||||
- name: Speed Test
|
||||
run: DEV=CPU:LLVM THREADS=0 python3 test/speed/external_test_speed_v_torch.py
|
||||
run: CPU=1 CPU_LLVM=1 THREADS=0 python3 test/speed/external_test_speed_v_torch.py
|
||||
- name: Speed Test (BEAM=2)
|
||||
run: BEAM=2 DEV=CPU:LLVM THREADS=0 python3 test/speed/external_test_speed_v_torch.py
|
||||
run: BEAM=2 CPU=1 CPU_LLVM=1 THREADS=0 python3 test/speed/external_test_speed_v_torch.py
|
||||
|
||||
docs:
|
||||
name: Docs
|
||||
@@ -41,7 +41,7 @@ jobs:
|
||||
CHECK_OOB: 0
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -70,7 +70,7 @@ jobs:
|
||||
source venv/bin/activate
|
||||
pip install $GITHUB_WORKSPACE
|
||||
cp $GITHUB_WORKSPACE/examples/beautiful_mnist.py .
|
||||
BS=2 STEPS=10 MAX_BUFFER_SIZE=0 python beautiful_mnist.py
|
||||
BS=2 STEPS=10 python beautiful_mnist.py
|
||||
- name: Test Docs Build
|
||||
run: python -m mkdocs build --strict
|
||||
- name: Test Docs
|
||||
@@ -83,7 +83,7 @@ jobs:
|
||||
run: DEBUG=100 python3 -c "from tinygrad import Tensor; N = 1024; a, b = Tensor.rand(N, N), Tensor.rand(N, N); c = (a.reshape(N, 1, N) * b.T.reshape(1, N, N)).sum(axis=2); print((c.numpy() - (a.numpy() @ b.numpy())).mean())"
|
||||
- name: Compile EfficientNet to C and test it
|
||||
run: |
|
||||
DEV=CPU python examples/compile_efficientnet.py > recognize.c
|
||||
CPU=1 CPU_LLVM=0 python examples/compile_efficientnet.py > recognize.c
|
||||
clang -O2 recognize.c -lm -o recognize
|
||||
cat test/models/efficientnet/Chicken.jpg | ./recognize | grep cock
|
||||
|
||||
@@ -93,7 +93,7 @@ jobs:
|
||||
timeout-minutes: 15
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -114,11 +114,11 @@ jobs:
|
||||
- name: Test one op in torch tests
|
||||
run: DEBUG=2 python3 extra/torch_backend/torch_tests.py TestTinyBackendPRIVATEUSE1.test_unary_log_tiny_float32
|
||||
- name: Test Ops with TINY_BACKEND
|
||||
run: DEV=CPU:LLVM LLVMOPT=0 TINY_BACKEND=1 python3 -m pytest -n auto test/backend/test_ops.py --durations=20
|
||||
run: CPU=1 CPU_LLVM=1 LLVMOPT=0 TINY_BACKEND=1 python3 -m pytest -n auto test/backend/test_ops.py --durations=20
|
||||
- name: Test in-place operations on views
|
||||
run: TORCH_DEBUG=1 python3 extra/torch_backend/test_inplace.py
|
||||
- name: Test multi-gpu
|
||||
run: DEV=CPU:LLVM GPUS=4 TORCH_DEBUG=1 python3 extra/torch_backend/test_multigpu.py
|
||||
run: CPU=1 CPU_LLVM=1 GPUS=4 TORCH_DEBUG=1 python3 extra/torch_backend/test_multigpu.py
|
||||
- name: Test kernel fusion
|
||||
run: python3 extra/torch_backend/test_kernel_fusion.py
|
||||
|
||||
@@ -129,7 +129,7 @@ jobs:
|
||||
timeout-minutes: 15
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -141,7 +141,7 @@ jobs:
|
||||
sudo apt update || true
|
||||
sudo apt install -y --no-install-recommends ninja-build
|
||||
- name: Test beautiful_mnist in torch with TINY_BACKEND
|
||||
run: STEPS=20 DEV=CPU TARGET_EVAL_ACC_PCT=90.0 MAX_BUFFER_SIZE=0 TINY_BACKEND=1 python3 examples/other_mnist/beautiful_mnist_torch.py
|
||||
run: STEPS=20 CPU=1 TARGET_EVAL_ACC_PCT=90.0 TINY_BACKEND=1 python3 examples/other_mnist/beautiful_mnist_torch.py
|
||||
- name: Test some torch tests (expect failure)
|
||||
run: python3 -m pytest extra/torch_backend/torch_tests.py -v --tb=no || true
|
||||
|
||||
@@ -151,67 +151,67 @@ jobs:
|
||||
timeout-minutes: 15
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
key: be-minimal
|
||||
deps: testing_unit
|
||||
- name: Test dtype with Python emulator
|
||||
run: DEBUG=1 DEV=PYTHON python3 -m pytest -n=auto test/backend/test_dtype.py test/backend/test_dtype_alu.py
|
||||
run: DEBUG=1 PYTHON=1 python3 -m pytest -n=auto test/backend/test_dtype.py test/backend/test_dtype_alu.py
|
||||
- name: Test ops with Python emulator
|
||||
run: DEBUG=2 SKIP_SLOW_TEST=1 DEV=PYTHON python3 -m pytest -n=auto test/backend/test_ops.py --durations=20
|
||||
run: DEBUG=2 SKIP_SLOW_TEST=1 PYTHON=1 python3 -m pytest -n=auto test/backend/test_ops.py --durations=20
|
||||
- name: Test uops with Python emulator
|
||||
run: DEV=PYTHON python3 -m pytest test/backend/test_uops.py --durations=20
|
||||
run: PYTHON=1 python3 -m pytest test/backend/test_uops.py --durations=20
|
||||
- name: Test symbolic with Python emulator
|
||||
run: DEV=PYTHON python3 test/backend/test_symbolic_ops.py
|
||||
run: PYTHON=1 python3 test/backend/test_symbolic_ops.py
|
||||
- name: test_renderer_failures with Python emulator
|
||||
run: DEV=PYTHON python3 -m pytest -rA test/backend/test_renderer_failures.py::TestRendererFailures
|
||||
- name: Test IMAGE support
|
||||
run: PYTHON=1 python3 -m pytest -rA test/backend/test_renderer_failures.py::TestRendererFailures
|
||||
- name: Test IMAGE=2 support
|
||||
run: |
|
||||
IMAGE=1 DEV=PYTHON python3 test/backend/test_ops.py TestOps.test_gemm
|
||||
IMAGE=1 DEV=PYTHON python3 test/backend/test_ops.py TestOps.test_simple_conv2d
|
||||
IMAGE=2 PYTHON=1 python3 test/backend/test_ops.py TestOps.test_gemm
|
||||
IMAGE=2 PYTHON=1 python3 test/backend/test_ops.py TestOps.test_simple_conv2d
|
||||
- name: Test emulated METAL tensor cores
|
||||
run: |
|
||||
DEBUG=2 FORWARD_ONLY=1 DEV=PYTHON::METAL python3 test/backend/test_ops.py TestOps.test_big_gemm
|
||||
DEBUG=2 FORWARD_ONLY=1 DEV=PYTHON::METAL python3 test/opt/test_tensor_cores.py
|
||||
DEBUG=2 EMULATE=METAL FORWARD_ONLY=1 PYTHON=1 python3 test/backend/test_ops.py TestOps.test_big_gemm
|
||||
DEBUG=2 EMULATE=METAL FORWARD_ONLY=1 PYTHON=1 python3 test/opt/test_tensor_cores.py
|
||||
- name: Test emulated AMX tensor cores
|
||||
run: DEBUG=2 AMX=1 FORWARD_ONLY=1 DEV=PYTHON::AMX python3 test/backend/test_ops.py TestOps.test_gemm
|
||||
run: DEBUG=2 AMX=1 EMULATE=AMX FORWARD_ONLY=1 PYTHON=1 python3 test/backend/test_ops.py TestOps.test_gemm
|
||||
- name: Test emulated AMD tensor cores
|
||||
run: |
|
||||
DEBUG=2 FORWARD_ONLY=1 DEV=PYTHON::gfx1100 N=16 HALF=1 ACC_HALF=0 python3 ./extra/gemm/simple_matmul.py
|
||||
DEBUG=2 FORWARD_ONLY=1 DEV=PYTHON::gfx1100 N=64 HALF=1 ACC_HALF=0 python3 ./extra/gemm/simple_matmul.py
|
||||
DEBUG=2 FORWARD_ONLY=1 DEV=PYTHON::gfx1100 N=16 HALF=1 ACC_HALF=1 ATOL=1e-3 python3 ./extra/gemm/simple_matmul.py
|
||||
DEBUG=2 FORWARD_ONLY=1 DEV=PYTHON::gfx1100 N=64 HALF=1 ACC_HALF=1 ATOL=1e-3 python3 ./extra/gemm/simple_matmul.py
|
||||
DEBUG=2 FORWARD_ONLY=1 DEV=PYTHON::gfx1100 python3 test/opt/test_tensor_cores.py
|
||||
DEBUG=2 EMULATE=AMD FORWARD_ONLY=1 PYTHON=1 N=16 HALF=1 ACC_HALF=0 python3 ./extra/gemm/simple_matmul.py
|
||||
DEBUG=2 EMULATE=AMD FORWARD_ONLY=1 PYTHON=1 N=64 HALF=1 ACC_HALF=0 python3 ./extra/gemm/simple_matmul.py
|
||||
DEBUG=2 EMULATE=AMD FORWARD_ONLY=1 PYTHON=1 N=16 HALF=1 ACC_HALF=1 ATOL=1e-3 python3 ./extra/gemm/simple_matmul.py
|
||||
DEBUG=2 EMULATE=AMD FORWARD_ONLY=1 PYTHON=1 N=64 HALF=1 ACC_HALF=1 ATOL=1e-3 python3 ./extra/gemm/simple_matmul.py
|
||||
DEBUG=2 EMULATE=AMD FORWARD_ONLY=1 PYTHON=1 python3 test/opt/test_tensor_cores.py
|
||||
- name: Test emulated AMD MFMA tensor cores
|
||||
run: |
|
||||
DEBUG=2 FORWARD_ONLY=1 DEV=PYTHON::gfx950 N=64 HALF=1 ACC_HALF=0 python3 ./extra/gemm/simple_matmul.py
|
||||
DEBUG=2 FORWARD_ONLY=1 DEV=PYTHON::gfx950 python3 test/opt/test_tensor_cores.py
|
||||
DEBUG=2 EMULATE=AMD_MFMA FORWARD_ONLY=1 PYTHON=1 N=64 HALF=1 ACC_HALF=0 python3 ./extra/gemm/simple_matmul.py
|
||||
DEBUG=2 EMULATE=AMD_MFMA FORWARD_ONLY=1 PYTHON=1 python3 test/opt/test_tensor_cores.py
|
||||
- name: Test emulated AMD RDNA4 tensor cores
|
||||
run: |
|
||||
DEBUG=2 FORWARD_ONLY=1 DEV=PYTHON::gfx1201 N=16 HALF=1 ACC_HALF=0 python3 ./extra/gemm/simple_matmul.py
|
||||
DEBUG=2 FORWARD_ONLY=1 DEV=PYTHON::gfx1201 N=64 HALF=1 ACC_HALF=0 python3 ./extra/gemm/simple_matmul.py
|
||||
DEBUG=2 FORWARD_ONLY=1 DEV=PYTHON::gfx1201 N=16 HALF=1 ACC_HALF=1 ATOL=1e-3 python3 ./extra/gemm/simple_matmul.py
|
||||
DEBUG=2 FORWARD_ONLY=1 DEV=PYTHON::gfx1201 N=64 HALF=1 ACC_HALF=1 ATOL=1e-3 python3 ./extra/gemm/simple_matmul.py
|
||||
DEBUG=2 FORWARD_ONLY=1 DEV=PYTHON::gfx1201 python3 test/opt/test_tensor_cores.py
|
||||
DEBUG=2 EMULATE=AMD_RDNA4 FORWARD_ONLY=1 PYTHON=1 N=16 HALF=1 ACC_HALF=0 python3 ./extra/gemm/simple_matmul.py
|
||||
DEBUG=2 EMULATE=AMD_RDNA4 FORWARD_ONLY=1 PYTHON=1 N=64 HALF=1 ACC_HALF=0 python3 ./extra/gemm/simple_matmul.py
|
||||
DEBUG=2 EMULATE=AMD_RDNA4 FORWARD_ONLY=1 PYTHON=1 N=16 HALF=1 ACC_HALF=1 ATOL=1e-3 python3 ./extra/gemm/simple_matmul.py
|
||||
DEBUG=2 EMULATE=AMD_RDNA4 FORWARD_ONLY=1 PYTHON=1 N=64 HALF=1 ACC_HALF=1 ATOL=1e-3 python3 ./extra/gemm/simple_matmul.py
|
||||
DEBUG=2 EMULATE=AMD_RDNA4 FORWARD_ONLY=1 PYTHON=1 python3 test/opt/test_tensor_cores.py
|
||||
- name: Test emulated CUDA tensor cores
|
||||
run: |
|
||||
DEBUG=2 FORWARD_ONLY=1 DEV=PYTHON::sm_80 python3 test/backend/test_ops.py TestOps.test_gemm_fp16
|
||||
DEBUG=2 ALLOW_TF32=1 FORWARD_ONLY=1 DEV=PYTHON::sm_80 python3 test/backend/test_ops.py TestOps.test_gemm
|
||||
DEBUG=2 FORWARD_ONLY=1 DEV=PYTHON::sm_75 python3 test/backend/test_ops.py TestOps.test_gemm_fp16
|
||||
DEBUG=2 ALLOW_TF32=1 FORWARD_ONLY=1 DEV=PYTHON::sm_89 python3 test/opt/test_tensor_cores.py
|
||||
DEBUG=2 EMULATE=CUDA FORWARD_ONLY=1 PYTHON=1 python3 test/backend/test_ops.py TestOps.test_gemm_fp16
|
||||
DEBUG=2 EMULATE=CUDA ALLOW_TF32=1 FORWARD_ONLY=1 PYTHON=1 python3 test/backend/test_ops.py TestOps.test_gemm
|
||||
DEBUG=2 EMULATE=CUDA_SM75 FORWARD_ONLY=1 PYTHON=1 python3 test/backend/test_ops.py TestOps.test_gemm_fp16
|
||||
DEBUG=2 EMULATE=CUDA_SM89 ALLOW_TF32=1 FORWARD_ONLY=1 PYTHON=1 python3 test/opt/test_tensor_cores.py
|
||||
- name: Test emulated INTEL OpenCL tensor cores
|
||||
run: DEBUG=2 FORWARD_ONLY=1 DEV=PYTHON::INTEL HALF=1 N=64 python3 ./extra/gemm/simple_matmul.py
|
||||
run: DEBUG=2 EMULATE=INTEL FORWARD_ONLY=1 PYTHON=1 HALF=1 N=64 python3 ./extra/gemm/simple_matmul.py
|
||||
- name: Test emulated AMX tensor cores
|
||||
run: DEBUG=2 AMX=1 FORWARD_ONLY=1 DEV=PYTHON::AMX python3 test/opt/test_tensor_cores.py
|
||||
run: DEBUG=2 AMX=1 EMULATE=AMX FORWARD_ONLY=1 PYTHON=1 python3 test/opt/test_tensor_cores.py
|
||||
- name: Test device flop counts
|
||||
run: |
|
||||
DEBUG=2 DEV=PYTHON::METAL python3 ./test/null/test_uops_stats.py TestUOpsStatsMatmulHalf
|
||||
DEBUG=2 DEV=PYTHON::gfx1100 python3 ./test/null/test_uops_stats.py TestUOpsStatsMatmulHalf
|
||||
DEBUG=2 DEV=PYTHON::sm_80 python3 ./test/null/test_uops_stats.py TestUOpsStatsMatmulHalf
|
||||
DEBUG=2 DEV=PYTHON::INTEL python3 ./test/null/test_uops_stats.py TestUOpsStatsMatmulHalf
|
||||
DEBUG=2 AMX=1 DEV=PYTHON::AMX python3 ./test/null/test_uops_stats.py TestUOpsStats.test_simple_matmul
|
||||
DEBUG=2 EMULATE=METAL PYTHON=1 python3 ./test/null/test_uops_stats.py TestUOpsStatsMatmulHalf
|
||||
DEBUG=2 EMULATE=AMD PYTHON=1 python3 ./test/null/test_uops_stats.py TestUOpsStatsMatmulHalf
|
||||
DEBUG=2 EMULATE=CUDA PYTHON=1 python3 ./test/null/test_uops_stats.py TestUOpsStatsMatmulHalf
|
||||
DEBUG=2 EMULATE=INTEL PYTHON=1 python3 ./test/null/test_uops_stats.py TestUOpsStatsMatmulHalf
|
||||
DEBUG=2 AMX=1 EMULATE=AMX PYTHON=1 python3 ./test/null/test_uops_stats.py TestUOpsStats.test_simple_matmul
|
||||
|
||||
linter:
|
||||
name: Linters
|
||||
@@ -220,7 +220,7 @@ jobs:
|
||||
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -251,7 +251,7 @@ jobs:
|
||||
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -261,19 +261,19 @@ jobs:
|
||||
llvm: 'true'
|
||||
amd: 'true'
|
||||
- name: Run NULL backend tests
|
||||
run: DEV=NULL python -m pytest -n=auto test/null/ --durations=20
|
||||
- name: Run targeted tests on NULL backend
|
||||
run: DEV=NULL python3 -m unittest test.backend.test_multitensor.TestMultiTensor.test_data_parallel_resnet_train_step
|
||||
run: NULL=1 python -m pytest -n=auto test/null/ --durations=20
|
||||
- name: Run targetted tests on NULL backend
|
||||
run: NULL=1 python3 -m unittest test.backend.test_multitensor.TestMultiTensor.test_data_parallel_resnet_train_step
|
||||
# TODO: too slow
|
||||
# - name: Run SDXL on NULL backend
|
||||
# run: DEV=NULL DEBUG=1 python3 examples/sdxl.py --seed 0 --noshow --timing --fakeweights
|
||||
# run: NULL=1 DEBUG=1 python3 examples/sdxl.py --seed 0 --noshow --timing --fakeweights
|
||||
- name: Run Clip tests for SD MLPerf on NULL backend
|
||||
run: DEV=NULL python -m pytest -n=auto test/external/mlperf_stable_diffusion/external_test_models.py::TestOpenClip --durations=20
|
||||
run: NULL=1 python -m pytest -n=auto test/external/mlperf_stable_diffusion/external_test_models.py::TestOpenClip --durations=20
|
||||
- name: Run AMD emulated BERT training on NULL backend
|
||||
run: DEV=NULL::gfx1201 NULL_ALLOW_COPYOUT=1 CAPTURE_PROCESS_REPLAY=0 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=66 GPUS=1 BERT_LAYERS=2 MODEL=bert python3 examples/mlperf/model_train.py
|
||||
run: EMULATE=AMD_RDNA4 NULL=1 NULL_ALLOW_COPYOUT=1 CAPTURE_PROCESS_REPLAY=0 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=66 GPUS=1 BERT_LAYERS=2 MODEL=bert python3 examples/mlperf/model_train.py
|
||||
# TODO: support fake weights
|
||||
#- name: Run LLaMA 7B on 4 fake devices
|
||||
# run: DEV=NULL python3 examples/llama.py --gen 1 --size 7B --shard 4 --prompt "Hello." --count 3 --temperature 0 --timing
|
||||
# run: NULL=1 python3 examples/llama.py --gen 1 --size 7B --shard 4 --prompt "Hello." --count 3 --temperature 0 --timing
|
||||
|
||||
unittest:
|
||||
name: Unit Tests
|
||||
@@ -282,7 +282,7 @@ jobs:
|
||||
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -297,8 +297,8 @@ jobs:
|
||||
run: python -c "from tinygrad import Device; assert Device.DEFAULT == 'CPU', Device.DEFAULT"
|
||||
- name: Run unit tests
|
||||
run: |
|
||||
DEV=CPU python test/null/test_device.py TestRunAsModule.test_module_runs
|
||||
DEV=CPU python -m pytest -n=auto test/unit/ --durations=20
|
||||
CPU=1 python test/null/test_device.py TestRunAsModule.test_module_runs
|
||||
CPU=1 python -m pytest -n=auto test/unit/ --durations=20
|
||||
- name: Run GC tests
|
||||
run: python test/external/external_uop_gc.py
|
||||
- name: External Benchmark Schedule
|
||||
@@ -325,7 +325,7 @@ jobs:
|
||||
timeout-minutes: 15
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -341,7 +341,7 @@ jobs:
|
||||
timeout-minutes: 10
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -362,17 +362,18 @@ jobs:
|
||||
timeout-minutes: 15
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
key: gpu-image
|
||||
deps: testing_unit
|
||||
opencl: 'true'
|
||||
- name: Test CL IMAGE=1 ops
|
||||
- name: Test CL IMAGE=2 ops
|
||||
run: |
|
||||
DEV=CL IMAGE=1 python -m pytest -n=auto test/backend/test_ops.py --durations=20
|
||||
DEV=CL IMAGE=1 python test/models/test_end2end.py TestEnd2End.test_linear_mnist
|
||||
CL=1 IMAGE=2 python -m pytest -n=auto test/backend/test_ops.py --durations=20
|
||||
# TODO: training is broken
|
||||
# CL=1 IMAGE=2 python test/models/test_end2end.py TestEnd2End.test_linear_mnist
|
||||
- name: Run process replay tests
|
||||
uses: ./.github/actions/process-replay
|
||||
|
||||
@@ -382,7 +383,7 @@ jobs:
|
||||
timeout-minutes: 10
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -390,13 +391,13 @@ jobs:
|
||||
deps: testing
|
||||
opencl: 'true'
|
||||
- name: Generate Dataset
|
||||
run: DEV=CL extra/optimization/generate_dataset.sh
|
||||
run: CL=1 extra/optimization/generate_dataset.sh
|
||||
- name: Run Kernel Count Test
|
||||
