Commit Graph

Select branches

Hide Pull Requests

abstractions_4

add_local_buffer

after_qol

algebraic_upat

all_bv_no_idx

all_sink

amd_isel

amd_refac

amd_smu_reset_patch

amd_sqtt

amd_uop

amd_uop_cleanups

assign_copy

assign_try_3

before2

better_spec

better_syn

bitcast_spec

broadcast_shape_2

bt_ei

buf_heur

bump_nv_2404

cache_reshape

call_inline

call_is_kernels

callify

cast_vec_as_cast

cat_mop

changes_from_rangify

check_clsetarg

check_dims

cifar_30

cifar_30_clean

clean_load

clean_spec

clone_tg

codegen_try_2

contig_on_after

continue_mat

deepsek

default_late_allreduce

delete_kernel_py

delete_ones

delete_slow_rangeify

delta_attention

delta_attention_scan

demote_op

disable_rules

dsp_search

dsp_search_merged

dtype_shape

earlier_gater

emu_cdna

emulate_ctx

end_all_ranges

endrange

er_prioity

execution_unit

failing_assign

fancy_usb

fast_llm

fix_fa_fwd_clang_22

fix_fuse

fix_llvm_cast

fix_range_merging_sd

fix_rangeify_tests

fix_rc

fix_sched_copy

fix_slice_store

fix_src_spec

fixup_fix_kernel

full_spec

fun_w_egraphs

gate-on-load-store

gate_rewrite

gemma4_gpt

gen_pdf_fast

gh-pages

gl_dims

gldims

glm_flash

grad_outerworld

grok_x86_novec

group_div_rules

hcq2_vram_copies

heur_4096

heur_group

hk_gemm

http_threading

hwtest_fixes

id_store

if_range_cleanups

ignore_oob

image_idx_xy

image_in_mc

image_no_vec

image_refactors_stack

index_to_slice

invalid_try_3

jinja_template

kernel4

kernel_is_call

kernelless

kimi_slop

kitten_matmul

late_Locals

late_add_load

late_allreduce

late_codegen

late_decomp_is_late

late_lower_idx

late_view

late_weakint

lds_double_buffering

less_wmma_arg

lil_dg_change

lil_rangeify_td

llama_trainer

llm_262k_minimal

llm_flash_attn

llm_qwen

llm_serve_fixes

load_view_late

local_cache

locals_have_device

loop_end_op

loop_is_rng

lowerer_hacks

mac_pytest

manual_kernel_3

master

mega_lowerer

min_new_lin

min_outer_range

minor_views

mixins7

mlperf_training_v4.1

mlperf_training_v5.0

mlperf_training_v6.0

modern_scan

more_apps

more_early_comps

more_folding_rules

more_kernelcount

more_pcode

more_rangeify_pm_tests

more_ren_clean

more_uop_gc_test

more_uop_mm

more_vmap

move_devec

move_dtype_decomps

move_expander

move_gates_to_load_store

move_mop_cleanup

move_schedule

move_simp_views

moveleftright

mt_typing

multi_range

multioutput

new_ck

new_devec

new_invalid_and_image

new_linearizer

new_shape

new_test

no_assign_2

no_buf_0

no_decimals

no_late_simplify_marg

no_merge_ends

no_merge_views

no_mstack_in_allreduce

no_one_ranges

no_sip_usbgpu

no_tsink_base

no_tuplize

no_user_ck_contig

no_vec_x86_4

no_weakint_variable

noop_index

novecimg

num_batches_tracked_shape

olmo3_support

one_compiler

one_hcopt

only_assign_buffers

only_loop_backends

only_reg_emu

only_reg_emu2

outer_range

outerworld_work

pad_invalid_work

pad_mixin

param_call_is_function

parse_sqtt_more

pci_iommu

port_nir

postrange_hcopts

precompile_backward

prefill_qwen_fast

print_special_ops

ps_clean

pyremu_try

python_prg

python_speed

q4k_support

qcom-openpilot-13ms