run: DEV=CL python -m pytest -n=auto test/external/external_test_opt.py
|
||||
run: CL=1 python -m pytest -n=auto test/external/external_test_opt.py
|
||||
- name: Run fused optimizer tests
|
||||
run: DEV=CL FUSE_OPTIM=1 python -m pytest -n=auto test/models/test_mnist.py test/backend/test_optim.py -k "not muon"
|
||||
run: CL=1 FUSE_OPTIM=1 python -m pytest -n=auto test/models/test_mnist.py test/backend/test_optim.py -k "not muon"
|
||||
- name: Upload artifact
|
||||
uses: actions/upload-artifact@v7
|
||||
uses: actions/upload-artifact@v4
|
||||
with:
|
||||
name: sops.gz
|
||||
path: /tmp/sops.gz
|
||||
@@ -407,7 +408,7 @@ jobs:
|
||||
timeout-minutes: 15
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -417,13 +418,13 @@ jobs:
|
||||
llvm: 'true'
|
||||
- name: Test openpilot model kernel count and gate usage
|
||||
run: |
|
||||
ALLOWED_KERNEL_COUNT=123 ALLOWED_READ_IMAGE=1486 ALLOWED_GATED_READ_IMAGE=17 FLOAT16=1 DEV=CL IMAGE=1 python examples/openpilot/compile3.py https://gitlab.com/commaai/openpilot-lfs.git/gitlab-lfs/objects/cf6376aa9a090f0da26c280ef69eabf9bbdd51d1faac9ed392919c3db69be916
|
||||
ALLOWED_KERNEL_COUNT=123 ALLOWED_READ_IMAGE=1397 ALLOWED_GATED_READ_IMAGE=94 FLOAT16=1 CL=1 IMAGE=2 python examples/openpilot/compile3.py https://gitlab.com/commaai/openpilot-lfs.git/gitlab-lfs/objects/cf6376aa9a090f0da26c280ef69eabf9bbdd51d1faac9ed392919c3db69be916
|
||||
- name: Test openpilot CL compile fp16
|
||||
run: FLOAT16=1 DEV=CL IMAGE=1 python examples/openpilot/compile3.py https://gitlab.com/commaai/openpilot-lfs.git/gitlab-lfs/objects/cf6376aa9a090f0da26c280ef69eabf9bbdd51d1faac9ed392919c3db69be916
|
||||
run: FLOAT16=1 DEBUGCL=1 CL=1 IMAGE=2 python examples/openpilot/compile3.py https://gitlab.com/commaai/openpilot-lfs.git/gitlab-lfs/objects/cf6376aa9a090f0da26c280ef69eabf9bbdd51d1faac9ed392919c3db69be916
|
||||
- name: Test openpilot CL compile fp32 (test correctness)
|
||||
run: DEV=CL IMAGE=1 SELFTEST=1 python examples/openpilot/compile3.py https://github.com/haraschax/filedump/raw/refs/heads/master/driving_vision_fp32.onnx
|
||||
run: DEBUGCL=1 CL=1 IMAGE=2 SELFTEST=1 python examples/openpilot/compile3.py https://github.com/haraschax/filedump/raw/refs/heads/master/driving_vision_fp32.onnx
|
||||
- name: Test openpilot LLVM compile fp16
|
||||
run: IMAGE=1 FLOAT16=1 DEV=CPU:LLVM python examples/openpilot/compile3.py https://gitlab.com/commaai/openpilot-lfs.git/gitlab-lfs/objects/cf6376aa9a090f0da26c280ef69eabf9bbdd51d1faac9ed392919c3db69be916
|
||||
run: FLOAT16=1 CPU=1 CPU_LLVM=1 python examples/openpilot/compile3.py https://gitlab.com/commaai/openpilot-lfs.git/gitlab-lfs/objects/cf6376aa9a090f0da26c280ef69eabf9bbdd51d1faac9ed392919c3db69be916
|
||||
- name: Run process replay tests
|
||||
uses: ./.github/actions/process-replay
|
||||
|
||||
@@ -436,7 +437,7 @@ jobs:
|
||||
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -445,15 +446,15 @@ jobs:
|
||||
python-version: '3.12'
|
||||
llvm: 'true'
|
||||
- name: Test ONNX (CPU)
|
||||
run: DEV=CPU python -m pytest -n=auto test/external/external_test_onnx_backend.py --durations=20
|
||||
run: CPU=1 CPU_LLVM=0 python -m pytest -n=auto test/external/external_test_onnx_backend.py --durations=20
|
||||
- name: Test ONNX (LLVM)
|
||||
run: DEV=CPU:LLVM python -m pytest -n=auto test/external/external_test_onnx_backend.py --durations=20
|
||||
run: CPU=1 CPU_LLVM=1 python -m pytest -n=auto test/external/external_test_onnx_backend.py --durations=20
|
||||
- name: Test ONNX Runner (CPU)
|
||||
run: DEV=CPU python3 test/external/external_test_onnx_runner.py
|
||||
run: CPU=1 CPU_LLVM=0 python3 test/external/external_test_onnx_runner.py
|
||||
- name: Test Additional ONNX Ops (CPU)
|
||||
run: DEV=CPU python3 test/external/external_test_onnx_ops.py
|
||||
run: CPU=1 CPU_LLVM=0 python3 test/external/external_test_onnx_ops.py
|
||||
- name: Test Quantize ONNX
|
||||
run: DEV=CPU python3 test/backend/test_quantize_onnx.py
|
||||
run: CPU=1 CPU_LLVM=0 python3 test/backend/test_quantize_onnx.py
|
||||
- name: Run process replay tests
|
||||
uses: ./.github/actions/process-replay
|
||||
|
||||
@@ -463,7 +464,7 @@ jobs:
|
||||
timeout-minutes: 20
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -473,21 +474,21 @@ jobs:
|
||||
python-version: '3.12'
|
||||
opencl: 'true'
|
||||
- name: Test ONNX (CL)
|
||||
run: DEV=CL python -m pytest -n=auto test/external/external_test_onnx_backend.py --durations=20
|
||||
run: CL=1 python -m pytest -n=auto test/external/external_test_onnx_backend.py --durations=20
|
||||
#- name: Test Optimization Helpers
|
||||
# run: DEBUG=1 python3 extra/optimization/test_helpers.py
|
||||
#- name: Test Action Space
|
||||
# run: DEBUG=1 DEV=CL python3 extra/optimization/get_action_space.py
|
||||
# run: DEBUG=1 CL=1 python3 extra/optimization/get_action_space.py
|
||||
- name: Test Beam Search
|
||||
run: DEV=CL IGNORE_BEAM_CACHE=1 python3 -m pytest extra/optimization/test_beam_search.py
|
||||
run: CL=1 IGNORE_BEAM_CACHE=1 python3 -m pytest extra/optimization/test_beam_search.py
|
||||
- name: Test MLPerf stuff
|
||||
run: DEV=CL python -m pytest -n=auto test/external/external_test_optim.py test/external/external_test_losses.py test/external/external_test_metrics.py test/external/external_test_datasets.py --durations=20
|
||||
- name: DEV=NULL beautiful_mnist_multigpu
|
||||
run: DEV=NULL NULL_ALLOW_COPYOUT=1 python examples/beautiful_mnist_multigpu.py
|
||||
run: CL=1 python -m pytest -n=auto test/external/external_test_optim.py test/external/external_test_losses.py test/external/external_test_metrics.py test/external/external_test_datasets.py --durations=20
|
||||
- name: NULL=1 beautiful_mnist_multigpu
|
||||
run: NULL=1 NULL_ALLOW_COPYOUT=1 python examples/beautiful_mnist_multigpu.py
|
||||
- name: Test Bert training
|
||||
run: DEV=NULL NULL_ALLOW_COPYOUT=1 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=24 GPUS=4 BERT_LAYERS=2 MODEL=bert python3 examples/mlperf/model_train.py
|
||||
run: NULL=1 NULL_ALLOW_COPYOUT=1 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=24 GPUS=4 BERT_LAYERS=2 MODEL=bert python3 examples/mlperf/model_train.py
|
||||
- name: Test llama 3 training
|
||||
run: DEV=NULL NULL_ALLOW_COPYOUT=1 SAMPLES=300 BS=8 SEQLEN=512 GRADIENT_ACC_STEPS=1 FAKEDATA=1 DEFAULT_FLOAT=bfloat16 OPTIM_DTYPE=bfloat16 LLAMA3_SIZE=1B MODEL=llama3 python3 examples/mlperf/model_train.py
|
||||
run: NULL=1 NULL_ALLOW_COPYOUT=1 SAMPLES=300 BS=8 SEQLEN=512 GRADIENT_ACC_STEPS=1 FAKEDATA=1 DEFAULT_FLOAT=bfloat16 OPTIM_DTYPE=bfloat16 LLAMA3_SIZE=1B MODEL=llama3 python3 examples/mlperf/model_train.py
|
||||
- name: Run process replay tests
|
||||
uses: ./.github/actions/process-replay
|
||||
|
||||
@@ -499,18 +500,13 @@ jobs:
|
||||
CHECK_OOB: 0
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
key: apps_llm
|
||||
- name: Test 1B LLM (llama)
|
||||
run: echo "What's a male chicken called? Answer with only one word." | MAX_BUFFER_SIZE=0 python3 -m tinygrad.apps.llm --model llama3.2:1b | tee /dev/stderr | grep -i rooster
|
||||
- name: Test 1B LLM (llama q4)
|
||||
run: echo "What's a male chicken called? Answer with only one word." | MAX_BUFFER_SIZE=0 python3 -m tinygrad.apps.llm --model llama3.2:1b-q4 | tee /dev/stderr | grep -i rooster
|
||||
- name: Test 1B LLM (qwen)
|
||||
# NOTE: qwen is dumb and only knows about female chickens
|
||||
run: echo "What's a female chicken called? Answer with only one word." | MAX_BUFFER_SIZE=0 python3 -m tinygrad.apps.llm --model qwen3:0.6b | tee /dev/stderr | grep -i hen
|
||||
- name: Test 1B LLM
|
||||
run: echo "What's a male chicken called? Answer with only one word." | MAX_BUFFER_SIZE=0 python3 -m tinygrad.apps.llm | grep -i rooster
|
||||
|
||||
# ****** Models Tests ******
|
||||
|
||||
@@ -520,7 +516,7 @@ jobs:
|
||||
timeout-minutes: 15
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -529,11 +525,11 @@ jobs:
|
||||
opencl: 'true'
|
||||
llvm: 'true'
|
||||
- name: Test models (llvm)
|
||||
run: DEV=CPU:LLVM python -m pytest -n=auto test/models --durations=20
|
||||
run: CPU=1 CPU_LLVM=1 python -m pytest -n=auto test/models --durations=20
|
||||
- name: Test models (opencl)
|
||||
run: DEV=CL python -m pytest -n=auto test/models --durations=20
|
||||
run: CL=1 python -m pytest -n=auto test/models --durations=20
|
||||
- name: Test models (cpu)
|
||||
run: DEV=CPU python -m pytest -n=auto test/models --durations=20
|
||||
run: CPU=1 CPU_LLVM=0 python -m pytest -n=auto test/models --durations=20
|
||||
- name: Run process replay tests
|
||||
uses: ./.github/actions/process-replay
|
||||
|
||||
@@ -543,7 +539,7 @@ jobs:
|
||||
timeout-minutes: 20
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -551,9 +547,9 @@ jobs:
|
||||
deps: testing
|
||||
python-version: '3.12'
|
||||
- name: Test models (Metal)
|
||||
run: DEV=METAL python -m pytest -n=auto test/models --durations=20
|
||||
run: METAL=1 python -m pytest -n=auto test/models --durations=20
|
||||
- name: Test LLaMA compile speed
|
||||
run: DEV=METAL python test/external/external_test_speed_llama.py
|
||||
run: METAL=1 python test/external/external_test_speed_llama.py
|
||||
|
||||
# ****** Feature Tests ******
|
||||
|
||||
@@ -563,7 +559,7 @@ jobs:
|
||||
timeout-minutes: 15
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -572,11 +568,11 @@ jobs:
|
||||
pydeps: "pillow"
|
||||
llvm: "true"
|
||||
- name: Test LLVM=1 DEVECTORIZE=0
|
||||
run: DEV=CPU:LLVM DEVECTORIZE=0 python3 -m pytest -n auto test/test_tiny.py test/backend/test_ops.py
|
||||
run: CPU=1 CPU_LLVM=1 DEVECTORIZE=0 python3 -m pytest -n auto test/test_tiny.py test/backend/test_ops.py
|
||||
- name: Test LLVM=1 DEVECTORIZE=0 for model
|
||||
run: DEV=CPU:LLVM DEVECTORIZE=0 python3 test/models/test_efficientnet.py
|
||||
- name: Test DEV=CPU DEVECTORIZE=0
|
||||
run: DEV=CPU DEVECTORIZE=0 python3 -m pytest -n auto test/test_tiny.py test/backend/test_ops.py
|
||||
run: CPU=1 CPU_LLVM=1 DEVECTORIZE=0 python3 test/models/test_efficientnet.py
|
||||
- name: Test CPU=1 DEVECTORIZE=0
|
||||
run: CPU=1 CPU_LLVM=0 DEVECTORIZE=0 python3 -m pytest -n auto test/test_tiny.py test/backend/test_ops.py
|
||||
|
||||
testdsp:
|
||||
name: Linux (DSP)
|
||||
@@ -584,7 +580,7 @@ jobs:
|
||||
timeout-minutes: 15
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -593,9 +589,9 @@ jobs:
|
||||
pydeps: "onnx==1.18.0 onnxruntime ml_dtypes"
|
||||
llvm: "true"
|
||||
- name: Set up Docker Buildx
|
||||
uses: docker/setup-buildx-action@v4
|
||||
uses: docker/setup-buildx-action@v3
|
||||
- name: Build QEMU Docker with cache
|
||||
uses: docker/build-push-action@v7
|
||||
uses: docker/build-push-action@v4
|
||||
with:
|
||||
file: extra/dsp/Dockerfile
|
||||
push: false
|
||||
@@ -606,11 +602,11 @@ jobs:
|
||||
- name: Set MOCKDSP env
|
||||
run: printf "MOCKDSP=1" >> $GITHUB_ENV
|
||||
- name: Run test_tiny on DSP
|
||||
run: DEBUG=2 DEV=DSP python test/test_tiny.py
|
||||
run: DEBUG=2 DSP=1 python test/test_tiny.py
|
||||
- name: Test transcendentals
|
||||
run: CC=clang-20 DEBUG=2 DEV=DSP python test/backend/test_transcendental.py TestTranscendentalVectorized
|
||||
run: CC=clang-20 DEBUG=2 DSP=1 python test/backend/test_transcendental.py TestTranscendentalVectorized
|
||||
- name: Test quantize onnx
|
||||
run: DEBUG=2 DEV=DSP python3 test/backend/test_quantize_onnx.py
|
||||
run: DEBUG=2 DSP=1 python3 test/backend/test_quantize_onnx.py
|
||||
|
||||
testwebgpu:
|
||||
name: Linux (WebGPU)
|
||||
@@ -618,7 +614,7 @@ jobs:
|
||||
timeout-minutes: 20
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -628,11 +624,11 @@ jobs:
|
||||
webgpu: 'true'
|
||||
- name: Check Device.DEFAULT (WEBGPU) and print some source
|
||||
run: |
|
||||
DEV=WEBGPU python -c "from tinygrad import Device; assert Device.DEFAULT == 'WEBGPU', Device.DEFAULT"
|
||||
DEV=WEBGPU DEBUG=4 FORWARD_ONLY=1 python3 test/test_tiny.py TestTiny.test_plus
|
||||
WEBGPU=1 python -c "from tinygrad import Device; assert Device.DEFAULT == 'WEBGPU', Device.DEFAULT"
|
||||
WEBGPU=1 DEBUG=4 FORWARD_ONLY=1 python3 test/test_tiny.py TestTiny.test_plus
|
||||
- name: Run selected webgpu tests
|
||||
run: |
|
||||
DEV=WEBGPU WEBGPU_BACKEND="WGPUBackendType_Vulkan" python3 -m pytest -n=auto test/backend --durations=20
|
||||
WEBGPU=1 WEBGPU_BACKEND="WGPUBackendType_Vulkan" python3 -m pytest -n=auto test/backend --durations=20
|
||||
- name: Run process replay tests
|
||||
uses: ./.github/actions/process-replay
|
||||
|
||||
@@ -641,12 +637,12 @@ jobs:
|
||||
runs-on: ubuntu-24.04
|
||||
timeout-minutes: 20
|
||||
env:
|
||||
DEV: AMD
|
||||
AMD: 1
|
||||
PYTHON_REMU: 1
|
||||
MOCKGPU: 1
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -667,34 +663,33 @@ jobs:
|
||||
- name: Install rocprof-trace-decoder
|
||||
run: sudo PYTHONPATH="." ./extra/sqtt/install_rocprof_decoder.py
|
||||
- name: Run AMD renderer tests
|
||||
run: python -m pytest -n=auto test/amd/ --durations 20
|
||||
- name: Run AMD renderer tests (AMD:LLVM)
|
||||
run: DEV=AMD:LLVM python -m pytest -n=auto test/amd/ --durations 20
|
||||
run: AMD_LLVM=0 python -m pytest -n=auto test/amd/ --durations 20
|
||||
- name: Run AMD renderer tests (AMD_LLVM=1)
|
||||
run: AMD_LLVM=1 python -m pytest -n=auto test/amd/ --durations 20
|
||||
- name: Run SQTT profiling tests
|
||||
run: PROFILE=1 SQTT=1 python3 -m pytest -n=auto test/amd/test_sqtt_profiler.py
|
||||
- name: Run AMD emulated tests on NULL backend
|
||||
env:
|
||||
AMD: 0
|
||||
run: |
|
||||
PYTHONPATH=. DEV=NULL::gfx1100 python extra/mmapeak/mmapeak.py
|
||||
PYTHONPATH=. DEV=NULL::gfx1201 python3 -m pytest -n=auto test/testextra/test_tk.py test/backend/test_asm_gemm.py
|
||||
- name: Run matmul on MOCKGPU
|
||||
run: |
|
||||
PYTHONPATH="." DEV=AMD MOCKGPU=1 N=256 python3 extra/gemm/amd_asm_matmul.py
|
||||
PYTHONPATH="." DEV=AMD MOCKGPU=1 N=256 python3 extra/gemm/amd_copy_matmul.py
|
||||
PYTHONPATH=. NULL=1 EMULATE=AMD python extra/mmapeak/mmapeak.py
|
||||
PYTHONPATH=. NULL=1 EMULATE=AMD_CDNA4 python3 -m pytest -n=auto test/testextra/test_tk.py test/backend/test_asm_gemm.py
|
||||
- name: Run ASM matmul on MOCKGPU
|
||||
run: PYTHONPATH="." AMD=1 MOCKGPU=1 N=256 python3 extra/gemm/amd_asm_matmul.py
|
||||
- name: Run LLVM test
|
||||
run: DEV=AMD:LLVM python test/device/test_amd_llvm.py
|
||||
run: AMD_LLVM=1 python test/device/test_amd_llvm.py
|
||||
|
||||
testmockam:
|
||||
name: Linux (am)
|
||||
runs-on: ubuntu-24.04
|
||||
timeout-minutes: 15
|
||||
env:
|
||||
DEV: PCI+AMD
|
||||
AMD: 1
|
||||
MOCKGPU: 1
|
||||
AMD_IFACE: PCI
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -704,35 +699,30 @@ jobs:
|
||||
- name: Run test_tiny on MOCKAM
|
||||
run: python test/test_tiny.py
|
||||
- name: Run test_tiny on MOCKAM USB
|
||||
run: GMMU=0 DEV=USB+AMD python test/test_tiny.py
|
||||
run: AMD_IFACE=USB python test/test_tiny.py
|
||||
- name: Run test_hcq on MOCKAM
|
||||
run: python -m pytest test/device/test_hcq.py
|
||||
- name: Run disk copy tests on MOCKAM
|
||||
run: python -m pytest test/unit/test_disk_tensor.py -k test_copy_from_disk
|
||||
- name: Run test_tiny on MOCKAM Remote
|
||||
run: |
|
||||
python extra/remote/serve.py 6667 &
|
||||
sleep 2
|
||||
REMOTE=127.0.0.1:6667 python test/test_tiny.py
|
||||
REMOTE=127.0.0.1:6667 python -m pytest test/unit/test_disk_tensor.py -k test_copy_from_disk; kill %1
|
||||
|
||||