qcom_830

qcom_mmapeak

qcom_speed_codex

qol_string

qwen36_27b_amd_900

qwen36_27b_amd_minimal

qwen36_amd_bugfixes

qwen_mergable

r_cleanups

r_new_opt

ranged_call

rangeify_2

rangeify_const_folding

rdna3_vibes

rdna4_asm

rdna4_gemm

real_pyrender

recurse_sub

reg_coalese

relu_pattern

remove_assign

remove_cast_hack

remove_ck_contig

remove_define_var

remove_forced_re

remove_on_stack

remove_programspec

remove_pylint

remove_slow_tests

remove_vec_2

remove_vec_3

renderers_use_shapes_2

replace_if_with_range

reprocess_node_experiments

reshape_trait

resnet-ci

revert-11547-fix_mismatch_red

revert-11715-revert-11711-amd_llvm_tc_cast

revert-11809-no_ki_in_gpudims

revert-11811-better_div_nesting

revert-11876-memplan_vbig

revert-12104-simplify_valid_in_symbolic

revert-12308-bufferize_early

revert-12731-jitlink_init_check

revert-12813-after_cleanups

revert-13046-better_cleanup_arange_buffers

revert-13311-prof_link_sqtt

revert-13583-rocr_desc_2

revert-14734-tload

revert-17141-full_rewrite_to_sink

revert-17408-_broadcasted-weak

revert-9438-simple_matmul_np_rand

revert_wrong_ck_fix

rewrite_rangeify2

rtoposort

scan_assign

scan_op_8

sched_cache

sched_lin

sd_load_simple

self_type

set_testpath

shrink_in_render

simp_spec

simple_gidxs

simple_pool

simple_priority

simple_rangeify

simpler-sd-compile

simpler_end

simpler_fusion

simpler_image

simpler_postrange

simpler_reduce

slice_like_shrink

spec_3

spec_cleanup

spec_ren

speed_one_linear

split_buf_idx

split_decomp_pass

split_sub

stack_0

store_is_void

string_ranges

support_opts_in_contig

swizzle_in_kernel

sym_llm

sym_work

symbolic-chunk-split

szdiff_bot_test

tc_is_heur

tensor_graph_single_sink

test_amd_crash

test_fa

test_from_postopt

test_hgemm

test_unit

tiny_scan

tmr

tool_call_support

topoprop

ttomsa-new_linearizer

tuplegtuple

two_stage_remove

typed_checks

typing_work

unused_permute_arg

uop_prg

upd_test4

update_benchmark

update_benchmark_search

update_mlperf

user_failure

vecless_x86_glm

verify_sink_dims_clean

vf_ish

view_after_const

view_in_codegen

viz_cbroadcast

warp_fun

wgpu-f16-packed

working_postopt

x86_moves

x86_numel

v0.10.0

v0.10.1

v0.10.2

v0.10.3

v0.11.0

v0.12.0

v0.13.0

v0.4.0

v0.5.0

v0.6.0

v0.7.0

v0.8.0

v0.9.0

v0.9.1

v0.9.2

Mono Color

• ee502c8055 fixup to_movement_ops and add back to CI (#3881) chenyuandGitHub 2024-03-22 18:14:49 -04:00
• 16e31f7f0d init multidevice cuda graph (#3858) 3 people 2024-03-22 23:49:48 +03:00
• 0c197b9cf3 hotfix: hip bfloat formatting geohot 2024-03-22 11:52:05 -07:00
• 54dc48aa47 fix assign (#3878) George HotzandGitHub 2024-03-22 11:48:48 -07:00
• 5587594a00 fuzz_linearizer: add --ast and --file params to read kernels (#3877) Francis LamandGitHub 2024-03-22 11:27:40 -07:00
• c5467e5bd6 diverse test value in test_dtype DATA based on dtype (#3864) chenyuandGitHub 2024-03-22 14:22:06 -04:00
• 86ee36e697 preschedule all (#3875) George HotzandGitHub 2024-03-22 11:20:06 -07:00
• d8c3f1894a Use UOpGraph in test (#3876) Szymon OżógandGitHub 2024-03-22 19:12:38 +01:00
• 1c51d586ea replace raise Exception with specific errors (#3874) chenyuandGitHub 2024-03-22 12:32:21 -04:00
• 8ef5490ec8 cuda tranfer + async copyin (#3873) nimlgenandGitHub 2024-03-22 19:01:37 +03:00
• 624bc89910 PTX - implement float 4, ptr arithmetics and other speed improvements (#3775) 3 people 2024-03-22 16:54:02 +01:00
• f4055439dc don't include hip common (#3851) George HotzandGitHub 2024-03-22 08:50:50 -07:00
• 4a27ce6ec9 tiny version of amd_hip_bfloat16 (#3868) qazalandGitHub 2024-03-22 17:37:30 +02:00
• 82ce60e172 use JIT_BATCH_SIZE=4 for GPT2 3090 benchmark (#3870) chenyuandGitHub 2024-03-22 00:40:06 -04:00
• fe6ceff15f proposal: multioutput JIT spec (#3856) qazalandGitHub 2024-03-22 06:28:30 +02:00
• a26090d404 search: change to use "spawn" and limit the number of tasks per child (#3862) Francis LamandGitHub 2024-03-21 21:23:36 -07:00
• dca69df197 hot fix use DEBUG >= 3 for allreduce message (#3869) chenyuandGitHub 2024-03-21 23:40:44 -04:00
• 6729f20aab Ring allreduce try 2 (#3852) 5 people 2024-03-22 01:17:51 +02:00
• 3c0478bfab fuzz_linearizer: add additional DEBUG info for comparison errors (#3866) Francis LamandGitHub 2024-03-21 15:58:10 -07:00
• bc482729d0 lower hlb_cifar acc to 93.3 (#3865) chenyuandGitHub 2024-03-21 17:58:53 -04:00
• e50b7abe4f diversed buf inputs based on dtype in fuzz_linearizer (#3863) chenyuandGitHub 2024-03-21 16:23:11 -04:00
• c40f78499f reuse fuzz_linearizer.compare_linearizer in test_linearizer_failures (#3861) chenyuandGitHub 2024-03-21 14:23:37 -04:00
• 30fa03243e reuse fuzz_linearizer.compare_linearizer in test_linearizer_failures (#3861) chenyuandGitHub 2024-03-21 14:12:27 -04:00
• 33dd99acf4 remove helper_add_store from test_linearizer_failures (#3860) chenyuandGitHub 2024-03-21 12:53:31 -04:00
• 6bf0b82267 alloc new output in fuzz_linearizer between baseline and real one (#3859) chenyuandGitHub 2024-03-21 11:36:05 -04:00
• b78352b423 do not create structs every call in CUDAProgram (#3855) nimlgenandGitHub 2024-03-21 17:51:40 +03:00
• e5745c1a0d fix nan on multigpus cuda (#3854) nimlgenandGitHub 2024-03-21 15:21:55 +03:00
• 4e0819e40b fixing the benchmark not printing in handcode resnet50 opt example (#3850) Anurag LamsalandGitHub 2024-03-21 10:40:31 +05:45
• 85691c8e20 fix hsa sync issue (#3847) nimlgenandGitHub 2024-03-21 04:00:30 +03:00
• f271cd682b user _resolve_dim in argmax (#3846) chenyuandGitHub 2024-03-20 20:17:30 -04:00
• 5c4cf62d2c fix View.pad arg type (#3845) chenyuandGitHub 2024-03-20 19:36:02 -04:00
• 6d5dec2fef log optimized kernels and a script to compare with non-optimized ones (#3829) Francis LamandGitHub 2024-03-20 16:22:08 -07:00
• 9d1d08fbb0 show llama bandwith with timing (#3844) chenyuandGitHub 2024-03-20 17:19:15 -04:00
• 7ff47e45a1 cifar TARGET_EVAL_ACC_PCT=93.5 (#3843) chenyuandGitHub 2024-03-20 16:56:51 -04:00
• 92c5067439 conceptual small refactor (#3842) qazalandGitHub 2024-03-20 22:46:14 +02:00
• 519336cfea factor out partial in SumNode div int (#3841) chenyuandGitHub 2024-03-20 16:34:33 -04:00
• 8cb5215885 Revert "Ring allreduce in multitensor (#3000)" (#3840) George HotzandGitHub 2024-03-20 11:41:49 -07:00