testamd:
|
||||
strategy:
|
||||
fail-fast: false
|
||||
matrix:
|
||||
backend: [amd, amdllvm]
|
||||
arch: [gfx1100, gfx1201, gfx950]
|
||||
arch: [rdna3, rdna4]
|
||||
#arch: [rdna3, rdna4, cdna4]
|
||||
|
||||
name: Linux (${{ matrix.backend }} ${{ matrix.arch }})
|
||||
runs-on: ubuntu-22.04
|
||||
timeout-minutes: 15
|
||||
env:
|
||||
DEV: AMD:${{ matrix.backend == 'amdllvm' && 'LLVM' || '' }}:${{ matrix.arch }}
|
||||
AMD: 1
|
||||
MOCKGPU: 1
|
||||
MOCKGPU_ARCH: ${{ matrix.arch }}
|
||||
SKIP_SLOW_TEST: 1
|
||||
AMD_LLVM: ${{ matrix.backend == 'amdllvm' && '1' || matrix.backend != 'amdllvm' && '0' }}
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -745,9 +735,7 @@ jobs:
|
||||
python3 -c "from tinygrad import Device; assert Device.DEFAULT in ['AMD'], Device.DEFAULT"
|
||||
DEBUG=5 FORWARD_ONLY=1 python3 test/test_tiny.py TestTiny.test_plus
|
||||
- name: Run pytest (amd)
|
||||
run: python -m pytest -n=auto test/backend/test_ops.py test/backend/test_dtype.py test/backend/test_dtype_alu.py test/backend/test_linearizer.py test/backend/test_randomness.py test/backend/test_jit.py test/backend/test_graph.py test/backend/test_multitensor.py test/device/test_hcq.py test/external/external_test_am.py test/backend/test_asm_gemm.py::TestAsmGEMM --durations=20
|
||||
- name: Run disk copy tests
|
||||
run: python -m pytest test/unit/test_disk_tensor.py -k test_copy_from_disk
|
||||
run: python -m pytest -n=auto test/backend/test_ops.py test/backend/test_dtype.py test/backend/test_dtype_alu.py test/backend/test_linearizer.py test/backend/test_randomness.py test/backend/test_jit.py test/backend/test_graph.py test/backend/test_multitensor.py test/device/test_hcq.py test/testextra/test_cfg_viz.py test/external/external_test_am.py --durations=20
|
||||
- name: Run TRANSCENDENTAL math
|
||||
run: TRANSCENDENTAL=2 python -m pytest -n=auto test/backend/test_ops.py::TestOps::test_sin test/backend/test_ops.py::TestOps::test_cos test/backend/test_ops.py::TestOps::test_tan test/backend/test_ops.py::TestOps::test_exp test/backend/test_ops.py::TestOps::test_log --durations=20
|
||||
- name: Run process replay tests
|
||||
@@ -767,7 +755,7 @@ jobs:
|
||||
FORWARD_ONLY: 1
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -776,7 +764,7 @@ jobs:
|
||||
cuda: 'true'
|
||||
ocelot: 'true'
|
||||
- name: Set env
|
||||
run: printf "${{ matrix.backend == 'ptx' && 'DEV=CUDA:PTX' || matrix.backend == 'nv' && 'DEV=NV\nSKIP_SLOW_TEST=1' }}" >> $GITHUB_ENV
|
||||
run: printf "${{ matrix.backend == 'PTX' && 'CUDA=1\nCUDA_PTX=1' || matrix.backend == 'nv' && 'NV=1\nSKIP_SLOW_TEST=1' }}" >> $GITHUB_ENV
|
||||
- name: Check Device.DEFAULT and print some source
|
||||
run: |
|
||||
python3 -c "from tinygrad import Device; assert Device.DEFAULT in ['CUDA','NV'], Device.DEFAULT"
|
||||
@@ -800,7 +788,7 @@ jobs:
|
||||
timeout-minutes: 20
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -810,7 +798,7 @@ jobs:
|
||||
llvm: ${{ matrix.backend == 'llvm' || matrix.backend == 'lvp' }}
|
||||
mesa: ${{ matrix.backend == 'lvp' && 'true' }}
|
||||
- name: Set env
|
||||
run: printf "${{ matrix.backend == 'llvm' && 'DEV=CPU:LLVM' || matrix.backend == 'cpu' && 'DEV=CPU\nCPU_COUNT=2' || matrix.backend == 'opencl' && 'DEV=CL' || matrix.backend == 'lvp' && 'DEV=CPU:LVP' }}" >> $GITHUB_ENV
|
||||
run: printf "${{ matrix.backend == 'llvm' && 'CPU=1\nCPU_LLVM=1' || matrix.backend == 'cpu' && 'CPU=1\nCPU_LLVM=0\nCPU_COUNT=2' || matrix.backend == 'opencl' && 'CL=1' || matrix.backend == 'lvp' && 'CPU=1\nCPU_LVP=1' }}" >> $GITHUB_ENV
|
||||
- name: Check Device.DEFAULT and print some source
|
||||
run: |
|
||||
python3 -c "from tinygrad import Device; assert Device.DEFAULT in ['CPU','CL'], Device.DEFAULT"
|
||||
@@ -830,7 +818,7 @@ jobs:
|
||||
timeout-minutes: 20
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -842,41 +830,44 @@ jobs:
|
||||
ocelot: 'true'
|
||||
llvm: 'true'
|
||||
- name: Run unit tests
|
||||
run: DEV=METAL python -m pytest -n=auto test/unit/ --durations=20
|
||||
run: METAL=1 python -m pytest -n=auto test/unit/ --durations=20
|
||||
- name: Run NULL backend tests
|
||||
run: DEV=NULL python -m pytest -n=auto test/null/ --durations=20
|
||||
run: NULL=1 python -m pytest -n=auto test/null/ --durations=20
|
||||
- name: Run ONNX
|
||||
run: DEV=METAL python -m pytest -n=auto test/external/external_test_onnx_backend.py --durations=20
|
||||
run: METAL=1 python -m pytest -n=auto test/external/external_test_onnx_backend.py --durations=20
|
||||
- name: Test tensor core ops (fake)
|
||||
run: DEV=METAL DEBUG=3 TC=2 python test/backend/test_ops.py TestOps.test_gemm
|
||||
run: METAL=1 DEBUG=3 TC=2 python test/backend/test_ops.py TestOps.test_gemm
|
||||
- name: Test tensor core ops (real)
|
||||
run: DEV=METAL DEBUG=3 python test/backend/test_ops.py TestOps.test_big_gemm
|
||||
run: METAL=1 DEBUG=3 python test/backend/test_ops.py TestOps.test_big_gemm
|
||||
- name: Test Beam Search
|
||||
run: DEV=METAL IGNORE_BEAM_CACHE=1 python3 -m pytest extra/optimization/test_beam_search.py
|
||||
run: METAL=1 IGNORE_BEAM_CACHE=1 python3 -m pytest extra/optimization/test_beam_search.py
|
||||
- name: Test Device Specific
|
||||
run: DEV=METAL python3 -m pytest test/device/test_metal.py
|
||||
run: METAL=1 python3 -m pytest test/device/test_metal.py
|
||||
#- name: Fuzz Test linearizer
|
||||
# run: DEV=METAL DEPTH=4 FUZZ_N=50 FUZZ_MAX_SIZE=1000000 python test/external/fuzz_linearizer.py
|
||||
# run: METAL=1 DEPTH=4 FUZZ_N=50 FUZZ_MAX_SIZE=1000000 python test/external/fuzz_linearizer.py
|
||||
- name: Run TRANSCENDENTAL math
|
||||
run: DEV=METAL TRANSCENDENTAL=2 python -m pytest -n=auto test/backend/test_ops.py::TestOps::test_sin test/backend/test_ops.py::TestOps::test_cos test/backend/test_ops.py::TestOps::test_tan test/backend/test_ops.py::TestOps::test_exp test/backend/test_ops.py::TestOps::test_log --durations=20
|
||||
run: METAL=1 TRANSCENDENTAL=2 python -m pytest -n=auto test/backend/test_ops.py::TestOps::test_sin test/backend/test_ops.py::TestOps::test_cos test/backend/test_ops.py::TestOps::test_tan test/backend/test_ops.py::TestOps::test_exp test/backend/test_ops.py::TestOps::test_log --durations=20
|
||||
- name: Run pytest (amd)
|
||||
env:
|
||||
MOCKGPU: 1
|
||||
DEV: AMD
|
||||
AMD: 1
|
||||
AMD_LLVM: 0
|
||||
FORWARD_ONLY: 1
|
||||
run: |
|
||||
python3 -m pytest -n=auto test/device/test_hcq.py test/test_tiny.py --durations=20
|
||||
- name: Run pytest (amd with llvm backend)
|
||||
env:
|
||||
MOCKGPU: 1
|
||||
DEV: "AMD:LLVM"
|
||||
AMD: 1
|
||||
AMD_LLVM: 1
|
||||
FORWARD_ONLY: 1
|
||||
run: |
|
||||
python -m pytest -n=auto test/device/test_hcq.py test/test_tiny.py test/device/test_amd_llvm.py --durations=20
|
||||
- name: Run pytest (ptx)
|
||||
env:
|
||||
MOCKGPU: 1
|
||||
DEV: "NV:PTX"
|
||||
NV_PTX: 1
|
||||
NV: 1
|
||||
FORWARD_ONLY: 1
|
||||
# TODO: failing due to library loading error
|
||||
CAPTURE_PROCESS_REPLAY: 0
|
||||
@@ -891,7 +882,7 @@ jobs:
|
||||
timeout-minutes: 10
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -899,9 +890,9 @@ jobs:
|
||||
deps: testing
|
||||
webgpu: 'true'
|
||||
- name: Build WEBGPU Efficientnet
|
||||
run: DEV=WEBGPU WEBGPU_BACKEND="WGPUBackendType_Metal" python3 -m examples.compile_efficientnet
|
||||
run: WEBGPU=1 WEBGPU_BACKEND="WGPUBackendType_Metal" python3 -m examples.compile_efficientnet
|
||||
- name: Run selected webgpu tests
|
||||
run: DEV=WEBGPU WEBGPU_BACKEND="WGPUBackendType_Metal" python3 -m pytest -n=auto test/backend --durations=20
|
||||
run: WEBGPU=1 WEBGPU_BACKEND="WGPUBackendType_Metal" python3 -m pytest -n=auto test/backend --durations=20
|
||||
#- name: Clean npm cache
|
||||
# run: npm cache clean --force
|
||||
#- name: Install Puppeteer
|
||||
@@ -920,7 +911,7 @@ jobs:
|
||||
# cp $GITHUB_WORKSPACE/test/web/test_viz.js .
|
||||
# node test_viz.js
|
||||
- name: Test ONNX Runner (WEBGPU)
|
||||
run: DEV=WEBGPU python3 test/external/external_test_onnx_runner.py
|
||||
run: WEBGPU=1 python3 test/external/external_test_onnx_runner.py
|
||||
|
||||
osxtests:
|
||||
strategy:
|
||||
@@ -932,7 +923,7 @@ jobs:
|
||||
timeout-minutes: 20
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -941,7 +932,7 @@ jobs:
|
||||
llvm: ${{ matrix.backend == 'llvm' || matrix.backend == 'lvp' }}
|
||||
mesa: ${{ matrix.backend == 'lvp' && 'true' }}
|
||||
- name: Set env
|
||||
run: printf "${{ matrix.backend == 'llvm' && 'DEV=CPU:LLVM' || matrix.backend == 'cpu' && 'DEV=CPU\nCPU_COUNT=2' || matrix.backend == 'metal' && 'DEV=METAL' || matrix.backend == 'lvp' && 'DEV=CPU:LVP' }}" >> $GITHUB_ENV
|
||||
run: printf "${{ matrix.backend == 'llvm' && 'CPU=1\nCPU_LLVM=1' || matrix.backend == 'cpu' && 'CPU=1\nCPU_LLVM=0\nCPU_COUNT=2' || matrix.backend == 'metal' && 'METAL=1' || matrix.backend == 'lvp' && 'CPU=1\nCPU_LVP=1' }}" >> $GITHUB_ENV
|
||||
- name: Check Device.DEFAULT and print some source
|
||||
run: |
|
||||
python -c "from tinygrad import Device; assert Device.DEFAULT == {'LLVM':'CPU','LVP':'CPU'}.get(x:='${{ matrix.backend }}'.upper(), x), Device.DEFAULT"
|
||||
@@ -967,7 +958,7 @@ jobs:
|
||||
timeout-minutes: 15
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -976,7 +967,7 @@ jobs:
|
||||
pydeps: ${{ matrix.backend == 'webgpu' && 'dawn-python' || '' }}
|
||||
- name: Set env
|
||||
shell: bash
|
||||
run: printf "${{ matrix.backend == 'llvm' && 'DEV=CPU:LLVM' || matrix.backend == 'cpu' && 'DEV=CPU\nCPU_COUNT=2' || matrix.backend == 'webgpu' && 'DEV=WEBGPU'}}" >> $GITHUB_ENV
|
||||
run: printf "${{ matrix.backend == 'llvm' && 'CPU=1\nCPU_LLVM=1' || matrix.backend == 'cpu' && 'CPU=1\nCPU_LLVM=0\nCPU_COUNT=2' || matrix.backend == 'webgpu' && 'WEBGPU=1'}}" >> $GITHUB_ENV
|
||||
- name: Run unit tests
|
||||
if: matrix.backend=='llvm'
|
||||
# test_newton_schulz hits RecursionError
|
||||
@@ -984,7 +975,7 @@ jobs:
|
||||
- name: Run NULL backend tests
|
||||
if: matrix.backend=='llvm'
|
||||
shell: bash
|
||||
run: DEV=NULL python -m pytest -n=auto test/null/ --ignore=test/null/test_elf.py --durations=20
|
||||
run: CPU=0 CPU_LLVM=0 NULL=1 python -m pytest -n=auto test/null/ --ignore=test/null/test_elf.py --durations=20
|
||||
- name: Run pytest (${{ matrix.backend }})
|
||||
shell: bash
|
||||
run: |
|
||||
@@ -1003,7 +994,7 @@ jobs:
|
||||
timeout-minutes: 15
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -1013,7 +1004,7 @@ jobs:
|
||||
python-version: '3.12'
|
||||
- name: Set env
|
||||
shell: bash
|
||||
run: printf "NULL_ALLOW_COPYOUT=1\n${{ matrix.backend == 'ir3' && 'DEV=NULL:IR3:a630' || matrix.backend == 'nak' && 'DEV=NULL:NAK:sm_120' }}" >> $GITHUB_ENV
|
||||
run: printf "NULL=1\nNULL_ALLOW_COPYOUT=1\n${{ matrix.backend == 'ir3' && 'NULL_IR3=1' || matrix.backend == 'nak' && 'NULL_NAK=1' }}" >> $GITHUB_ENV
|
||||
- name: Run test_ops
|
||||
shell: bash
|
||||
run: |
|
||||
@@ -1026,7 +1017,7 @@ jobs:
|
||||
timeout-minutes: 15
|
||||
steps:
|
||||
- name: Checkout Code
|
||||
uses: actions/checkout@v6
|
||||
uses: actions/checkout@v4
|
||||
- name: Setup Environment
|
||||
uses: ./.github/actions/setup-tinygrad
|
||||
with:
|
||||
@@ -1036,7 +1027,7 @@ jobs:
|
||||
python-version: '3.12'
|
||||
- name: Set env
|
||||
shell: bash
|
||||
run: printf "DEV=NULL:QCOMCL:a630\nNULL_ALLOW_COPYOUT=1" >> $GITHUB_ENV
|
||||
run: printf "NULL=1\nNULL_ALLOW_COPYOUT=1\nNULL_QCOMCL=1" >> $GITHUB_ENV
|
||||
- name: Run test_ops
|
||||
shell: bash
|
||||
run: |
|
||||
|
||||
@@ -3,7 +3,7 @@
|
||||
AM driver is a userspace driver targeting AMD's RDNA3/RDNA4. You only need tinygrad to send compute tasks to your GPU!
|
||||
|
||||
## How to run?
|
||||
Make sure that amdgpu module is unloaded and just run tinygrad with `DEV=AMD`!
|
||||
Make sure that amdgpu module is unloaded and just run tinygrad with `AMD=1`!
|
||||
|
||||
Optional requirements:
|
||||
|
||||
|
||||
+8
-19
@@ -3,7 +3,7 @@
|
||||
This is a list of environment variable that control the runtime behavior of tinygrad and its examples.
|
||||
Most of these are self-explanatory, and are usually used to set an option at runtime.
|
||||
|
||||
Example: `DEV=CL DEBUG=4 python3 -m pytest`
|
||||
Example: `CL=1 DEBUG=4 python3 -m pytest`
|
||||
|
||||
However you can also decorate a function to set a value only inside that function.
|
||||
|
||||
@@ -31,7 +31,12 @@ These control the behavior of core tinygrad even when used as a library.
|
||||
Variable | Possible Value(s) | Description
|
||||
---|---|---
|
||||
DEBUG | [1-7] | enable debugging output (operations, timings, speed, generated code and more)
|
||||
DEV | [AMD, NV, ...] | enable a specific backend, see [below](#dev-variable)
|
||||
CL | [1] | enable OpenCL backend
|
||||
CUDA | [1] | enable CUDA backend
|
||||
AMD | [1] | enable AMD backend
|
||||
NV | [1] | enable NV backend
|
||||
METAL | [1] | enable Metal backend (for Mac M1 and after)
|
||||
CPU | [1] | enable CPU backend
|
||||
BEAM | [#] | number of beams in kernel beam search
|
||||
DEFAULT_FLOAT | [HALF, ...]| specify the default float dtype (FLOAT32, HALF, BFLOAT16, FLOAT64, ...), default to FLOAT32
|
||||
IMAGE | [1-2] | enable 2d specific optimizations
|
||||
@@ -43,23 +48,7 @@ ALLOW_TF32 | [1] | enable TensorFloat-32 tensor cores on Ampere
|
||||
WEBGPU_BACKEND | [WGPUBackendType_Metal, ...] | Force select a backend for WebGPU (Metal, DirectX, OpenGL, Vulkan...)
|
||||
CUDA_PATH | str | Use `CUDA_PATH/include` for CUDA headers for CUDA and NV backends. If not set, TinyGrad will use `/usr/local/cuda/include`, `/usr/include` and `/opt/cuda/include`.
|
||||
|
||||
### DEV variable
|
||||
|
||||
The `DEV` variable deserves special note due to its more nuanced syntax.
|
||||
`DEV` is used to specify the target device, target renderer and target architecture for said device, separated by colons.
|
||||
Specifying the renderer and architecture is optional, omitting a preference will cause tinygrad to automatically determine a suitable setting.
|
||||
The `DEV` variable may also be used to specify the interface through which to access the device (eg. `PCI`, `USB`). Interfaces may be specified preceding the target triple,
|
||||
separated by a plus (eg. `DEV=USB+AMD:LLVM`). Similarly as above, the interface may be omitted. Example usage follows:
|
||||
|
||||
`DEV` contents | Interpretation
|
||||
--- | ---
|
||||
AMD | use the AMD device
|
||||
AMD:LLVM | use the AMD device with the LLVM renderer
|
||||
NV:CUDA:sm_70 | use the NV device with the CUDA renderer targetting sm_70
|
||||
AMD::gfx950 | use the AMD device targetting gfx950
|
||||
USB+AMD | use the AMD device over the USB interface
|
||||
|
||||
### Debug breakdown
|
||||
## Debug breakdown
|
||||
|
||||
Variable | Value | Description
|
||||
---|---|---
|
||||
|
||||
+6
-6
@@ -1,16 +1,16 @@
|
||||
# Runtimes
|
||||
|
||||
tinygrad supports various runtimes, enabling your code to scale across a wide range of devices. The default runtime can be automatically selected based on the available hardware, or you can force a specific runtime to be default using environment variables (e.g., `DEV=CPU`).
|
||||
tinygrad supports various runtimes, enabling your code to scale across a wide range of devices. The default runtime can be automatically selected based on the available hardware, or you can force a specific runtime to be default using environment variables (e.g., `CPU=1`).