• c5bf9e4c96 Ring allreduce in multitensor (#3000) 5 people 2024-03-20 20:20:01 +02:00
• 455f7bea9b test example from half resnet that idx has number outside of int32 (#3838) chenyuandGitHub 2024-03-20 13:44:20 -04:00
• 727de5ba1e llama 7B on 3090 benchmark (#3837) chenyuandGitHub 2024-03-20 12:48:22 -04:00
• 9452994201 add a better error message for resnet training (#3836) qazalandGitHub 2024-03-20 18:22:15 +02:00
• 47b9cc2dfe use float32 for rand buffer in test_beam_search and test in metal (#3831) chenyuandGitHub 2024-03-19 23:22:58 -04:00
• d17900bc45 use int32 instead of default_int in simplify_phi_loops (#3828) chenyuandGitHub 2024-03-19 17:49:58 -04:00
• 2d54e4d747 clean up hsa driver (#3818) nimlgenandGitHub 2024-03-20 00:17:41 +03:00
• 99cbc24390 use dtypes.int32 as return dtype for functions that return indices (#3827) chenyuandGitHub 2024-03-19 17:06:57 -04:00
• fa1921ec7d move test_dtype tests to test dtype and output value (#3826) chenyuandGitHub 2024-03-19 16:31:27 -04:00
• 131bbb6563 test_linearizer_failure: add failure 27 from a gpt2 kernel (#3825) Francis LamandGitHub 2024-03-19 13:29:50 -07:00
• 3fb13ff892 HIP -> HSA in docs/env_vars (#3824) nimlgenandGitHub 2024-03-19 22:53:33 +03:00
• 9851e2c3b9 test_linearizer_failure: add failure 26 from a gpt2 kernel (#3821) Francis LamandGitHub 2024-03-19 10:19:54 -07:00
• b436c9792f Fix factoring bug (O(n) arange related) (#3817) 3 people 2024-03-19 08:49:42 -07:00
• e12bc85014 use BS=128 and BS=768 for resent benchmark (#3815) chenyuandGitHub 2024-03-18 23:49:55 -04:00
• e6d55932ca hotfix: this makes beautiful mnist work again, not okay geohot 2024-03-18 18:22:44 -07:00
• a6ed2ae3c6 use old cumsum optimization for arange (#3813) chenyuandGitHub 2024-03-18 20:01:03 -04:00
• ac866eaf5a disable simplify_phi_loops (#3812) chenyuandGitHub 2024-03-18 19:25:26 -04:00
• 4c4d3cb3e3 restrict assignment to base (#3809) George HotzandGitHub 2024-03-18 15:33:06 -07:00
• 20681d5c4a remove old dist multigpu (#3811) chenyuandGitHub 2024-03-18 18:31:05 -04:00
• 5dd048a378 remove HIP in core tinygrad (#3810) chenyuandGitHub 2024-03-18 18:19:27 -04:00
• a7afd2f6bf test_linearizer_failures: add failing kernel from GPT2 CUDA (#3808) Francis LamandGitHub 2024-03-18 14:16:40 -07:00
• d8296d4a3f simple assign tests (#3807) George HotzandGitHub 2024-03-18 13:57:01 -07:00
• a0ab755317 threefry again (#3785) 3 people 2024-03-18 16:47:07 -04:00
• 7af7467f22 Fix: BEAM search with PTX fails (#3786) Elias WahlandGitHub 2024-03-18 21:13:32 +01:00
• 629757eaa1 hotfix: update inputs of correct transfers in hsagraph (#3800) nimlgenandGitHub 2024-03-18 22:52:27 +03:00
• 1711274654 7B llama on 4 gpus on benchmark (#3804) chenyuandGitHub 2024-03-18 14:32:37 -04:00
• d79a1d315b add outbufs back (#3803) qazalandGitHub 2024-03-18 19:30:53 +02:00
• 7fe08fa5a0 hotfix: put that assign hack back geohot 2024-03-18 09:25:05 -07:00
• 0183a05f0a test assign (#3798) George HotzandGitHub 2024-03-18 08:58:04 -07:00
• 1e1beb888c Revert "add failing assign test (#3796)" (#3797) George HotzandGitHub 2024-03-18 08:55:36 -07:00
• 2dea12832c add failing assign test (#3796) George HotzandGitHub 2024-03-18 08:47:30 -07:00