|
||||
|
||||
| Runtime | Description | Compiler Options | Requirements |
|
||||
|---------|-------------|------------------|--------------|
|
||||
| [NV](https://github.com/tinygrad/tinygrad/tree/master/tinygrad/runtime/ops_nv.py) | Provides acceleration for NVIDIA GPUs | nvrtc (default)<br>PTX (`DEV=NV:PTX`) | Ampere/Ada/Blackwell series GPUs.<br>You can select an interface via [the `DEV` variable](env_vars.md#dev-variable). See [NV interfaces](#nv-interfaces) for details. |
|
||||
| [AMD](https://github.com/tinygrad/tinygrad/tree/master/tinygrad/runtime/ops_amd.py) | Provides acceleration for AMD GPUs | LLVM (`DEV=AMD:LLVM`)<br>HIP/COMGR (`DEV=AMD:HIP`) | RDNA2 or newer GPUs.<br>You can select an interface via [the `DEV` variable](env_vars.md#dev-variable). See [AMD interfaces](#amd-interfaces) for details. |
|
||||
| [NV](https://github.com/tinygrad/tinygrad/tree/master/tinygrad/runtime/ops_nv.py) | Provides acceleration for NVIDIA GPUs | nvrtc (default)<br>PTX (`NV_PTX=1`) | Ampere/Ada/Blackwell series GPUs.<br>You can select an interface via `NV_IFACE=(NVK\|PCI)`. See [NV interfaces](#nv-interfaces) for details. |
|
||||
| [AMD](https://github.com/tinygrad/tinygrad/tree/master/tinygrad/runtime/ops_amd.py) | Provides acceleration for AMD GPUs | LLVM (`AMD_LLVM=1`)<br>HIP/COMGR (`AMD_HIP=1`) | RDNA2 or newer GPUs.<br>You can select an interface via `AMD_IFACE=(KFD\|PCI\|USB)`. See [AMD interfaces](#amd-interfaces) for details. |
|
||||
| [QCOM](https://github.com/tinygrad/tinygrad/tree/master/tinygrad/runtime/ops_qcom.py) | Provides acceleration for QCOM GPUs | - | 6xx series GPUs |
|
||||
| [METAL](https://github.com/tinygrad/tinygrad/tree/master/tinygrad/runtime/ops_metal.py) | Utilizes Metal for acceleration on Apple devices | - | M1+ Macs; Metal 3.0+ for `bfloat` support |
|
||||
| [CUDA](https://github.com/tinygrad/tinygrad/tree/master/tinygrad/runtime/ops_cuda.py) | Utilizes CUDA for acceleration on NVIDIA GPUs | nvrtc (default)<br> PTX (`DEV=CUDA:PTX`) | NVIDIA GPU with CUDA support |
|
||||
| [CUDA](https://github.com/tinygrad/tinygrad/tree/master/tinygrad/runtime/ops_cuda.py) | Utilizes CUDA for acceleration on NVIDIA GPUs | nvrtc (default)<br> PTX (`CUDA_PTX=1`) | NVIDIA GPU with CUDA support |
|
||||
| [CL](https://github.com/tinygrad/tinygrad/tree/master/tinygrad/runtime/ops_cl.py) | Accelerates computations using OpenCL on GPUs | - | OpenCL 2.0 compatible device |
|
||||
| [CPU](https://github.com/tinygrad/tinygrad/tree/master/tinygrad/runtime/ops_cpu.py) | Runs on CPU using the clang or llvm compiler | Clang JIT (default)<br>LLVM IR (`DEV=CPU:LLVM`) | `clang` compiler in system `PATH` |
|
||||
| [CPU](https://github.com/tinygrad/tinygrad/tree/master/tinygrad/runtime/ops_cpu.py) | Runs on CPU using the clang or llvm compiler | Clang JIT (default)<br>LLVM IR (`CPU_LLVM=1`) | `clang` compiler in system `PATH` |
|
||||
| [WEBGPU](https://github.com/tinygrad/tinygrad/tree/master/tinygrad/runtime/ops_webgpu.py) | Runs on GPU using the Dawn WebGPU engine (used in Google Chrome) | - | Dawn library installed and discoverable. Binaries: [pydawn v0.3.0](https://github.com/wpmed92/pydawn/releases/tag/v0.3.0) |
|
||||
|
||||
|
||||
@@ -72,7 +72,7 @@ AMD backend supports several interfaces for communicating with devices:
|
||||
* `PCI`: uses the [AM driver](developer/am.md)
|
||||
* `USB`: USB3 interface for asm24xx chips.
|
||||
|
||||
You can force an interface by setting the interface component of [the `DEV` environment variable](env_vars.md#dev-variable) to one of these values. When set to `PCI`, this may unbind your GPU from the amdgpu driver.
|
||||
You can force an interface by setting `AMD_IFACE` to one of these values. In the case of `AMD_IFACE=PCI`, this may unbind your GPU from the amdgpu driver.
|
||||
|
||||
## NV Interfaces
|
||||
NV backend supports several interfaces for communicating with devices:
|
||||
|
||||
@@ -1,59 +0,0 @@
|
||||
# TinyGPU
|
||||
|
||||
TinyGPU app lets you use AMD and NVIDIA GPUs on macOS over USB4/Thunderbolt with tinygrad.
|
||||
|
||||
## Requirements
|
||||
|
||||
- macOS (12.1+)
|
||||
- USB4/Thunderbolt port
|
||||
- A supported GPU (AMD RDNA3+ or NVIDIA Ampere+)
|
||||
|
||||
## Setup
|
||||
|
||||
### 1. Connect your GPU
|
||||
|
||||
Plug the supported GPU into your Mac over USB4/Thunderbolt.
|
||||
|
||||
### 2. Initiate the driver install
|
||||
|
||||
> **Note:** If tinygrad is cloned but not installed, run commands with `PYTHONPATH=.`
|
||||
|
||||
```bash
|
||||
curl -fsSL https://raw.githubusercontent.com/tinygrad/tinygrad/master/extra/setup_tinygpu_osx.sh | sh
|
||||
```
|
||||
|
||||
This downloads TinyGPU.app and triggers a system prompt to install the driver extension.
|
||||
|
||||
### 3. Enable the driver
|
||||
|
||||
You should see a system prompt: **"TinyGPU" would like to use a new driver extension**. Click **Open System Settings** and toggle TinyGPU on.
|
||||
|
||||
If you missed the prompt, go to **System Settings > General > Login Items & Extensions > Driver Extensions** and toggle TinyGPU on.
|
||||
|
||||
### 4. Compiler Setup
|
||||
|
||||
#### AMD
|
||||
|
||||
```bash
|
||||
curl -fsSL https://raw.githubusercontent.com/tinygrad/tinygrad/master/extra/setup_hipcomgr_osx.sh | sh
|
||||
```
|
||||
|
||||
#### NV
|
||||
|
||||
Install [Docker Desktop](https://www.docker.com/products/docker-desktop/) if you don't have it.
|
||||
|
||||
```bash
|
||||
curl -fsSL https://raw.githubusercontent.com/tinygrad/tinygrad/master/extra/setup_nvcc_osx.sh | sh
|
||||
```
|
||||
|
||||
Make sure `~/.local/bin` is on your `PATH`:
|
||||
|
||||
```bash
|
||||
export PATH="$HOME/.local/bin:$PATH"
|
||||
```
|
||||
|
||||
### 5. Use it!
|
||||
|
||||
```bash
|
||||
DEV={AMD|NV} python3 tinygrad/apps/llm.py
|
||||
```
|
||||
+14
-15
@@ -1,6 +1,6 @@
|
||||
# model based off https://medium.com/data-science/going-beyond-99-mnist-handwritten-digits-recognition-cfff96337392
|
||||
from typing import Callable
|
||||
from tinygrad import Tensor, TinyJit, nn, GlobalCounters, function
|
||||
from tinygrad import Tensor, TinyJit, nn, GlobalCounters
|
||||
from tinygrad.helpers import getenv, colored, trange
|
||||
from tinygrad.nn.datasets import mnist
|
||||
|
||||
@@ -15,31 +15,30 @@ class Model:
|
||||
nn.BatchNorm(64), Tensor.max_pool2d,
|
||||
lambda x: x.flatten(1), nn.Linear(576, 10)]
|
||||
|
||||
@function
|
||||
def __call__(self, x:Tensor) -> Tensor: return x.sequential(self.layers)
|
||||
|
||||
@TinyJit
|
||||
@Tensor.train()
|
||||
def train_step(self, X_train:Tensor, Y_train:Tensor) -> Tensor:
|
||||
opt.zero_grad()
|
||||
samples = Tensor.randint(getenv("BS", 512), high=X_train.shape[0])
|
||||
loss = self(X_train[samples]).sparse_categorical_crossentropy(Y_train[samples]).backward()
|
||||
return loss.realize(*opt.schedule_step())
|
||||
|
||||
@TinyJit
|
||||
def get_test_acc(self, X_test:Tensor, Y_test:Tensor) -> Tensor: return (self(X_test).argmax(axis=1) == Y_test).mean()*100
|
||||
|
||||
if __name__ == "__main__":
|
||||
X_train, Y_train, X_test, Y_test = mnist(fashion=getenv("FASHION"))
|
||||
|
||||
model = Model()
|
||||
opt = (nn.optim.Muon if getenv("MUON") else nn.optim.SGD if getenv("SGD") else nn.optim.Adam)(nn.state.get_parameters(model))
|
||||
|
||||
@TinyJit
|
||||
@Tensor.train()
|
||||
def train_step() -> Tensor:
|
||||
opt.zero_grad()
|
||||
samples = Tensor.randint(getenv("BS", 512), high=X_train.shape[0])
|
||||
loss = model(X_train[samples]).sparse_categorical_crossentropy(Y_train[samples]).backward()
|
||||
return loss.realize(*opt.schedule_step())
|
||||
|
||||
@TinyJit
|
||||
def get_test_acc() -> Tensor: return (model(X_test).argmax(axis=1) == Y_test).mean()*100
|
||||
|
||||
test_acc = float('nan')
|
||||
for i in (t:=trange(getenv("STEPS", 70))):
|
||||
GlobalCounters.reset() # NOTE: this makes it nice for DEBUG=2 timing
|
||||
loss = model.train_step(X_train, Y_train)
|
||||
if i%10 == 9: test_acc = model.get_test_acc(X_test, Y_test).item()
|
||||
loss = train_step()
|
||||
if i%10 == 9: test_acc = get_test_acc().item()
|
||||
t.set_description(f"loss: {loss.item():6.2f} test_accuracy: {test_acc:5.2f}%")
|
||||
|
||||
# verify eval acc
|
||||
|
||||
@@ -5,7 +5,7 @@ from extra.onnx_helpers import get_example_inputs, validate
|
||||
|
||||
def load_onnx_model(onnx_file):
|
||||
run_onnx = OnnxRunner(onnx_file)
|
||||
run_onnx_jit = TinyJit(lambda **kwargs: next(iter(run_onnx({k:v.to(None) for k,v in kwargs.items()}).values())), prune=True)
|
||||
run_onnx_jit = TinyJit(lambda **kwargs: next(iter(run_onnx({k:v.to(None) for k,v in kwargs.items()}).values())), prune=True, optimize=True)
|
||||
return run_onnx_jit, run_onnx.graph_inputs
|
||||
|
||||
if __name__ == "__main__":
|
||||
|
||||
@@ -1,10 +1,9 @@
|
||||
from pathlib import Path
|
||||
from extra.models.efficientnet import EfficientNet
|
||||
from tinygrad.tensor import Tensor
|
||||
from tinygrad.device import Device
|
||||
from tinygrad.nn.state import get_state_dict, safe_save, safe_load, load_state_dict
|
||||
from extra.export_model import export_model
|
||||
from tinygrad.helpers import fetch
|
||||
from tinygrad.helpers import getenv, fetch
|
||||
import ast
|
||||
|
||||
if __name__ == "__main__":
|
||||
@@ -13,13 +12,13 @@ if __name__ == "__main__":
|
||||
dirname = Path(__file__).parent
|
||||
# exporting a model that's loaded from safetensors doesn't work without loading in from safetensors first
|
||||
# loading the state dict from a safetensor file changes the generated kernels
|
||||
if Device.DEFAULT == "WEBGPU":
|
||||
if getenv("WEBGPU"):
|
||||
safe_save(get_state_dict(model), (dirname / "net.safetensors").as_posix())
|
||||
load_state_dict(model, safe_load(str(dirname / "net.safetensors")))
|
||||
mode = "clang" if Device.DEFAULT == "CPU" else "webgpu" if Device.DEFAULT == "WEBGPU" else ""
|
||||
mode = "clang" if getenv("CPU", "") != "" else "webgpu" if getenv("WEBGPU", "") != "" else ""
|
||||
prg, inp_sizes, out_sizes, state = export_model(model, mode, Tensor.randn(1,3,224,224))
|
||||
if Device.DEFAULT != "CPU":
|
||||
ext = "js" if Device.DEFAULT == "WEBGPU" else "json"
|
||||
if getenv("CPU", "") == "":
|
||||
ext = "js" if getenv("WEBGPU", "") != "" else "json"
|
||||
with open(dirname / f"net.{ext}", "w") as text_file:
|
||||
text_file.write(prg)
|
||||
else:
|
||||
@@ -69,6 +68,6 @@ if __name__ == "__main__":
|
||||
else printf("%s\\n", lbls[best_idx]);
|
||||
}""")
|
||||
|
||||
# DEV=CPU python3 examples/compile_efficientnet.py | clang -O2 -lm -x c - -o recognize && DEBUG=1 time ./recognize docs/showcase/stable_diffusion_by_tinygrad.jpg
|
||||
# CPU=1 python3 examples/compile_efficientnet.py | clang -O2 -lm -x c - -o recognize && DEBUG=1 time ./recognize docs/showcase/stable_diffusion_by_tinygrad.jpg
|
||||
# category : 281 (tabby, tabby cat) with 9.452788
|
||||
print('\n'.join(cprog))
|
||||
|
||||
@@ -2,14 +2,13 @@
|
||||
import os
|
||||
if "NOOPT" not in os.environ: os.environ["NOOPT"] = "1"
|
||||
from tinygrad import Device, nn, Tensor, dtypes
|
||||
Device.DEFAULT = "CPU"
|
||||
from train_gpt2 import GPT, GPTConfig
|
||||
from tinygrad.helpers import DEV, dedup, flatten, getenv, GlobalCounters, to_function_name
|
||||
from tinygrad.helpers import dedup, flatten, getenv, GlobalCounters, to_function_name
|
||||
from tinygrad.engine.realize import get_kernel
|
||||
from tinygrad.engine.memory import memory_planner
|
||||
from tinygrad.uop.ops import Ops
|
||||
|
||||
DEV.value = "CPU"
|
||||
|
||||
TIMING = getenv("TIMING")
|
||||
|
||||
if __name__ == "__main__":
|
||||
|
||||
@@ -325,18 +325,19 @@ def eval_stable_diffusion():
|
||||
# NOTE: the clip weights are the same between model.cond_stage_model and clip_encoder
|
||||
eval_timesteps = list(reversed(range(1, 1000, 20)))
|
||||
|
||||
with Context(DEV="CPU"):
|
||||
# The choice of alphas_prev[0] = alphas_cumprod[0] seems arbitrary, but it's how the mlperf ref does it:
|
||||
# alphas_prev = np.asarray([alphacums[0]] + alphacums[ddim_timesteps[:-1]].tolist())
|
||||
eval_alphas_prev = model.alphas_cumprod[0:1].cat(model.alphas_cumprod[list(range(1, 1000, 20))[:-1]]).to(GPUS).realize()
|
||||
inception = FidInceptionV3().load_from_pretrained(CKPTDIR / "inception" / "pt_inception-2015-12-05-6726825d.pth")
|
||||
vision_cfg = {'width': 1280, 'layers': 32, 'd_head': 80, 'image_size': 224, 'patch_size': 14}
|
||||
text_cfg = {'width': 1024, 'n_heads': 16, 'layers': 24, 'vocab_size': 49408, 'ctx_length': 77}
|
||||
clip.gelu = gelu_erf
|
||||
clip_encoder = OpenClipEncoder(1024, text_cfg, vision_cfg)
|
||||
loaded = torch_load(CKPTDIR / "clip" / "open_clip_pytorch_model.bin")
|
||||
loaded.update({"attn_mask": clip_encoder.attn_mask, "mean": clip_encoder.mean, "std": clip_encoder.std})
|
||||
load_state_dict(clip_encoder, loaded)
|
||||
original_device, Device.DEFAULT = Device.DEFAULT, "CPU"
|
||||
# The choice of alphas_prev[0] = alphas_cumprod[0] seems arbitrary, but it's how the mlperf ref does it:
|
||||
# alphas_prev = np.asarray([alphacums[0]] + alphacums[ddim_timesteps[:-1]].tolist())
|
||||
eval_alphas_prev = model.alphas_cumprod[0:1].cat(model.alphas_cumprod[list(range(1, 1000, 20))[:-1]]).to(GPUS).realize()
|
||||
inception = FidInceptionV3().load_from_pretrained(CKPTDIR / "inception" / "pt_inception-2015-12-05-6726825d.pth")
|
||||
vision_cfg = {'width': 1280, 'layers': 32, 'd_head': 80, 'image_size': 224, 'patch_size': 14}
|
||||
text_cfg = {'width': 1024, 'n_heads': 16, 'layers': 24, 'vocab_size': 49408, 'ctx_length': 77}
|
||||
clip.gelu = gelu_erf
|
||||
clip_encoder = OpenClipEncoder(1024, text_cfg, vision_cfg)
|
||||
loaded = torch_load(CKPTDIR / "clip" / "open_clip_pytorch_model.bin")
|
||||
loaded.update({"attn_mask": clip_encoder.attn_mask, "mean": clip_encoder.mean, "std": clip_encoder.std})
|
||||
load_state_dict(clip_encoder, loaded)
|
||||
Device.DEFAULT=original_device
|
||||
|
||||
@TinyJit
|
||||
def denoise_step(x:Tensor, x_x:Tensor, t_t:Tensor, uc_c:Tensor, sqrt_alphas_cumprod_t:Tensor, sqrt_one_minus_alphas_cumprod_t:Tensor,
|
||||
|
||||
+64
-103
@@ -1282,14 +1282,11 @@ def train_bert():
|
||||
previous_step = i
|
||||
|
||||
def train_llama3():
|
||||
from examples.mlperf.models.flat_llama import FlatTransformer, apply_grad, FP8
|
||||
from examples.mlperf.models.llama import Transformer
|
||||
from examples.llama3 import MODEL_PARAMS
|
||||
from examples.mlperf.lr_schedulers import CosineAnnealingLRWithWarmup
|
||||
from examples.mlperf.optim import GradAccClipAdamW
|
||||
|
||||
INITMLPERF = getenv("INITMLPERF")
|
||||
RUNMLPERF = getenv("RUNMLPERF")
|
||||
LOGMLPERF = getenv("LOGMLPERF")
|
||||
BENCHMARK = getenv("BENCHMARK")
|
||||
|
||||
config = {}
|
||||
@@ -1312,60 +1309,15 @@ def train_llama3():
|
||||
EVAL_BS = config["EVAL_BS"] = getenv("EVAL_BS", 16)
|
||||
EVAL_TARGET = config["EVAL_TARGET"] = getenv("EVAL_TARGET", 5.6)
|
||||
|
||||
if LOGMLPERF:
|
||||
from mlperf_logging import mllog
|
||||
import mlperf_logging.mllog.constants as mllog_constants
|
||||
|
||||
mllog.config(filename=f"result_llama31_{SEED}.log")
|
||||
mllog.config(root_dir=Path(__file__).parents[3].as_posix())
|
||||
MLLOGGER = mllog.get_mllogger()
|
||||
MLLOGGER.logger.propagate = False
|
||||
|
||||
LLAMA_BENCHMARK = mllog_constants.LLAMA31_405B if getenv("LLAMA3_SIZE", "8B") == "405B" else mllog_constants.LLAMA31_8B
|
||||
|
||||
if INITMLPERF:
|
||||
assert BENCHMARK, "BENCHMARK must be set for INITMLPERF"
|
||||
MLLOGGER.event(key=mllog_constants.SUBMISSION_ORG, value="tinycorp")
|
||||
MLLOGGER.event(key=mllog_constants.SUBMISSION_PLATFORM, value=getenv("SUBMISSION_PLATFORM", "tinybox"))
|
||||
MLLOGGER.event(key=mllog_constants.SUBMISSION_DIVISION, value=mllog_constants.CLOSED)
|
||||
MLLOGGER.event(key=mllog_constants.SUBMISSION_STATUS, value=mllog_constants.ONPREM)
|
||||
|
||||
MLLOGGER.event(key=mllog_constants.SUBMISSION_BENCHMARK, value=LLAMA_BENCHMARK)
|
||||
|
||||
diskcache_clear()
|
||||
MLLOGGER.event(key=mllog_constants.CACHE_CLEAR, value=True)
|
||||
MLLOGGER.start(key=mllog_constants.INIT_START, value=None)
|
||||
|
||||
if RUNMLPERF:
|
||||
MLLOGGER.start(key=mllog_constants.RUN_START, value=None)
|
||||
MLLOGGER.event(key=mllog_constants.SEED, value=SEED)
|
||||
|
||||
MLLOGGER.event(key=mllog_constants.GLOBAL_BATCH_SIZE, value=GBS)
|
||||
MLLOGGER.event(key=mllog_constants.MAX_SEQUENCE_LENGTH, value=SEQLEN)
|
||||
MLLOGGER.event(key=mllog_constants.MAX_STEPS, value=MAX_STEPS)
|
||||
MLLOGGER.event(key=mllog_constants.GRADIENT_ACCUMULATION_STEPS, value=grad_acc)
|
||||
MLLOGGER.event(key=mllog_constants.EVAL_SAMPLES, value=EVAL_SAMPLES)
|
||||
MLLOGGER.event(key=mllog_constants.TRAIN_SAMPLES, value=SAMPLES)
|
||||
|
||||
MLLOGGER.event(key=mllog_constants.OPT_NAME, value=mllog_constants.ADAMW)
|
||||
MLLOGGER.event(key=mllog_constants.OPT_BASE_LR, value=LR)
|
||||
MLLOGGER.event(key=mllog_constants.OPT_END_LR, value=END_LR)
|
||||
MLLOGGER.event(key=mllog_constants.OPT_ADAMW_BETA_1, value=0.9)
|
||||
MLLOGGER.event(key=mllog_constants.OPT_ADAMW_BETA_2, value=0.95)
|
||||
MLLOGGER.event(key=mllog_constants.OPT_ADAMW_EPSILON, value=1e-5)
|
||||
MLLOGGER.event(key=mllog_constants.OPT_ADAMW_WEIGHT_DECAY, value=0.1)
|
||||
MLLOGGER.event(key=mllog_constants.OPT_LR_WARMUP_STEPS, value=WARMUP_STEPS)
|
||||
MLLOGGER.event(key=mllog_constants.NUM_WARMUP_STEPS, value=WARMUP_STEPS)
|
||||
MLLOGGER.event(key=mllog_constants.OPT_LR_DECAY_STEPS, value=MAX_STEPS - WARMUP_STEPS)
|
||||
MLLOGGER.event(key=mllog_constants.OPT_GRADIENT_CLIP_NORM, value=1.0)
|
||||
else:
|
||||
MLLOGGER = None
|
||||
# LR=1e-4 TRAIN_ON_VAL=1 DEFAULT_FLOAT=bfloat16 JITBEAM=2 OPTIM_DTYPE=bfloat16 LLAMA3_SIZE=1B WARMUP_STEPS=36 DECAY_STEPS=360 SEQLEN=512 PYTHONPATH=. AMD=1 AMD_LLVM=0 MODEL=llama3 python3 examples/mlperf/model_train.py
|
||||
# trains to 7
|
||||
|
||||
opt_adamw_beta_1 = 0.9
|
||||
opt_adamw_beta_2 = 0.95
|
||||
opt_adamw_epsilon = 1e-5
|
||||
opt_adamw_weight_decay = 0.1
|
||||
|
||||
opt_gradient_clip_norm = 1.0
|
||||
opt_learning_rate_warmup_steps = WARMUP_STEPS
|
||||
opt_learning_rate_decay_steps = MAX_STEPS - opt_learning_rate_warmup_steps
|
||||
opt_base_learning_rate = LR
|
||||
@@ -1391,35 +1343,59 @@ def train_llama3():
|
||||
if (MP := getenv("MP", 1)) > 1: model_params['vocab_size'] = round_up(model_params['vocab_size'], 256 * MP)
|
||||
vocab_mask:Tensor = Tensor.arange(model_params['vocab_size']).reshape(1, 1, -1) >= real_vocab_size
|
||||
|
||||
model = FlatTransformer(**model_params, max_context=SEQLEN)
|
||||
model = Transformer(**model_params, max_context=SEQLEN)
|
||||
|
||||
params = get_parameters(model)
|
||||
# weights are all bfloat16 for now
|
||||
assert params and all(p.dtype == dtypes.bfloat16 for p in params)
|
||||
|
||||
if getenv("FAKEDATA"):
|
||||
for v in get_parameters(model):
|
||||
v = v.assign(Tensor.empty(v.shape, dtype=v.dtype))
|
||||
v = v.assign(Tensor.empty(v.shape))
|
||||
|
||||
is_dp = (DP := getenv("DP", 1)) > 1
|
||||
is_mp = (MP := getenv("MP", 1)) > 1
|
||||
is_sharding = is_dp or is_mp
|
||||
device_count = max(DP, MP)
|
||||
device = tuple(f"{Device.DEFAULT}:{i}" for i in range(device_count))
|
||||
if (DP := getenv("DP", 1)) > 1:
|
||||
device = tuple(f"{Device.DEFAULT}:{i}" for i in range(DP))
|
||||
for v in get_parameters(model):
|
||||
v.shard_(device, axis=None)
|
||||
|
||||
model.shard(device, is_mp)
|
||||
vocab_mask.shard_(device, axis=None)
|
||||
|
||||
if is_dp: vocab_mask.shard_(device, axis=None).realize()
|
||||
if is_mp: vocab_mask.shard_(device, axis=2).realize()
|
||||
if (MP := getenv("MP", 1)) > 1:
|
||||
device = tuple(f"{Device.DEFAULT}:{i}" for i in range(MP))
|
||||
for k,v in get_state_dict(model).items():
|
||||
if 'scale' in k: v.shard_(device, axis=None) # from quantized
|
||||
elif '.attention.wq' in k: v.shard_(device, axis=0)
|
||||
elif '.attention.wk' in k: v.shard_(device, axis=0)
|
||||
elif '.attention.wv' in k: v.shard_(device, axis=0)
|
||||
elif '.attention.wqkv' in k: v.shard_(device, axis=0)
|
||||
elif '.attention.wo' in k: v.shard_(device, axis=1)
|
||||
elif '.feed_forward.w1.' in k: v.shard_(device, axis=0)
|
||||
elif '.feed_forward.w2.' in k: v.shard_(device, axis=1)
|
||||
elif '.feed_forward.w3.' in k: v.shard_(device, axis=0)
|
||||
elif 'tok_embeddings.weight' in k: v.shard_(device, axis=0)
|
||||
elif 'output.weight' in k: v.shard_(device, axis=0)
|
||||
else:
|
||||
# attention_norm, ffn_norm, norm
|
||||
v.shard_(device, axis=None)
|
||||
# prevents memory spike on device 0
|
||||
v.realize()
|
||||
|
||||
vocab_mask.shard_(device, axis=2).realize()
|
||||
|
||||
is_offload_optim = bool(getenv("OFFLOAD_OPTIM"))
|
||||
is_fake_offload = Device.DEFAULT == "NULL"
|
||||
optim_device = ("CPU" if not is_fake_offload else "NULL:99") if is_offload_optim else None
|
||||
optim = GradAccClipAdamW(params, lr=0.0, b1=opt_adamw_beta_1, b2=opt_adamw_beta_2,
|
||||
optim = GradAccClipAdamW(get_parameters(model), lr=0.0, b1=opt_adamw_beta_1, b2=opt_adamw_beta_2,
|
||||
eps=opt_adamw_epsilon, weight_decay=opt_adamw_weight_decay, grad_acc=grad_acc, device=optim_device)
|
||||
|
||||
# init grads
|
||||
for p in optim.params:
|
||||
p.grad = Tensor.zeros(p.shape, dtype=p.dtype, device=p.device).contiguous()
|
||||
grads = [p.grad for p in optim.params]
|
||||
if is_offload_optim:
|
||||
for p in optim.params:
|
||||
p.grad = Tensor.zeros(p.shape, dtype=p.dtype, device=optim_device, requires_grad=False).contiguous().realize()
|
||||
else:
|
||||
for p in optim.params:
|
||||
p.grad = p.zeros_like().contiguous().realize()
|
||||
grads: list[Tensor] = [p.grad for p in optim.params]
|
||||
|
||||
scheduler = CosineAnnealingLRWithWarmup(optim, opt_base_learning_rate, opt_end_learning_rate, opt_learning_rate_warmup_steps, opt_learning_rate_decay_steps)
|
||||
|
||||
@@ -1432,28 +1408,30 @@ def train_llama3():
|
||||
print(f"loading optim checkpoint from {fn}")
|
||||
load_state_dict(scheduler, safe_load(fn), realize=False)
|
||||
|
||||
fp8_amax = [t for ts in model._fp8_amax.values() for t in ts] if FP8 else []
|
||||
|
||||
@TinyJit
|
||||
def minibatch(tokens:Tensor):
|
||||
if is_dp: tokens = tokens.to(None).shard(device, 0)
|
||||
if is_mp: tokens = tokens.shard(device)
|
||||
if not is_sharding: tokens = tokens.to(None)
|
||||
if (DP := getenv("DP", 1)) > 1:
|
||||
device = tuple(f"{Device.DEFAULT}:{i}" for i in range(DP))
|
||||
tokens = tokens.to(None).shard(device, 0)
|
||||
if (MP := getenv("MP", 1)) > 1:
|
||||
device = tuple(f"{Device.DEFAULT}:{i}" for i in range(MP))
|
||||
tokens = tokens.shard(device)
|
||||
if DP == 1 and MP == 1: tokens = tokens.to(None)
|
||||
logits:Tensor = model(tokens[:, :-1])
|
||||
loss = vocab_mask.where(-1e9, logits).sparse_categorical_crossentropy(tokens[:, 1:])
|
||||
|
||||
for g, new_g in zip(grads, loss.gradient(*optim.params)):
|
||||
apply_grad(g, new_g.uop)
|
||||
|
||||
loss.backward()
|
||||
assert all(p.grad is g for p,g in zip(optim.params, grads))
|
||||
loss_cpu = loss.flatten().float().to("CPU")
|
||||
return loss_cpu.realize(*grads, *fp8_amax)
|
||||
Tensor.realize(loss_cpu, *grads)
|
||||
return loss_cpu
|
||||
|
||||
@TinyJit
|
||||
def optim_step():
|
||||
grad_norm = optim.fstep(grads)
|
||||
scheduler.step()
|
||||
|
||||
for g in grads: g.assign(g.zeros_like())
|
||||
for g in grads:
|
||||
g.assign(g.zeros_like())
|
||||
|
||||
lr_cpu = optim.lr.float().to("CPU")
|
||||
grad_norm_cpu = grad_norm.float().to("CPU")
|
||||
@@ -1464,9 +1442,13 @@ def train_llama3():
|
||||
@TinyJit
|
||||
@Tensor.train(False)
|
||||
def eval_step(tokens:Tensor):
|
||||
if is_dp: tokens = tokens.to(None).shard(device, 0)
|
||||
if is_mp: tokens = tokens.shard(device)
|
||||
if not is_sharding: tokens = tokens.to(None)
|
||||
if (DP := getenv("DP", 1)) > 1:
|
||||
device = tuple(f"{Device.DEFAULT}:{i}" for i in range(DP))
|
||||
tokens = tokens.to(None).shard(device, 0)
|
||||
if (MP := getenv("MP", 1)) > 1:
|
||||
device = tuple(f"{Device.DEFAULT}:{i}" for i in range(MP))
|
||||
tokens = tokens.shard(device)
|
||||
if DP == 1 and MP == 1: tokens = tokens.to(None)
|
||||
logits:Tensor = model(tokens[:, :-1])
|
||||
loss = vocab_mask.where(-1e9, logits).sparse_categorical_crossentropy(tokens[:, 1:])
|
||||
return loss.flatten().float().to("CPU")
|
||||
@@ -1501,11 +1483,6 @@ def train_llama3():
|
||||
train_iter = get_train_iter()
|
||||
i, sequences_seen = resume_ckpt, 0
|
||||
step_times = []
|
||||
|
||||
if MLLOGGER and RUNMLPERF:
|
||||
MLLOGGER.start(key=mllog_constants.EPOCH_START, metadata={mllog_constants.SAMPLES_COUNT: sequences_seen})
|
||||
MLLOGGER.start(key=mllog_constants.BLOCK_START, metadata={mllog_constants.SAMPLES_COUNT: sequences_seen})
|
||||
|
||||
while i < MAX_STEPS:
|
||||
GlobalCounters.reset()
|
||||
actual_gbs = GBS if i >= 2 else BS
|
||||
@@ -1544,7 +1521,7 @@ def train_llama3():
|
||||
|
||||
mem_gb = GlobalCounters.mem_used / 1e9
|
||||
gflops = GlobalCounters.global_ops / 1e9 / dev_time
|
||||
mfu = ((6 * num_params * SEQLEN * GBS) / (dev_time * device_count * 2.3e15)) * 100
|
||||
mfu = ((6 * num_params * SEQLEN * GBS) / (dev_time * max(getenv("DP", 1), getenv("MP", 1)) * 2.3e15)) * 100
|
||||
tqdm.write(
|
||||
f"{i:5} {step_time:.3f} s step, {gbs_time:.3f} s gbs, {optim_time:.3f} s optim, {data_time:.3f} s data, {loss:.4f} loss, " \
|
||||
f"{lr:.12f} LR, {grad_norm:.6f} grad_norm, {mem_gb:.2f} GB used, {gflops:9.2f} GFLOPS, {mfu:5.2f}% MFU")
|
||||
@@ -1588,10 +1565,6 @@ def train_llama3():
|
||||
tqdm.write(f"evaluating after {sequences_seen} sequences")
|
||||
profile_marker(f"eval @ {i}")
|
||||
|
||||
if MLLOGGER and RUNMLPERF:
|
||||
MLLOGGER.end(key=mllog_constants.BLOCK_STOP, metadata={mllog_constants.SAMPLES_COUNT: sequences_seen})
|
||||
MLLOGGER.start(key=mllog_constants.EVAL_START, metadata={mllog_constants.SAMPLES_COUNT: sequences_seen})
|
||||
|
||||
# run eval
|
||||
eval_losses = []
|
||||
eval_iter = get_eval_iter()
|
||||
@@ -1601,34 +1574,22 @@ def train_llama3():
|
||||
eval_losses += eval_step(tokens).tolist()
|
||||
|
||||
if BENCHMARK and (j+1) == min(BENCHMARK, EVAL_SAMPLES//EVAL_BS):
|
||||
if MLLOGGER and INITMLPERF:
|
||||
MLLOGGER.end(key=mllog_constants.INIT_STOP, value=None)
|
||||
return
|
||||
|
||||
log_perplexity = sum(eval_losses) / len(eval_losses)
|
||||
log_perplexity = Tensor(eval_losses).mean().float().item()
|
||||
|
||||
tqdm.write(f"eval log perplexity: {log_perplexity:.4f}")
|
||||
|
||||
if MLLOGGER and RUNMLPERF:
|
||||
MLLOGGER.event(key=mllog_constants.EVAL_ACCURACY, value=log_perplexity, metadata={mllog_constants.SAMPLES_COUNT: sequences_seen})
|
||||
MLLOGGER.end(key=mllog_constants.EVAL_STOP, metadata={mllog_constants.SAMPLES_COUNT: sequences_seen})
|
||||
|
||||
if WANDB:
|
||||
wandb.log({"eval/log_perplexity": log_perplexity, "eval/sequences_seen": sequences_seen})
|
||||
|
||||
if log_perplexity < EVAL_TARGET:
|
||||
tqdm.write(f"target achieved after {sequences_seen} sequences")
|
||||
if MLLOGGER and RUNMLPERF:
|
||||
MLLOGGER.end(key=mllog_constants.EPOCH_STOP, metadata={mllog_constants.SAMPLES_COUNT: sequences_seen})
|
||||
MLLOGGER.event(key=mllog_constants.TRAIN_SAMPLES, value=sequences_seen)
|
||||
MLLOGGER.end(key=mllog_constants.RUN_STOP, metadata={mllog_constants.STATUS: mllog_constants.SUCCESS})
|
||||
if getenv("CKPT"):
|
||||
if not os.path.exists(ckpt_dir := "./ckpts"): os.mkdir(ckpt_dir)
|
||||
fn = f"{ckpt_dir}/llama3.safe"
|
||||
safe_save(get_state_dict(model), fn)
|
||||
break
|
||||
if MLLOGGER and RUNMLPERF:
|
||||
MLLOGGER.start(key=mllog_constants.BLOCK_START, metadata={mllog_constants.SAMPLES_COUNT: sequences_seen})
|
||||
|
||||
def train_stable_diffusion():
|
||||
from extra.models.unet import UNetModel
|
||||
|
||||
@@ -1,263 +0,0 @@
|
||||
import math, os
|
||||
if __name__ == "__main__":
|
||||
os.environ["DEFAULT_FLOAT"] = "bfloat16"
|
||||
os.environ["OPTIM_DTYPE"] = "bfloat16"
|
||||
if "DEV" not in os.environ: os.environ["DEV"] = "NULL"
|
||||
# CDNA
|
||||
os.environ["EMULATE"] = "AMD_CDNA4"
|
||||
os.environ["DEVICE_IN_FUNCTION_BUG"] = "1"
|
||||
os.environ["ALL2ALL"] = "1"
|
||||
os.environ["USE_ATOMICS"] = "1"
|
||||
if "HK_FLASH_ATTENTION" not in os.environ:
|
||||
os.environ["HK_FLASH_ATTENTION"] = "1"
|
||||
if "ASM_GEMM" not in os.environ:
|
||||
os.environ["ASM_GEMM"] = "1"
|
||||
from tinygrad import Tensor, nn, function, getenv, dtypes, TinyJit
|
||||
from tinygrad.helpers import Timing, colored, GlobalCounters, profile_marker
|
||||
from tinygrad.uop.ops import Ops, UOp
|
||||
from extra.models.llama import apply_rotary_emb, precompute_freqs_cis
|
||||
|
||||
FP8 = getenv("FP8", 0)
|
||||
WQKV = getenv("WQKV", 0)
|
||||
|
||||
FP8_DTYPE = dtypes.fp8e4m3
|
||||
FP8_GRAD_DTYPE = dtypes.fp8e5m2
|
||||
FP8_MAX = 448.0
|
||||
|
||||
def quantize_fp8(x:Tensor, amax_state:Tensor|None=None):
|
||||
if amax_state is not None:
|
||||
scale = FP8_MAX / (amax_state + 1e-8)
|
||||
amax_state.assign(x.abs().max().detach())
|
||||
else:
|
||||
scale = FP8_MAX / (x.abs().max().detach() + 1e-8)
|
||||
x_scaled = x * scale
|
||||
x_clamped = x_scaled + (x_scaled.detach().clamp(-FP8_MAX, FP8_MAX) - x_scaled.detach()) # STE
|
||||
return x_clamped.cast(FP8_DTYPE), scale.float().reciprocal()
|
||||
|
||||
def matmul(x:Tensor, w:Tensor, fp8=FP8, amax_x:Tensor|None=None, amax_w:Tensor|None=None) -> Tensor:
|
||||
if not fp8: return x @ w.T
|
||||
from tinygrad.helpers import ASM_GEMM
|
||||
x_fp8, x_scale = quantize_fp8(x, amax_state=amax_x)
|
||||
w_fp8, w_scale = quantize_fp8(w, amax_state=amax_w)
|
||||
combined_scale = x_scale * w_scale
|
||||
if ASM_GEMM:
|
||||
from extra.gemm.cdna_asm_gemm import can_use_asm_gemm, asm_gemm
|
||||
if can_use_asm_gemm(x_fp8, w_fp8.T): return asm_gemm(x_fp8, w_fp8.T, combined_scale=combined_scale)
|
||||
return x_fp8.dot(w_fp8.T, dtype=dtypes.float) * combined_scale
|
||||
|
||||
def rmsnorm(x_in:Tensor, eps:float):
|
||||
x = x_in.float()
|
||||
x = x * (x.square().mean(-1, keepdim=True) + eps).rsqrt()
|
||||
return x.cast(x_in.dtype)
|
||||
|
||||
class FlatTransformer:
|
||||
def __init__(self, dim:int, hidden_dim:int, n_heads:int, n_layers:int, norm_eps:float, vocab_size:int, n_kv_heads:int|None=None,
|
||||
rope_theta:int=10000, max_context:int=1024):
|
||||
self.vocab_size = vocab_size
|
||||
self.n_layers = n_layers
|
||||
self.n_heads = n_heads
|
||||
self.n_kv_heads = n_kv_heads if n_kv_heads is not None else n_heads # n_kv_heads != n_heads implies MQA [arxiv/2307.09288, A.2.1]
|
||||
self.head_dim = dim // n_heads
|
||||
self.n_rep = self.n_heads // self.n_kv_heads
|
||||
|
||||
scaled_std = 0.02 / math.sqrt(2 * n_layers)
|
||||
|
||||
# Attention
|
||||
if WQKV:
|
||||
self.wqkv = self.lin_per_layer(dim, self.n_heads * self.head_dim + self.n_kv_heads * self.head_dim * 2)
|
||||
else:
|
||||
self.wq = self.lin_per_layer(dim, self.n_heads * self.head_dim)
|
||||
self.wk = self.lin_per_layer(dim, self.n_kv_heads * self.head_dim)
|
||||
self.wv = self.lin_per_layer(dim, self.n_kv_heads * self.head_dim)
|
||||
self.wo = self.lin_per_layer(self.n_heads * self.head_dim, dim, std=scaled_std)
|
||||
|
||||
# FeedForward
|
||||
self.w1 = self.lin_per_layer(dim, hidden_dim)
|
||||
self.w2 = self.lin_per_layer(hidden_dim, dim, std=scaled_std)
|
||||
self.w3 = self.lin_per_layer(dim, hidden_dim)
|
||||
|
||||
self.norm_eps = norm_eps
|
||||
self.attention_norm = Tensor.ones(n_layers, dim).contiguous()
|
||||
self.ffn_norm = Tensor.ones(n_layers, dim).contiguous()
|
||||
|
||||
# output
|
||||
self.norm = nn.RMSNorm(dim, norm_eps)
|
||||
self.tok_embeddings = nn.Embedding(vocab_size, dim)
|
||||
self.tok_embeddings.weight = Tensor.normal(vocab_size, dim, mean=0.0, std=0.02, dtype=dtypes.bfloat16)
|
||||
self.output = Tensor.normal(1, vocab_size, dim, mean=0.0, std=0.02, dtype=dtypes.bfloat16)
|
||||
self.freqs_cis = precompute_freqs_cis(dim // n_heads, max_context * 2, rope_theta).contiguous().requires_grad_(False)
|
||||
|
||||
if FP8:
|
||||
def _amax(): return Tensor.full((), FP8_MAX).contiguous().requires_grad_(False)
|
||||
names = (["xqkv", "wqkv"] if WQKV else ["xq", "wq", "xk", "wk", "xv", "wv"]) + \
|
||||
["xo", "wo", "x1", "w1", "x2", "w2", "x3", "w3"]
|
||||
# _fp8_amax[name][layer_idx] = scalar amax tensor
|
||||
self._fp8_amax = {name: [_amax() for _ in range(n_layers)] for name in names}
|
||||
self._fp8_amax["xout"] = [_amax()]
|
||||
self._fp8_amax["wout"] = [_amax()]
|
||||
|
||||
def lin_per_layer(self, in_features:int, out_features:int, std:float=0.02):
|
||||
if getenv("ZEROS"): return Tensor.zeros(self.n_layers, out_features, in_features)
|
||||
return Tensor.normal(self.n_layers, out_features, in_features, mean=0.0, std=std)
|
||||
|
||||
def attention(self, x:Tensor, freqs_cis:Tensor, attention_norm:Tensor, wo:Tensor, wqkv:Tensor|None=None,
|
||||
wq:Tensor|None=None, wk:Tensor|None=None, wv:Tensor|None=None,
|
||||
amax_xqkv=None, amax_wqkv=None, amax_xq=None, amax_wq=None, amax_xk=None, amax_wk=None,
|
||||
amax_xv=None, amax_wv=None, amax_xo=None, amax_wo=None):
|
||||
x = rmsnorm(x, self.norm_eps) * attention_norm
|
||||
bsz, seqlen, _ = x.shape
|
||||
|
||||
if wqkv is not None:
|
||||
xqkv = matmul(x, wqkv, amax_x=amax_xqkv, amax_w=amax_wqkv)
|
||||
xqkv = xqkv.reshape(bsz, seqlen, self.n_kv_heads, self.n_rep + 2, self.head_dim)
|
||||
xq = xqkv[:, :, :, :self.n_rep].reshape(bsz, seqlen, self.n_heads, self.head_dim)
|
||||
xk = xqkv[:, :, :, self.n_rep].reshape(bsz, seqlen, self.n_kv_heads, self.head_dim)
|
||||
xv = xqkv[:, :, :, self.n_rep+1].reshape(bsz, seqlen, self.n_kv_heads, self.head_dim)
|
||||
else:
|
||||
assert wq is not None and wk is not None and wv is not None
|
||||