• e78df485c7 update inputs for transfers in hsagraph (#3560) nimlgenandGitHub 2024-03-18 18:01:04 +03:00
• 086291e8c6 hotfix: add test for JIT reset geohot 2024-03-17 21:35:49 -07:00
• dccefab23f remove mixtral weight to clang first (#3792) chenyuandGitHub 2024-03-17 23:33:17 -04:00
• bf3e1c4df2 support pickling tensors and others (#3787) George HotzandGitHub 2024-03-17 18:29:14 -07:00
• 5ac1fa933f apply the same fix_bf16 in llama and coder (#3789) chenyuandGitHub 2024-03-17 21:25:24 -04:00
• 639bd5dbfc move bf16 cast hack to Tensor.llvm_bf16_cast (#3788) chenyuandGitHub 2024-03-17 18:51:22 -04:00
• 311cf2b7d3 Revert "threefry_2x32 (#2601)" (#3784) George HotzandGitHub 2024-03-17 10:27:20 -07:00
• db3de54bc4 threefry_2x32 (#2601) 3 people 2024-03-17 13:19:33 -04:00
• 53adcb34f5 remove hip backend (#3783) George HotzandGitHub 2024-03-17 10:12:16 -07:00
• 2a14d1b5e0 Revert "add outbufs info to CompiledASTRunner (#3781)" (#3782) George HotzandGitHub 2024-03-17 09:47:23 -07:00
• 722dd4276c add outbufs info to CompiledASTRunner (#3781) qazalandGitHub 2024-03-17 16:52:20 +02:00
• 9255332d9e use llvm as bridge to fix_bf16 loading (#3774) chenyuandGitHub 2024-03-16 15:22:19 -04:00
• 987a055c0d increase jit batch size progressivly (#3771) 3 people 2024-03-16 18:58:11 +03:00
• 77febb44e6 llama 7B on 6 gpus benchmark (#3773) chenyuandGitHub 2024-03-16 11:38:52 -04:00
• 07324b56d5 [experimenting] use contiguous instead of realize in optim (#3770) David HouandGitHub 2024-03-15 23:06:50 -07:00
• e3e89c244b multioutput uoping infra (#3706) qazalandGitHub 2024-03-16 06:56:59 +02:00
• e1c5aa9cce estimated resnet training time for BENCHMARK (#3769) chenyuandGitHub 2024-03-15 22:36:58 -04:00
• 0870dd5b3b hotfix: switch resnet training from HIP -> HSA in CI geohot 2024-03-15 13:35:52 -07:00
• d8070876d2 bfs scheduler, infra for multioutput (#3763) qazalandGitHub 2024-03-15 22:33:58 +02:00
• 91e181ee02 make alignment readable (#3766) nimlgenandGitHub 2024-03-15 23:18:40 +03:00
• 8ea53951c1 bfloat16 Tensor.rand (#3764) chenyuandGitHub 2024-03-15 15:05:13 -04:00
• a2d3cf64a5 move is_dtype_supported to test.helpers (#3762) chenyuandGitHub 2024-03-15 14:33:26 -04:00
• 8af87e20a0 unrealized, assign is replace (#3761) George HotzandGitHub 2024-03-15 11:17:59 -07:00
• 922f8319cb Run test_real_world in METAL test (#3760) chenyuandGitHub 2024-03-15 13:56:52 -04:00
• 4bd5535d72 update mlperf resnet default hparams (#3758) chenyuandGitHub 2024-03-15 12:09:26 -04:00
• aad9332e21 remove that extra assign line, is it fixed? (#3757) George HotzandGitHub 2024-03-15 08:59:41 -07:00
• ba79a3c09a some hsa lines saving + fixes (#3752) nimlgenandGitHub 2024-03-15 18:12:18 +03:00
• ca19eb3e82 where fold try 2 (#3748) 3 people 2024-03-15 07:46:26 -07:00
• 6b8c66e04f fix broken loops in llvm (#3751) nimlgenandGitHub 2024-03-15 11:57:51 +03:00
• d3a6319630 bf16 tests in test_dtype.py (#3749) chenyuandGitHub 2024-03-15 00:17:11 -04:00
• 33c01c9db0 Fix kwargs in JIT (#3730) Rohan PotdarandGitHub 2024-03-14 23:55:19 -04:00
• 641f347232 simple LoadOps.ASSIGN (#3745) George HotzandGitHub 2024-03-14 20:44:34 -07:00