xq = matmul(x, wq, amax_x=amax_xq, amax_w=amax_wq).reshape(bsz, seqlen, self.n_heads, self.head_dim)
|
||||
xk = matmul(x, wk, amax_x=amax_xk, amax_w=amax_wk).reshape(bsz, seqlen, self.n_kv_heads, self.head_dim)
|
||||
xv = matmul(x, wv, amax_x=amax_xv, amax_w=amax_wv).reshape(bsz, seqlen, self.n_kv_heads, self.head_dim)
|
||||
|
||||
xq, xk = apply_rotary_emb(xq, xk, freqs_cis)
|
||||
if FP8: xq, xk, xv = xq.cast(dtypes.bfloat16), xk.cast(dtypes.bfloat16), xv.cast(dtypes.bfloat16)
|
||||
xq, xk, xv = xq.transpose(1, 2), xk.transpose(1, 2), xv.transpose(1, 2)
|
||||
attn = xq.scaled_dot_product_attention(xk, xv, is_causal=True, enable_gqa=True).transpose(1, 2)
|
||||
attn = attn.reshape(bsz, seqlen, -1)
|
||||
return matmul(attn, wo, amax_x=amax_xo, amax_w=amax_wo)
|
||||
|
||||
def feed_forward(self, x:Tensor, ffn_norm:Tensor, w1:Tensor, w2:Tensor, w3:Tensor,
|
||||
amax_x1=None, amax_w1=None, amax_x2=None, amax_w2=None, amax_x3=None, amax_w3=None):
|
||||
x = rmsnorm(x, self.norm_eps) * ffn_norm
|
||||
x_w1 = matmul(x, w1, amax_x=amax_x1, amax_w=amax_w1).silu()
|
||||
x_w3 = matmul(x.contiguous_backward(), w3, amax_x=amax_x3, amax_w=amax_w3)
|
||||
return matmul(x_w1 * x_w3, w2, amax_x=amax_x2, amax_w=amax_w2)
|
||||
|
||||
@function(precompile=True, precompile_backward=True)
|
||||
def run_layer(self, x:Tensor, freqs_cis:Tensor,
|
||||
attention_norm:Tensor, wo:Tensor,
|
||||
ffn_norm:Tensor, w1:Tensor, w2:Tensor, w3:Tensor,
|
||||
wqkv:Tensor|None=None, wq:Tensor|None=None, wk:Tensor|None=None, wv:Tensor|None=None,
|
||||
amax_xqkv=None, amax_wqkv=None, amax_xq=None, amax_wq=None, amax_xk=None, amax_wk=None,
|
||||
amax_xv=None, amax_wv=None, amax_xo=None, amax_wo=None,
|
||||
amax_x1=None, amax_w1=None, amax_x2=None, amax_w2=None, amax_x3=None, amax_w3=None):
|
||||
h = x + self.attention(x, freqs_cis, attention_norm, wo, wqkv=wqkv, wq=wq, wk=wk, wv=wv,
|
||||
amax_xqkv=amax_xqkv, amax_wqkv=amax_wqkv, amax_xq=amax_xq, amax_wq=amax_wq,
|
||||
amax_xk=amax_xk, amax_wk=amax_wk, amax_xv=amax_xv, amax_wv=amax_wv,
|
||||
amax_xo=amax_xo, amax_wo=amax_wo)
|
||||
return h + self.feed_forward(h, ffn_norm, w1, w2, w3,
|
||||
amax_x1=amax_x1, amax_w1=amax_w1, amax_x2=amax_x2, amax_w2=amax_w2,
|
||||
amax_x3=amax_x3, amax_w3=amax_w3)
|
||||
|
||||
def shard(self, device:tuple[str, ...], mp:bool=False):
|
||||
from tinygrad.nn.state import get_parameters
|
||||
if not mp:
|
||||
for v in get_parameters(self): v.shard_(device, axis=None)
|
||||
else:
|
||||
# flat per-layer weights: axis 0 is n_layers, so shard axes are +1 vs per-layer Transformer
|
||||
if WQKV:
|
||||
self.wqkv.shard_(device, axis=1).realize() # (n_layers, out, dim) shard out
|
||||
else:
|
||||
self.wq.shard_(device, axis=1).realize() # (n_layers, n_heads*head_dim, dim) shard out
|
||||
self.wk.shard_(device, axis=1).realize() # (n_layers, n_kv_heads*head_dim, dim) shard out
|
||||
self.wv.shard_(device, axis=1).realize() # (n_layers, n_kv_heads*head_dim, dim) shard out
|
||||
self.wo.shard_(device, axis=2).realize() # (n_layers, dim, in) shard in
|
||||
self.w1.shard_(device, axis=1).realize() # (n_layers, hidden, dim) shard out
|
||||
self.w2.shard_(device, axis=2).realize() # (n_layers, dim, hidden) shard in
|
||||
self.w3.shard_(device, axis=1).realize() # (n_layers, hidden, dim) shard out
|
||||
self.attention_norm.shard_(device, axis=None).realize()
|
||||
self.ffn_norm.shard_(device, axis=None).realize()
|
||||
self.norm.weight.shard_(device, axis=None).realize()
|
||||
self.tok_embeddings.weight.shard_(device, axis=0).realize()
|
||||
self.output.shard_(device, axis=1).realize()
|
||||
self.freqs_cis.shard_(device, axis=None).realize()
|
||||
|
||||
def __call__(self, tokens:Tensor):
|
||||
h = self.tok_embeddings(tokens)
|
||||
freqs_cis = self.freqs_cis.cast(h.dtype)[:, :tokens.shape[1], :, :, :]
|
||||
a = self._fp8_amax if FP8 else None
|
||||
for i in range(self.n_layers):
|
||||
if WQKV:
|
||||
attn_kwargs = {"wqkv": self.wqkv[i]}
|
||||
amax_attn = {"amax_xqkv": a["xqkv"][i], "amax_wqkv": a["wqkv"][i]} if a else {}
|
||||
else:
|
||||
attn_kwargs = {"wq": self.wq[i], "wk": self.wk[i], "wv": self.wv[i]}
|
||||
amax_attn = {"amax_xq": a["xq"][i], "amax_wq": a["wq"][i],
|
||||
"amax_xk": a["xk"][i], "amax_wk": a["wk"][i],
|
||||
"amax_xv": a["xv"][i], "amax_wv": a["wv"][i]} if a else {}
|
||||
amax_layer = {"amax_xo": a["xo"][i], "amax_wo": a["wo"][i],
|
||||
"amax_x1": a["x1"][i], "amax_w1": a["w1"][i],
|
||||
"amax_x2": a["x2"][i], "amax_w2": a["w2"][i],
|
||||
"amax_x3": a["x3"][i], "amax_w3": a["w3"][i]} if a else {}
|
||||
h = self.run_layer(h, freqs_cis,
|
||||
self.attention_norm[i], self.wo[i],
|
||||
self.ffn_norm[i], self.w1[i], self.w2[i], self.w3[i],
|
||||
**attn_kwargs, **amax_attn, **amax_layer)
|
||||
logits = (self.norm(h).contiguous().contiguous_backward() @ self.output[0].T).contiguous_backward()
|
||||
return logits
|
||||
|
||||
def _get_pads(uop:UOp) -> list[UOp]:
|
||||
if uop.op == Ops.ADD: return _get_pads(uop.src[0]) + _get_pads(uop.src[1])
|
||||
return [uop]
|
||||
|
||||
def apply_grad(grad_buf:Tensor, new_grad:UOp):
|
||||
pads = _get_pads(new_grad)
|
||||
new_grad = new_grad.cast(grad_buf.dtype)
|
||||
if len(pads) <= 1:
|
||||
store = grad_buf.uop.store(grad_buf.uop + new_grad)
|
||||
grad_buf.uop = grad_buf.uop.after(store)
|
||||
return
|
||||
sorted_pads = sorted(pads, key=lambda p: p.marg[0][0] if p.op == Ops.PAD else 0)
|
||||
inners = [Tensor(p.src[0] if p.op == Ops.PAD else p, device=grad_buf.device).cast(grad_buf.dtype) for p in sorted_pads]
|
||||
grad_buf.assign(grad_buf + inners[0].cat(*inners[1:], dim=0))
|
||||
|
||||
if __name__ == "__main__":
|
||||
config = {}
|
||||
BS = config["BS"] = getenv("BS", 16)
|
||||
SEQLEN = config["SEQLEN"] = getenv("SEQLEN", 8192)
|
||||
|
||||
from examples.llama3 import MODEL_PARAMS
|
||||
model_params = MODEL_PARAMS[getenv("LLAMA3_SIZE", "8B")]["args"]
|
||||
if (llama_layers:=getenv("LLAMA_LAYERS")) != 0: model_params['n_layers'] = llama_layers
|
||||
model = FlatTransformer(**model_params, max_context=SEQLEN)
|
||||
state = nn.state.get_state_dict(model)
|
||||
print("tensor count:", len(state))
|
||||
|
||||
# shard the model
|
||||
from tinygrad import Device
|
||||
if (DP := getenv("DP", 1)) > 1:
|
||||
model.shard(tuple(f"{Device.DEFAULT}:{i}" for i in range(DP)))
|
||||
if (MP := getenv("MP", 1)) > 1:
|
||||
model.shard(tuple(f"{Device.DEFAULT}:{i}" for i in range(MP)), mp=True)
|
||||
|
||||
# preallocate all the grad buffers and zero them out
|
||||
grads = {x:Tensor.zeros(x.shape, dtype=x.dtype, device=x.device).contiguous()
|
||||
for x in state.values() if x.requires_grad is None}
|
||||
|
||||
# print model size
|
||||
sz = 0
|
||||
for k,v in state.items():
|
||||
print(f"{colored(k, 'green' if v in grads else 'white'):30s} {str(v.shape):30s} {str(v.dtype):20s} {v.device} {v.nbytes()/1e9:.2f} GB")
|
||||
sz += v.nbytes()
|
||||
print(f"total sz: {sz/1e9:.2f} GB")
|
||||
|
||||
with Timing("fake data: "): tokens = Tensor.randint(BS, SEQLEN+1, low=0, high=model.vocab_size, dtype=dtypes.int)
|
||||
with Timing("realize weights/grads/data: "): Tensor.realize(*state.values(), *grads.values(), tokens)
|
||||
print("mem per device: " + ', '.join(f"{dev}: {mem/1e9:.2f} GB" for dev, mem in sorted(GlobalCounters.mem_used_per_device.items())))
|
||||
if DP > 1: tokens = tokens.shard(tuple(f"{Device.DEFAULT}:{i}" for i in range(DP)), axis=0)
|
||||
if MP > 1: tokens = tokens.shard(tuple(f"{Device.DEFAULT}:{i}" for i in range(MP)))
|
||||
|
||||
@TinyJit
|
||||
def jit_step(tokens:Tensor):
|
||||
with Timing("python forward: "): loss = model(tokens[:, :-1]).sparse_categorical_crossentropy(tokens[:, 1:])
|
||||
with Timing("python backward: "):
|
||||
for t,g in zip(grads, loss.gradient(*grads)):
|
||||
apply_grad(grads[t], g.uop)
|
||||
with Timing("run step: "): loss.realize(*grads.values())
|
||||
|
||||
for i in range(6):
|
||||
GlobalCounters.reset()
|
||||
profile_marker(f"step {i}")
|
||||
with Timing(colored(f"*** step {i}: ", "red")):
|
||||
jit_step(tokens)
|
||||
print("mem per device: " + ', '.join(f"{dev}: {mem/1e9:.2f} GB" for dev, mem in sorted(GlobalCounters.mem_used_per_device.items())))
|
||||
@@ -1,140 +0,0 @@
|
||||
import os
|
||||
os.environ["WQKV"] = "1"
|
||||
import unittest
|
||||
import numpy as np
|
||||
from tinygrad import Tensor, nn, dtypes
|
||||
from tinygrad.nn.state import get_parameters
|
||||
from tinygrad.device import is_dtype_supported, Device
|
||||
from examples.mlperf.models.llama import Transformer
|
||||
from examples.mlperf.models.flat_llama import FlatTransformer
|
||||
|
||||
def copy_weights(flat:FlatTransformer, ref:Transformer):
|
||||
n_layers = flat.n_layers
|
||||
Tensor.realize(*nn.state.get_state_dict(ref).values())
|
||||
flat.wqkv.assign(Tensor(np.stack([ref.layers[i].attention.wqkv.weight.numpy() for i in range(n_layers)])))
|
||||
flat.wo.assign(Tensor(np.stack([ref.layers[i].attention.wo.weight.numpy() for i in range(n_layers)])))
|
||||
flat.w1.assign(Tensor(np.stack([ref.layers[i].feed_forward.w1.weight.numpy() for i in range(n_layers)])))
|
||||
flat.w2.assign(Tensor(np.stack([ref.layers[i].feed_forward.w2.weight.numpy() for i in range(n_layers)])))
|
||||
flat.w3.assign(Tensor(np.stack([ref.layers[i].feed_forward.w3.weight.numpy() for i in range(n_layers)])))
|
||||
flat.attention_norm.assign(Tensor(np.stack([ref.layers[i].attention_norm.weight.numpy() for i in range(n_layers)])))
|
||||
flat.ffn_norm.assign(Tensor(np.stack([ref.layers[i].ffn_norm.weight.numpy() for i in range(n_layers)])))
|
||||
flat.norm.weight.assign(Tensor(ref.norm.weight.numpy()))
|
||||
flat.tok_embeddings.weight.assign(Tensor(ref.tok_embeddings.weight.numpy()))
|
||||
flat.output.weight.assign(Tensor(ref.output.weight.numpy()))
|
||||
|
||||
class TestFlatLlama(unittest.TestCase):
|
||||
def test_forward_match(self):
|
||||
Tensor.manual_seed(42)
|
||||
params = dict(dim=128, hidden_dim=256, n_heads=4, n_kv_heads=2, n_layers=2, norm_eps=1e-5, vocab_size=1024, rope_theta=10000, max_context=64)
|
||||
ref = Transformer(**params)
|
||||
flat = FlatTransformer(**params)
|
||||
copy_weights(flat, ref)
|
||||
Tensor.realize(*nn.state.get_state_dict(flat).values())
|
||||
|
||||
tokens = Tensor([[1, 50, 100, 999, 2]])
|
||||
ref_logits = ref(tokens).realize()
|
||||
flat_logits = flat(tokens).realize()
|
||||
self.assertEqual(ref_logits.shape, flat_logits.shape)
|
||||
diff = (ref_logits - flat_logits).abs().max().item()
|
||||
self.assertLess(diff, 1e-5, f"forward mismatch: max abs diff {diff}")
|
||||
|
||||
def test_backward_match(self):
|
||||
Tensor.manual_seed(42)
|
||||
params = dict(dim=128, hidden_dim=256, n_heads=4, n_kv_heads=2, n_layers=2, norm_eps=1e-5, vocab_size=1024, rope_theta=10000, max_context=64)
|
||||
ref = Transformer(**params)
|
||||
flat = FlatTransformer(**params)
|
||||
copy_weights(flat, ref)
|
||||
|
||||
for p in get_parameters(ref): p.requires_grad_(True)
|
||||
for p in get_parameters(flat): p.requires_grad_(True)
|
||||
Tensor.realize(*nn.state.get_state_dict(flat).values())
|
||||
|
||||
tokens = Tensor([[1, 50, 100, 999, 2, 10]])
|
||||
|
||||
ref_loss = ref(tokens[:, :-1]).sparse_categorical_crossentropy(tokens[:, 1:])
|
||||
ref_loss.backward()
|
||||
ref_grads = {k: v.grad.numpy() for k, v in nn.state.get_state_dict(ref).items() if v.grad is not None}
|
||||
|
||||
flat_loss = flat(tokens[:, :-1]).sparse_categorical_crossentropy(tokens[:, 1:])
|
||||
flat_loss.backward()
|
||||
flat_grads = {k: v.grad.numpy() for k, v in nn.state.get_state_dict(flat).items() if v.grad is not None}
|
||||
|
||||
# check loss matches
|
||||
self.assertAlmostEqual(ref_loss.item(), flat_loss.item(), places=4)
|
||||
|
||||
# check output weight grad matches
|
||||
diff = abs(ref_grads["output.weight"] - flat_grads["output.weight"]).max()
|
||||
self.assertLess(diff, 1e-4, f"output.weight grad mismatch: max abs diff {diff}")
|
||||
|
||||
# check per-layer weight grads match
|
||||
for i in range(params["n_layers"]):
|
||||
for flat_key, ref_key in [
|
||||
("wqkv", f"layers.{i}.attention.wqkv.weight"),
|
||||
("wo", f"layers.{i}.attention.wo.weight"),
|
||||
("w1", f"layers.{i}.feed_forward.w1.weight"),
|
||||
("w2", f"layers.{i}.feed_forward.w2.weight"),
|
||||
("w3", f"layers.{i}.feed_forward.w3.weight"),
|
||||
]:
|
||||
diff = abs(ref_grads[ref_key] - flat_grads[flat_key][i]).max()
|
||||
self.assertLess(diff, 1e-4, f"layer {i} {flat_key} grad mismatch: max abs diff {diff}")
|
||||
|
||||
@unittest.skipUnless(Device.DEFAULT == "CPU", "multi-device CPU test")
|
||||
def test_forward_match_mp(self):
|
||||
Tensor.manual_seed(42)
|
||||
params = dict(dim=128, hidden_dim=256, n_heads=4, n_kv_heads=2, n_layers=2, norm_eps=1e-5, vocab_size=1024, rope_theta=10000, max_context=64)
|
||||
from tinygrad import Device
|
||||
devices = (f"{Device.DEFAULT}:0", f"{Device.DEFAULT}:1")
|
||||
ref = Transformer(**params)
|
||||
flat = FlatTransformer(**params)
|
||||
copy_weights(flat, ref)
|
||||
Tensor.realize(*nn.state.get_state_dict(flat).values())
|
||||
flat.shard(devices, mp=True)
|
||||
|
||||
tokens = Tensor([[1, 50, 100, 999, 2]], device=devices[0])
|
||||
ref_logits = ref(tokens.to(devices[0])).numpy()
|
||||
flat_logits = flat(tokens.shard(devices)).numpy()
|
||||
self.assertEqual(ref_logits.shape, flat_logits.shape)
|
||||
np.testing.assert_allclose(flat_logits, ref_logits, atol=1e-4, rtol=1e-4)
|
||||
|
||||
@unittest.skipUnless(Device.DEFAULT == "CPU", "multi-device CPU test")
|
||||
def test_forward_match_dp(self):
|
||||
Tensor.manual_seed(42)
|
||||
params = dict(dim=128, hidden_dim=256, n_heads=4, n_kv_heads=2, n_layers=2, norm_eps=1e-5, vocab_size=1024, rope_theta=10000, max_context=64)
|
||||
from tinygrad import Device
|
||||
devices = (f"{Device.DEFAULT}:0", f"{Device.DEFAULT}:1")
|
||||
ref = Transformer(**params)
|
||||
flat = FlatTransformer(**params)
|
||||
copy_weights(flat, ref)
|
||||
Tensor.realize(*nn.state.get_state_dict(flat).values())
|
||||
flat.shard(devices)
|
||||
|
||||
tokens = Tensor([[1, 50, 100, 999, 2], [2, 100, 50, 1, 999]], device=devices[0])
|
||||
ref_logits = ref(tokens.to(devices[0])).numpy()
|
||||
flat_logits = flat(tokens.shard(devices, axis=0)).numpy()
|
||||
self.assertEqual(ref_logits.shape, flat_logits.shape)
|
||||
np.testing.assert_allclose(flat_logits, ref_logits, atol=1e-4, rtol=1e-4)
|
||||
|
||||
@unittest.skipUnless(is_dtype_supported(dtypes.fp8e4m3), "fp8 not supported on this device")
|
||||
def test_forward_fp8(self):
|
||||
import examples.mlperf.models.flat_llama as flat_llama_mod
|
||||
old_fp8 = flat_llama_mod.FP8
|
||||
try:
|
||||
flat_llama_mod.FP8 = 1
|
||||
Tensor.manual_seed(42)
|
||||
params = dict(dim=128, hidden_dim=256, n_heads=4, n_kv_heads=2, n_layers=2, norm_eps=1e-5, vocab_size=1024, rope_theta=10000, max_context=64)
|
||||
ref = Transformer(**params)
|
||||
flat = FlatTransformer(**params)
|
||||
copy_weights(flat, ref)
|
||||
Tensor.realize(*nn.state.get_state_dict(flat).values())
|
||||
|
||||
tokens = Tensor([[1, 50, 100, 999, 2]])
|
||||
ref_logits = ref(tokens).numpy()
|
||||
flat_logits = flat(tokens).numpy()
|
||||
self.assertEqual(ref_logits.shape, flat_logits.shape)
|
||||
# FP8 has lower precision, allow larger tolerance
|
||||
np.testing.assert_allclose(flat_logits, ref_logits, atol=1.0, rtol=0.1)
|
||||
finally:
|
||||
flat_llama_mod.FP8 = old_fp8
|
||||
|
||||
if __name__ == "__main__":
|
||||
unittest.main()
|
||||
+13
-36
@@ -1,21 +1,7 @@
|
||||
from tinygrad.tensor import Tensor
|
||||
from tinygrad.dtype import dtypes
|
||||
from tinygrad.nn.optim import Optimizer
|
||||
from tinygrad.helpers import FUSE_OPTIM, getenv
|
||||
from tinygrad.uop.ops import UOp, Ops
|
||||
|
||||
STOCHASTIC_ROUND = getenv("STOCHASTIC_ROUND", 0)
|
||||
MASTER_WEIGHTS = getenv("MASTER_WEIGHTS", 0)
|
||||
|
||||
def stochastic_round_bf16(x:Tensor) -> Tensor:
|
||||
bits = x.bitcast(dtypes.uint32)
|
||||
if isinstance(x.device, tuple):
|
||||
shape = x.uop.shard_shape if x.uop.axis is not None else x.shape
|
||||
noise = Tensor(UOp(Ops.MSTACK, dtypes.default_float, tuple(Tensor.rand(*shape, device=d).uop for d in x.device)))
|
||||
else:
|
||||
noise = x.rand_like()
|
||||
noise = (noise * 0xFFFF).cast(dtypes.uint32)
|
||||
return ((bits + noise) & 0xFFFF0000).bitcast(dtypes.float32).cast(dtypes.bfloat16)
|
||||
from tinygrad.helpers import FUSE_OPTIM
|
||||
|
||||
class GradAccClipAdamW(Optimizer):
|
||||
def __init__(self, params:list[Tensor], lr=0.001, b1=0.9, b2=0.999, eps=1e-6, weight_decay=0.0, grad_acc=1, clip_norm=1.0, device=None, fused=FUSE_OPTIM):
|
||||
@@ -25,7 +11,6 @@ class GradAccClipAdamW(Optimizer):
|
||||
self.m = self._new_optim_param()
|
||||
self.v = self._new_optim_param()
|
||||
self.grad_acc, self.clip_norm = grad_acc, clip_norm
|
||||
self.master_params:list[Tensor]|None = [p.float().contiguous() for p in self.params] if MASTER_WEIGHTS and self.params[0].dtype != dtypes.float32 else None
|
||||
|
||||
def fstep(self, grads:list[Tensor]):
|
||||
if self.fused:
|
||||
@@ -33,17 +18,15 @@ class GradAccClipAdamW(Optimizer):
|
||||
updates = [out[0][self.pos_params[i]:self.pos_params[i+1]].reshape(tt.shape) for i, tt in enumerate(self.params)]
|
||||
else:
|
||||
updates, extra = self._step([], grads)
|
||||
for i, tt in enumerate(self.params): tt.assign(self._apply_update(tt, updates[i], self.master_params[i] if self.master_params else None))
|
||||
to_realize = extra+self.params+self.buffers+(self.master_params or [])
|
||||
for i, tt in enumerate(self.params): tt.assign(self._apply_update(tt, updates[i]))
|
||||
to_realize = extra+self.params+self.buffers
|
||||
|
||||
Tensor.realize(*to_realize)
|
||||
return extra[-1]
|
||||
|
||||
def _step(self, params:list[Tensor], grads:list[Tensor]) -> tuple[list[Tensor], list[Tensor]]:
|
||||
grads = list(grads)
|
||||
|
||||
for i in range(len(grads)):
|
||||
if grads[i].device != self.m[i].device: grads[i] = grads[i].to(self.m[i].device)
|
||||
if grads[i].device != self.m[i].device: grads[i].assign(grads[i].to(self.m[i].device))
|
||||
|
||||
if self.fused:
|
||||
grads[0].assign(grads[0] / self.grad_acc)
|
||||
@@ -60,21 +43,15 @@ class GradAccClipAdamW(Optimizer):
|
||||
self.b1_t *= self.b1
|
||||
self.b2_t *= self.b2
|
||||
for i, g in enumerate(grads):
|
||||
m_new = self.b1 * self.m[i].float() + (1.0 - self.b1) * g.float()
|
||||
v_new = self.b2 * self.v[i].float() + (1.0 - self.b2) * (g.float() * g.float())
|
||||
self.m[i].assign(m_new.cast(self.m[i].dtype))
|
||||
self.v[i].assign(v_new.cast(self.v[i].dtype))
|
||||
m_hat = m_new / (1.0 - self.b1_t)
|
||||
v_hat = v_new / (1.0 - self.b2_t)
|
||||
self.m[i].assign((self.b1 * self.m[i] + (1.0 - self.b1) * g).cast(self.m[i].dtype))
|
||||
self.v[i].assign((self.b2 * self.v[i] + (1.0 - self.b2) * (g * g)).cast(self.v[i].dtype))
|
||||
m_hat = (self.m[i] / (1.0 - self.b1_t)).cast(self.m[i].dtype)
|
||||
v_hat = (self.v[i] / (1.0 - self.b2_t)).cast(self.v[i].dtype)
|
||||
up = m_hat / (v_hat.sqrt() + self.eps)
|
||||
ret.append(self.lr * up)
|
||||
ret.append((self.lr * up).cast(g.dtype))
|
||||
return ret, [self.b1_t, self.b2_t] + self.m + self.v + [total_norm]
|
||||
|
||||
def _apply_update(self, t:Tensor, up:Tensor, master:Tensor|None=None) -> Tensor:
|
||||
w = master if master is not None else t
|
||||
wd = self.wd if t.ndim >= 3 else 0.0
|
||||
up = up.float().shard_like(w) + self.lr.to(w.device) * wd * w.detach()
|
||||
new_w = w.detach() - up
|
||||
if master is not None: master.assign(new_w)
|
||||
if STOCHASTIC_ROUND and t.dtype == dtypes.bfloat16: return stochastic_round_bf16(new_w)
|
||||
return new_w.cast(t.dtype)
|
||||
def _apply_update(self, t:Tensor, up:Tensor) -> Tensor:
|
||||
wd = self.wd if t.ndim >= 2 else 0.0
|
||||
up = up.shard_like(t) + self.lr.to(t.device) * wd * t.detach()
|
||||
return t.detach() - up.cast(t.dtype)
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="bert"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=1 BS=128 EVAL_BS=128
|
||||
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="bert"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=8 BS=1024 EVAL_BS=1024
|
||||
export OPT_BASE_LEARNING_RATE=0.0011 OPT_LAMB_BETA_1=0.60466 OPT_LAMB_BETA_2=0.85437 DECAY=0.1
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="bert"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=8 BS=1024 EVAL_BS=1024
|
||||
|
||||
|
||||
+1
-1
@@ -1,7 +1,7 @@
|
||||
#!/bin/bash
|
||||
set -e # Exit on any error
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="bert"
|
||||
export SUBMISSION_PLATFORM="tinybox_8xMI300X"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=8 BS=1024 EVAL_BS=1024
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="bert"
|
||||
export DEFAULT_FLOAT="HALF" SUM_DTYPE="HALF" GPUS=6 BS=96 EVAL_BS=96
|
||||
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="bert"
|
||||
export DEFAULT_FLOAT="HALF" SUM_DTYPE="HALF" GPUS=6 BS=96 EVAL_BS=96
|
||||
|
||||
|
||||
+1
-1
@@ -1,7 +1,7 @@
|
||||
#!/bin/bash
|
||||
set -e # Exit on any error
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="bert"
|
||||
export SUBMISSION_PLATFORM="tinybox_green"
|
||||
export DEFAULT_FLOAT="HALF" SUM_DTYPE="HALF" GPUS=6 BS=96 EVAL_BS=96
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="bert"
|
||||
export DEFAULT_FLOAT="HALF" SUM_DTYPE="HALF" GPUS=6 BS=96 EVAL_BS=96
|
||||
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="bert"
|
||||
export DEFAULT_FLOAT="HALF" SUM_DTYPE="HALF" GPUS=6 BS=96 EVAL_BS=96
|
||||
|
||||
|
||||
+1
-1
@@ -1,7 +1,7 @@
|
||||
#!/bin/bash
|
||||
set -e # Exit on any error
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="bert"
|
||||
export SUBMISSION_PLATFORM="tinybox_red"
|
||||
export DEFAULT_FLOAT="HALF" SUM_DTYPE="HALF" GPUS=6 BS=96 EVAL_BS=96
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="resnet"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=1536 EVAL_BS=192
|
||||
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="resnet"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=1536 EVAL_BS=192
|
||||
|
||||
|
||||
+1
-1
@@ -1,7 +1,7 @@
|
||||
#!/bin/bash
|
||||
set -e # Exit on any error
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="resnet"
|
||||
export SUBMISSION_PLATFORM="tinybox_green"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=1536 EVAL_BS=192
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="resnet"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=1536 EVAL_BS=192
|
||||
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="resnet"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=1536 EVAL_BS=192
|
||||
|
||||
|
||||
+1
-1
@@ -1,7 +1,7 @@
|
||||
#!/bin/bash
|
||||
set -e # Exit on any error
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="resnet"
|
||||
export SUBMISSION_PLATFORM="tinybox_red"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=1536 EVAL_BS=192
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="retinanet"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=96 EVAL_BS=96
|
||||
export BASEDIR="/raid/datasets/openimages"
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="retinanet"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=96 EVAL_BS=96
|
||||
export BASEDIR="/raid/datasets/openimages"
|
||||
|
||||
+1
-1
@@ -1,7 +1,7 @@
|
||||
#!/bin/bash
|
||||
set -e # Exit on any error
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="retinanet"
|
||||
export SUBMISSION_PLATFORM="tinybox_green"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=96 EVAL_BS=96
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="retinanet"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=96 EVAL_BS=96
|
||||
export BASEDIR="/raid/datasets/openimages"
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="retinanet"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=96 EVAL_BS=96
|
||||
export BASEDIR="/raid/datasets/openimages"
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="bert"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=1 BS=128 EVAL_BS=128
|
||||
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="bert"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=8 BS=1024 EVAL_BS=1024
|
||||
export OPT_BASE_LEARNING_RATE=0.0011 OPT_LAMB_BETA_1=0.60466 OPT_LAMB_BETA_2=0.85437 DECAY=0.1
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="bert"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=8 BS=1024 EVAL_BS=1024
|
||||
|
||||
|
||||
+1
-1
@@ -2,7 +2,7 @@
|
||||
set -e # Exit on any error
|
||||
set -o pipefail # Make pipeline fail if any command fails
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="bert"
|
||||
export SUBMISSION_PLATFORM="tinybox_8xMI300X"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=8 BS=1024 EVAL_BS=1024
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="bert"
|
||||
export DEFAULT_FLOAT="HALF" SUM_DTYPE="HALF" GPUS=6 BS=90 EVAL_BS=90
|
||||
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="bert"
|
||||
export DEFAULT_FLOAT="HALF" SUM_DTYPE="HALF" GPUS=6 BS=90 EVAL_BS=90
|
||||
|
||||
|
||||
+1
-1
@@ -2,7 +2,7 @@
|
||||
set -e # Exit on any error
|
||||
set -o pipefail # Make pipeline fail if any command fails
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="bert"
|
||||
export SUBMISSION_PLATFORM="tinybox_green"
|
||||
export DEFAULT_FLOAT="HALF" SUM_DTYPE="HALF" GPUS=6 BS=90 EVAL_BS=90
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="bert"
|
||||
export DEFAULT_FLOAT="HALF" SUM_DTYPE="HALF" GPUS=6 BS=90 EVAL_BS=90
|
||||
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="bert"
|
||||
export DEFAULT_FLOAT="HALF" SUM_DTYPE="HALF" GPUS=6 BS=90 EVAL_BS=90
|
||||
|
||||
|
||||
+1
-1
@@ -2,7 +2,7 @@
|
||||
set -e # Exit on any error
|
||||
set -o pipefail # Make pipeline fail if any command fails
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="bert"
|
||||
export SUBMISSION_PLATFORM="tinybox_red"
|
||||
export DEFAULT_FLOAT="HALF" SUM_DTYPE="HALF" GPUS=6 BS=90 EVAL_BS=90
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="resnet"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=1536 EVAL_BS=192
|
||||
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="resnet"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=1536 EVAL_BS=192
|
||||
|
||||
|
||||
+1
-1
@@ -2,7 +2,7 @@
|
||||
set -e # Exit on any error
|
||||
set -o pipefail # Make pipeline fail if any command fails
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="resnet"
|
||||
export SUBMISSION_PLATFORM="tinybox_green"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=1536 EVAL_BS=192
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="resnet"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=1536 EVAL_BS=192
|
||||
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="resnet"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=1536 EVAL_BS=192
|
||||
|
||||
|
||||
+1
-1
@@ -2,7 +2,7 @@
|
||||
set -e # Exit on any error
|
||||
set -o pipefail # Make pipeline fail if any command fails
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="resnet"
|
||||
export SUBMISSION_PLATFORM="tinybox_red"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=1536 EVAL_BS=192
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="retinanet"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=96 EVAL_BS=96
|
||||
export BASEDIR="/raid/datasets/openimages"
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="retinanet"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=96 EVAL_BS=96
|
||||
export BASEDIR="/raid/datasets/openimages"
|
||||
|
||||
+1
-1
@@ -2,7 +2,7 @@
|
||||
set -e # Exit on any error
|
||||
set -o pipefail # Make pipeline fail if any command fails
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="retinanet"
|
||||
export SUBMISSION_PLATFORM="tinybox_green"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=96 EVAL_BS=96
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="retinanet"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=96 EVAL_BS=96
|
||||
export BASEDIR="/raid/datasets/openimages"
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="retinanet"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=96 EVAL_BS=96
|
||||
export BASEDIR="/raid/datasets/openimages"
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="bert"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=1 BS=128 EVAL_BS=128
|
||||
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="bert"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=8 BS=1024 EVAL_BS=1024
|
||||
export OPT_BASE_LEARNING_RATE=0.0011 OPT_LAMB_BETA_1=0.60466 OPT_LAMB_BETA_2=0.85437 DECAY=0.1
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="bert"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=8 BS=1024 EVAL_BS=1024
|
||||
|
||||
|
||||
+1
-1
@@ -2,7 +2,7 @@
|
||||
set -e # Exit on any error
|
||||
set -o pipefail # Make pipeline fail if any command fails
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="bert"
|
||||
export SUBMISSION_PLATFORM="tinybox_8xMI300X"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=8 BS=1024 EVAL_BS=1024
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD DEBUG=0 JIT=1 FLASH_ATTENTION=1
|
||||
export PYTHONPATH="." AMD=1 DEBUG=0 JIT=1 FLASH_ATTENTION=1
|
||||
export MODEL="bert"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=8 BS=1024 EVAL_BS=1024
|
||||
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="bert"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=8 BS=1024 EVAL_BS=1024
|
||||
|
||||
|
||||
+1
-1
@@ -2,7 +2,7 @@
|
||||
set -e # Exit on any error
|
||||
set -o pipefail # Make pipeline fail if any command fails
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="bert"
|
||||
export SUBMISSION_PLATFORM="tinybox_8xMI350X"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=8 BS=1024 EVAL_BS=1024
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="bert"
|
||||
export DEFAULT_FLOAT="HALF" SUM_DTYPE="HALF" GPUS=6 BS=72 EVAL_BS=72
|
||||
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="bert"
|
||||
export DEFAULT_FLOAT="HALF" SUM_DTYPE="HALF" GPUS=6 BS=72 EVAL_BS=72
|
||||
|
||||
|
||||
+1
-1
@@ -2,7 +2,7 @@
|
||||
set -e # Exit on any error
|
||||
set -o pipefail # Make pipeline fail if any command fails
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="bert"
|
||||
export SUBMISSION_PLATFORM="tinybox_green"
|
||||
export DEFAULT_FLOAT="HALF" SUM_DTYPE="HALF" GPUS=6 BS=72 EVAL_BS=72
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="bert"
|
||||
export DEFAULT_FLOAT="HALF" SUM_DTYPE="HALF" GPUS=6 BS=96 EVAL_BS=96
|
||||
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="bert"
|
||||
export DEFAULT_FLOAT="HALF" SUM_DTYPE="HALF" GPUS=6 BS=96 EVAL_BS=96
|
||||
|
||||
|
||||
+1
-1
@@ -2,7 +2,7 @@
|
||||
set -e # Exit on any error
|
||||
set -o pipefail # Make pipeline fail if any command fails
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="bert"
|
||||
export SUBMISSION_PLATFORM="tinybox_red"
|
||||
export DEFAULT_FLOAT="HALF" SUM_DTYPE="HALF" GPUS=6 BS=96 EVAL_BS=96
|
||||
|
||||
-1
@@ -5,7 +5,6 @@ export DEV=${DEV:-AMD}
|
||||
export EMULATE="AMD_CDNA4"
|
||||
export CHECK_OOB=0
|
||||
export REWRITE_STACK_LIMIT=5000000 HCQDEV_WAIT_TIMEOUT_MS=240000
|
||||
export DEVICE_IN_FUNCTION_BUG=1
|
||||
|
||||
export DEBUG=${DEBUG:-2}
|
||||
export HK_FLASH_ATTENTION=${HK_FLASH_ATTENTION:-1}
|
||||
|
||||
+3
-5
@@ -12,12 +12,10 @@ export HK_FLASH_ATTENTION=${HK_FLASH_ATTENTION:-1}
|
||||
export ALL2ALL=${ALL2ALL:-1}
|
||||
export USE_ATOMICS=${USE_ATOMICS:-1}
|
||||
export ASM_GEMM=${ASM_GEMM:-1}
|
||||
export WQKV=${WQKV:-1}
|
||||
export MASTER_WEIGHTS=${MASTER_WEIGHTS:-1}
|
||||
export FP8=${FP8:-1}
|
||||
export WQKV=${WQKV:-0}
|
||||
|
||||
export DEFAULT_FLOAT="bfloat16" OPTIM_DTYPE="bfloat16"
|
||||
export DP=${DP:-8} MP=${MP:-1} BS=${BS:-8} EVAL_BS=${EVAL_BS:-8} GRADIENT_ACC_STEPS=${GRADIENT_ACC_STEPS:-4}
|
||||
export DP=${DP:-8} MP=${MP:-1} BS=${BS:-8} EVAL_BS=${EVAL_BS:-8} GRADIENT_ACC_STEPS=${GRADIENT_ACC_STEPS:-2}
|
||||
export GBS=$((BS * GRADIENT_ACC_STEPS))
|
||||
|
||||
export MODEL="llama3"
|
||||
@@ -36,7 +34,7 @@ export DATA_SEED=${DATA_SEED:-5760}
|
||||
export JITBEAM=${JITBEAM:-3}
|
||||
export BEAM_UOPS_MAX=6000 BEAM_UPCAST_MAX=256 BEAM_LOCAL_MAX=1024 BEAM_MIN_PROGRESS=5 BEAM_PADTO=1
|
||||
|
||||
export FAKEDATA=1 BENCHMARK=${BENCHMARK:-10}
|
||||
export FAKEDATA=1 BENCHMARK=10
|
||||
if [ -z "$FULL_LAYERS" ]; then
|
||||
export LLAMA_LAYERS=2
|
||||
fi
|
||||
|
||||
+1
-3
@@ -12,9 +12,7 @@ export HK_FLASH_ATTENTION=${HK_FLASH_ATTENTION:-1}
|
||||
export ALL2ALL=${ALL2ALL:-1}
|
||||
export USE_ATOMICS=${USE_ATOMICS:-1}
|
||||
export ASM_GEMM=${ASM_GEMM:-1}
|
||||
export WQKV=${WQKV:-1}
|
||||
export MASTER_WEIGHTS=${MASTER_WEIGHTS:-1}
|
||||
export FP8=${FP8:-1}
|
||||
export WQKV=${WQKV:-0}
|
||||
|
||||
export DEFAULT_FLOAT="bfloat16" OPTIM_DTYPE="bfloat16"
|
||||
export DP=${DP:-8} MP=${MP:-1} BS=${BS:-8} EVAL_BS=${EVAL_BS:-8} GRADIENT_ACC_STEPS=${GRADIENT_ACC_STEPS:-4}
|
||||
|
||||
+3
-2
@@ -1,5 +1,6 @@
|
||||
#!/bin/bash
|
||||
export BENCHMARK=5
|
||||
export EVAL_BS=0
|
||||
VIZ=${VIZ:--1} FULL_LAYERS=1 DEBUG=0 examples/mlperf/training_submission_v6.0/tinycorp/benchmarks/llama8b/implementations/tinybox_8xMI350X/dev_beam.sh
|
||||
extra/viz/cli.py --profile -s "${DEV:-AMD}"
|
||||
export VIZ=${VIZ:--1}
|
||||
examples/mlperf/training_submission_v6.0/tinycorp/benchmarks/llama8b/implementations/tinybox_8xMI350X/dev_run.sh
|
||||
extra/viz/cli.py --profile --device "AMD" --top 20
|
||||
|
||||
-49
@@ -1,49 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
set -e # Exit on any error
|
||||
set -o pipefail # Make pipeline fail if any command fails
|
||||
|
||||
export PYTHONPATH="."
|
||||
export DEV=AMD
|
||||
export EMULATE="AMD_CDNA4"
|
||||
export CHECK_OOB=0
|
||||
export REWRITE_STACK_LIMIT=5000000 HCQDEV_WAIT_TIMEOUT_MS=240000
|
||||
export DEVICE_IN_FUNCTION_BUG=1
|
||||
|
||||
export HK_FLASH_ATTENTION=1
|
||||
export ALL2ALL=1
|
||||
export USE_ATOMICS=1
|
||||
export ASM_GEMM=1
|
||||
export WQKV=1
|
||||
export MASTER_WEIGHTS=1
|
||||
export FP8=1
|
||||
|
||||
export DEFAULT_FLOAT="bfloat16" OPTIM_DTYPE="bfloat16"
|
||||
export DP=8 MP=1 BS=8 EVAL_BS=8 GRADIENT_ACC_STEPS=4
|
||||
export GBS=$((BS * GRADIENT_ACC_STEPS))
|
||||
|
||||
export MODEL="llama3"
|
||||
export BASEDIR="/raid/datasets/c4-8b/"
|
||||
export SMALL=1
|
||||
export LLAMA3_SIZE=8B
|
||||
export EVAL_TARGET=3.3 EVAL_FREQ=12288
|
||||
export LR="1e-3" END_LR="1e-4" WARMUP_SAMPLES=4096 MAX_STEPS=1200000
|
||||
export WARMUP_STEPS=$((WARMUP_SAMPLES / GBS))
|
||||
export SAMPLES=$((MAX_STEPS * GBS))
|
||||
export SEQLEN=8192
|
||||
|
||||
export SEED=$RANDOM
|
||||
export DATA_SEED=$SEED
|
||||
|
||||
export JITBEAM=3
|
||||
export BEAM_UOPS_MAX=6000 BEAM_UPCAST_MAX=256 BEAM_LOCAL_MAX=1024 BEAM_MIN_PROGRESS=5 BEAM_PADTO=1
|
||||
|
||||
export LOGMLPERF=1
|
||||
|
||||
DATETIME=$(date "+%m%d%H%M")
|
||||
LOGFILE="llama31_8b_8xMI350x_${DATETIME}_${SEED}.log"
|
||||
|
||||
# beam
|
||||
FAKEDATA=1 BENCHMARK=10 INITMLPERF=1 LLAMA_LAYERS=2 python3 examples/mlperf/model_train.py | tee "$LOGFILE"
|
||||
|
||||
# run
|
||||
RUNMLPERF=1 python3 examples/mlperf/model_train.py | tee -a "$LOGFILE"
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="resnet"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=1536 EVAL_BS=192
|
||||
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="resnet"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=1536 EVAL_BS=192
|
||||
|
||||
|
||||
+1
-1
@@ -2,7 +2,7 @@
|
||||
set -e # Exit on any error
|
||||
set -o pipefail # Make pipeline fail if any command fails
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="resnet"
|
||||
export SUBMISSION_PLATFORM="tinybox_green"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=1536 EVAL_BS=192
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="resnet"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=1536 EVAL_BS=192
|
||||
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="resnet"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=1536 EVAL_BS=192
|
||||
|
||||
|
||||
+1
-1
@@ -2,7 +2,7 @@
|
||||
set -e # Exit on any error
|
||||
set -o pipefail # Make pipeline fail if any command fails
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="resnet"
|
||||
export SUBMISSION_PLATFORM="tinybox_red"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=1536 EVAL_BS=192
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="retinanet"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=96 EVAL_BS=96
|
||||
export BASEDIR="/raid/datasets/openimages"
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="retinanet"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=96 EVAL_BS=96
|
||||
export BASEDIR="/raid/datasets/openimages"
|
||||
|
||||
+1
-1
@@ -2,7 +2,7 @@
|
||||
set -e # Exit on any error
|
||||
set -o pipefail # Make pipeline fail if any command fails
|
||||
|
||||
export PYTHONPATH="." DEV=NV
|
||||
export PYTHONPATH="." NV=1
|
||||
export MODEL="retinanet"
|
||||
export SUBMISSION_PLATFORM="tinybox_green"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=96 EVAL_BS=96
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="retinanet"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=96 EVAL_BS=96
|
||||
export BASEDIR="/raid/datasets/openimages"
|
||||
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
export PYTHONPATH="." DEV=AMD
|
||||
export PYTHONPATH="." AMD=1
|
||||
export MODEL="retinanet"
|
||||
export DEFAULT_FLOAT="HALF" GPUS=6 BS=96 EVAL_BS=96
|
||||
export BASEDIR="/raid/datasets/openimages"
|
||||
|
||||
@@ -1,38 +0,0 @@
|
||||
{
|
||||
"submitter": "tinycorp",
|
||||
"division": "closed",
|
||||
"status": "Available on-premise",
|
||||
"system_name": "tinybox 8xMI350X",
|
||||
"number_of_nodes": "1",
|
||||
"host_processors_per_node": "2",
|
||||
"host_processor_model_name": "AMD EPYC 9575F",
|
||||
"host_processor_core_count": "32",
|
||||
"host_processor_vcpu_count": "64",
|
||||
"host_processor_frequency": "",
|
||||
"host_processor_caches": "",
|
||||
"host_processor_interconnect": "",
|
||||
"host_memory_capacity": "3072 GiB",
|
||||
"host_storage_type": "NVMe SSD",
|
||||
"host_storage_capacity": "4TB",
|
||||
"host_networking": "",
|
||||
"host_networking_topology": "",
|
||||
"host_memory_configuration": "24x 128GB DDR5",
|
||||
"accelerators_per_node": "8",
|
||||
"accelerator_model_name": "AMD Instinct MI350X 288GB HBM3e",
|
||||
"accelerator_host_interconnect": "PCIe 5.0 x16",
|
||||
"accelerator_frequency": "",
|
||||
"accelerator_on-chip_memories": "",
|
||||
"accelerator_memory_configuration": "HBM3",
|
||||
"accelerator_memory_capacity": "288GB",
|
||||
"accelerator_interconnect": "",
|
||||
"accelerator_interconnect_topology": "",
|
||||
"cooling": "air",
|
||||
"hw_notes": "",
|
||||
"framework": "tinygrad, branch mlperf_training_v6.0",
|
||||
"other_software_stack": {
|
||||
"python": "3.12.3",
|
||||
"ROCm": "7.1.1"
|
||||
},
|
||||
"operating_system": "Ubuntu 24.04.3 LTS",
|
||||
"sw_notes": ""
|
||||
}
|
||||
+28
-27
@@ -1,11 +1,12 @@
|
||||
from pathlib import Path
|
||||
import numpy as np
|
||||
import torch
|
||||
from torchvision.utils import make_grid, save_image
|
||||
from tinygrad.nn.state import get_parameters
|
||||
from tinygrad.tensor import Tensor
|
||||
from tinygrad.helpers import trange
|
||||
from tinygrad.nn import optim
|
||||
from tinygrad.nn.datasets import mnist
|
||||
from extra.datasets import fetch_mnist
|
||||
|
||||
class LinearGen:
|
||||
def __init__(self):
|
||||
@@ -37,14 +38,14 @@ class LinearDisc:
|
||||
return x
|
||||
|
||||
def make_batch(images):
|
||||
sample = Tensor.randint(batch_size, low=0, high=images.shape[0])
|
||||
return images[sample].reshape(batch_size, 28*28).cast('float').div(127.5).sub(1.0)
|
||||
sample = np.random.randint(0, len(images), size=(batch_size))
|
||||
image_b = images[sample].reshape(-1, 28*28).astype(np.float32) / 127.5 - 1.0
|
||||
return Tensor(image_b)
|
||||
|
||||
def make_labels(bs, col, val=-2.0):
|
||||
y = Tensor.zeros(bs, 2)
|
||||
if col == 0: y = y + Tensor([val, 0.0])
|
||||
else: y = y + Tensor([0.0, val])
|
||||
return y
|
||||
y = np.zeros((bs, 2), np.float32)
|
||||
y[range(bs), [col] * bs] = val # Can we do label smoothing? i.e -2.0 changed to -1.98789.
|
||||
return Tensor(y)
|
||||
|
||||
def train_discriminator(optimizer, data_real, data_fake):
|
||||
real_labels = make_labels(batch_size, 1)
|
||||
@@ -70,12 +71,12 @@ def train_generator(optimizer, data_fake):
|
||||
|
||||
if __name__ == "__main__":
|
||||
# data for training and validation
|
||||
X_train, _, _, _ = mnist()
|
||||
images_real = np.vstack(fetch_mnist()[::2])
|
||||
ds_noise = Tensor.randn(64, 128, requires_grad=False)
|
||||
# parameters
|
||||
epochs, batch_size, k = 300, 512, 1
|
||||
sample_interval = epochs // 10
|
||||
n_steps = X_train.shape[0] // batch_size
|
||||
n_steps = len(images_real) // batch_size
|
||||
# models and optimizer
|
||||
generator = LinearGen()
|
||||
discriminator = LinearDisc()
|
||||
@@ -83,24 +84,24 @@ if __name__ == "__main__":
|
||||
output_dir = Path(".").resolve() / "outputs"
|
||||
output_dir.mkdir(exist_ok=True)
|
||||
# optimizers
|
||||
optim_g = optim.Adam(get_parameters(generator), lr=0.0002, b1=0.5) # 0.0002 for equilibrium!
|
||||
optim_d = optim.Adam(get_parameters(discriminator), lr=0.0002, b1=0.5)
|
||||
optim_g = optim.Adam(get_parameters(generator),lr=0.0002, b1=0.5) # 0.0002 for equilibrium!
|
||||
optim_d = optim.Adam(get_parameters(discriminator),lr=0.0002, b1=0.5)
|
||||
# training loop
|
||||
with Tensor.train():
|
||||
for epoch in (t := trange(epochs)):
|
||||
loss_g, loss_d = 0.0, 0.0
|
||||
for _ in range(n_steps):
|
||||
data_real = make_batch(X_train)
|
||||
for step in range(k): # Try with k = 5 or 7.
|
||||
noise = Tensor.randn(batch_size, 128)
|
||||
data_fake = generator.forward(noise).detach()
|
||||
loss_d += train_discriminator(optim_d, data_real, data_fake)
|
||||
Tensor.training = True
|
||||
for epoch in (t := trange(epochs)):
|
||||
loss_g, loss_d = 0.0, 0.0
|
||||
for _ in range(n_steps):
|
||||
data_real = make_batch(images_real)
|
||||
for step in range(k): # Try with k = 5 or 7.
|
||||
noise = Tensor.randn(batch_size, 128)
|
||||
data_fake = generator.forward(noise)
|
||||
loss_g += train_generator(optim_g, data_fake)
|
||||
if (epoch + 1) % sample_interval == 0:
|
||||
fake_images = generator.forward(ds_noise).detach().numpy()
|
||||
fake_images = (fake_images.reshape(-1, 1, 28, 28) + 1) / 2 # 0 - 1 range.
|
||||
save_image(make_grid(torch.tensor(fake_images)), output_dir / f"image_{epoch+1}.jpg")
|
||||
t.set_description(f"Generator loss: {loss_g/n_steps}, Discriminator loss: {loss_d/n_steps}")
|
||||
data_fake = generator.forward(noise).detach()
|
||||
loss_d += train_discriminator(optim_d, data_real, data_fake)
|
||||
noise = Tensor.randn(batch_size, 128)
|
||||
data_fake = generator.forward(noise)
|
||||
loss_g += train_generator(optim_g, data_fake)
|
||||
if (epoch + 1) % sample_interval == 0:
|
||||
fake_images = generator.forward(ds_noise).detach().numpy()
|
||||
fake_images = (fake_images.reshape(-1, 1, 28, 28) + 1) / 2 # 0 - 1 range.
|
||||
save_image(make_grid(torch.tensor(fake_images)), output_dir / f"image_{epoch+1}.jpg")
|
||||
t.set_description(f"Generator loss: {loss_g/n_steps}, Discriminator loss: {loss_d/n_steps}")
|
||||
print("Training Completed!")
|
||||
|
||||
Binary file not shown.
|
Before Width: | Height: | Size: 1.6 MiB After Width: | Height: | Size: 1.5 MiB |
Binary file not shown.
|
Before Width: | Height: | Size: 369 KiB After Width: | Height: | Size: 454 KiB |
@@ -3,7 +3,7 @@ from extra.export_model import export_model
|
||||
from examples.llama3 import build_transformer, Tokenizer
|
||||
from tinygrad.nn.state import get_state_dict, load_state_dict
|
||||
from tinygrad import Device, Variable, Tensor, dtypes, TinyJit
|
||||
from tinygrad.helpers import DEV, fetch, Context
|
||||
from tinygrad.helpers import fetch, Context
|
||||
from tiktoken.load import load_tiktoken_bpe, dump_tiktoken_bpe
|
||||
|
||||
def prepare_browser_chunks(model):
|
||||
@@ -115,7 +115,7 @@ if __name__=="__main__":
|
||||
start_pos = Variable("start_pos", 0, max_context).bind(0)
|
||||
model_input = lambda: [Tensor([[tok]]), start_pos, TEMPERATURE, TOP_K, TOP_P, ALPHA_F, ALPHA_P]
|
||||
|
||||
DEV.value = "CPU"
|
||||
Device.DEFAULT="CPU"
|
||||
model = build_transformer(model_path, model_size="1B", quantize="int8", scale_dtype=dtypes.float32, device=Device.DEFAULT, max_context=max_context)
|
||||
state_dict = get_state_dict(model)
|
||||
validate_model(model, tokenizer)
|
||||
@@ -129,7 +129,7 @@ if __name__=="__main__":
|
||||
with open(os.path.join(os.path.dirname(__file__), f"{model_name}.c"), "w") as f: f.write(cprog)
|
||||
with open(os.path.join(os.path.dirname(__file__), "net_clang.js"), "w") as f: f.write(js_wrapper)
|
||||
|
||||
DEV.value = "WEBGPU"
|
||||
Device.DEFAULT="WEBGPU"
|
||||
# float16 is not yet supported for dawn/Vulkan/NVIDIA stack, see: https://issues.chromium.org/issues/42251215
|
||||
# therefore for now, we used CLANG to quantize the float16 llama to int8 with float32 scales, then load to WEBGPU
|
||||
model = build_transformer(model_path, model_size="1B", quantize="int8", max_context=max_context, load_weights=False)
|
||||
|
||||
@@ -4,8 +4,8 @@ from extra.f16_decompress import u32_to_f16
|
||||
from examples.stable_diffusion import StableDiffusion
|
||||
from tinygrad.nn.state import get_state_dict, safe_save, safe_load_metadata, torch_load, load_state_dict
|
||||
from tinygrad.tensor import Tensor
|
||||
from tinygrad import dtypes
|
||||
from tinygrad.helpers import DEV, fetch
|
||||
from tinygrad import Device, dtypes
|
||||
from tinygrad.helpers import fetch
|
||||
from typing import NamedTuple, Any, List
|
||||
import requests
|
||||
import argparse
|
||||
@@ -80,7 +80,7 @@ if __name__ == "__main__":
|
||||
parser = argparse.ArgumentParser(description='Run Stable Diffusion', formatter_class=argparse.ArgumentDefaultsHelpFormatter)
|
||||
parser.add_argument('--remoteweights', action='store_true', help="Use safetensors from Huggingface, or from local")
|
||||
args = parser.parse_args()
|
||||
DEV.value = "WEBGPU"
|
||||
Device.DEFAULT = "WEBGPU"
|
||||
|
||||
model = StableDiffusion()
|
||||
|
||||
|
||||
@@ -4,11 +4,10 @@ from tinygrad.tensor import Tensor
|
||||
from tinygrad.nn.state import safe_save
|
||||
from extra.export_model import export_model
|
||||
from tinygrad.device import Device
|
||||
from tinygrad.helpers import DEV
|
||||
from tinygrad.nn.state import safe_load, load_state_dict
|
||||
|
||||
if __name__ == "__main__":
|
||||
DEV.value = "WEBGPU"
|
||||
Device.DEFAULT = "WEBGPU"
|
||||
yolo_variant = 'n'
|
||||
yolo_infer = YOLOv8(w=0.25, r=2.0, d=0.33, num_classes=80)
|
||||
state_dict = safe_load(get_weights_location(yolo_variant))
|
||||
|
||||
Some files were not shown because too many files have changed in this diff Show More
Reference in New Issue
Block a user