Commit Graph

Select branches

Hide Pull Requests

geohot/updates

gh-pages

master

remove_anchors

#1

#2

#3

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

• f502c9b08f minor cleanup of View.reshape (#3088) chenyuandGitHub 2024-01-11 13:05:54 -05:00
• f40299c3fe remove the third merging state in view._merge_dims (#3085) chenyuandGitHub 2024-01-11 12:07:43 -05:00
• 7f9590d357 hotfix disable flaky mac runner wino cifar (#3087) chenyuandGitHub 2024-01-11 11:57:05 -05:00
• adcc844755 cat works (#3086) Yixiang GaoandGitHub 2024-01-11 08:25:20 -08:00
• cdeab9ad97 mem_estimate is always int, not symbolic (#3083) chenyuandGitHub 2024-01-10 23:39:51 -05:00
• 162fa61a32 wmma: clean up device specific tensor core code (#3081) Francis LamandGitHub 2024-01-10 18:03:09 -08:00
• d218d13885 minor cleanups of lazy.py (#3080) chenyuandGitHub 2024-01-10 20:17:56 -05:00
• 56dda33fc6 Tensor.expand resolves the new_shape before shortcut return (#3078) chenyuandGitHub 2024-01-10 14:29:15 -05:00
• 6842476ca6 better test demonstration (#3077) Yixiang GaoandGitHub 2024-01-10 10:50:52 -08:00
• 507e0afba0 fix onehot and jit in examples/transformer (#3073) chenyuandGitHub 2024-01-10 02:22:41 -05:00
• 4342fccc83 filter_strides -> canonicalize_strides (#3072) chenyuandGitHub 2024-01-10 01:06:48 -05:00
• 023f5df0e9 simpler idxs_to_idx (#3071) chenyuandGitHub 2024-01-10 00:30:10 -05:00
• 2495ca95c7 early gate the graph (#3070) George HotzandGitHub 2024-01-09 20:17:13 -08:00
• ff0d6e4551 jit autorealizes output (#3069) George HotzandGitHub 2024-01-09 20:10:22 -08:00
• ae83733431 hotfix: examples/transformer.py geohot 2024-01-09 19:25:37 -08:00
• 145718a90f unbind view or shapetracker also returns var_val (#3067) chenyuandGitHub 2024-01-09 21:45:05 -05:00
• ef3aa6d7fb update gh actions (#3033) 3 people 2024-01-10 01:52:22 +00:00
• 3f80c1a098 speedtweaks3: apply shouldn't use the tensor constructor (#3065) George HotzandGitHub 2024-01-09 17:42:33 -08:00
• 0abe72b677 hotfix: use is for enum compare, a few more geohot 2024-01-09 16:53:13 -08:00
• b2b5849f74 hotfix: use is for enum compare geohot 2024-01-09 16:47:27 -08:00
• ac3f246c11 cached size (#3060) George HotzandGitHub 2024-01-09 16:37:37 -08:00
• 73b72b8de2 test scaled dot product attention (#3063) Yixiang GaoandGitHub 2024-01-09 14:30:57 -08:00
• 55ac2a2cf7 Tensor.cat with 0 shape tensors (#3062) chenyuandGitHub 2024-01-09 16:54:06 -05:00
• f0d7ad8aaa fix gpt2 attention with start_pos = 0 (#3061) chenyuandGitHub 2024-01-09 16:14:55 -05:00
• 39b91131bc Speed tweaks (#3059) George HotzandGitHub 2024-01-09 11:34:17 -08:00
• bf6281f316 hotfix: remove useless slow assert from ShapeTracker geohot 2024-01-09 10:56:36 -08:00
• 4b687af98f explicit lazybuffer caching (#3058) George HotzandGitHub 2024-01-09 10:52:37 -08:00
• 2c6f2e899d No extra vars call (#3054) v0.8.0 George HotzandGitHub 2024-01-09 09:52:58 -08:00
• 259bf9bffc add multigpu test for RMSNorm (#3056) Yixiang GaoandGitHub 2024-01-09 09:52:51 -08:00
• dab8214103 unit tests for Device.canonicalize (#3055) chenyuandGitHub 2024-01-09 12:47:20 -05:00
• 374f7659a7 remove unused reciprocal (#3053) George HotzandGitHub 2024-01-09 08:59:04 -08:00
• a686663657 make Embedding device aware for multigpu (#3051) Yixiang GaoandGitHub 2024-01-08 20:09:26 -08:00
• 19298e7a3f Device._buffers -> Device._devices (#3052) chenyuandGitHub 2024-01-08 21:30:38 -05:00
• 4f4e8634b8 use in_features directly in nn.Linear.__init__ bound check (#3050) chenyuandGitHub 2024-01-08 19:32:35 -05:00
• ee6a73826b clean up test_nn.py (#3049) chenyuandGitHub 2024-01-08 18:45:03 -05:00
• 3eb3664074 fix nn.Embedding with empty length input (#3048) chenyuandGitHub 2024-01-08 18:08:36 -05:00
• 7ea2e0035b move children for speed (#3047) George HotzandGitHub 2024-01-08 15:02:32 -08:00
• 655c6f61d3 St real size (#3046) George HotzandGitHub 2024-01-08 14:44:53 -08:00
• 1d730b8853 remove ACCUM_FP32 in simple_matmul.py (#3045) chenyuandGitHub 2024-01-08 17:37:57 -05:00
• 47d67da830 track the size in the lazybuffer (#3044) George HotzandGitHub 2024-01-08 13:44:55 -08:00
• c003be7309 Revert "track size in shapetracker" (#3043) George HotzandGitHub 2024-01-08 13:13:39 -08:00
• 50754f1494 add caches there (#3042) George HotzandGitHub 2024-01-08 13:02:16 -08:00
• c5a941d466 webgl backend in extra (#3041) 3 people 2024-01-08 09:29:13 -08:00
• 8cbcd1b342 Remove webgpu, back to 5k lines (#3040) George HotzandGitHub 2024-01-08 09:10:07 -08:00
• cf2eea961c more beautiful_cartpole with exposed hparams geohot 2024-01-07 17:41:09 -08:00
• 44618427f1 add bf16 type_map for both cuda and hip (#3036) Yixiang GaoandGitHub 2024-01-07 14:26:55 -08:00
• ef5f545fd8 add more Tensor.clip test cases (#3034) chenyuandGitHub 2024-01-07 13:08:59 -05:00
• c9371f0d31 hotfix llama conversation mode (#3031) chenyuandGitHub 2024-01-06 16:57:07 -05:00
• fa707c81e5 move beautiful cartpole action sampling inside jit (#3028) chenyuandGitHub 2024-01-06 00:39:55 -05:00
• ebb81e8f11 hotfix: st.size() -> st.size in llama geohot 2024-01-05 20:18:52 -08:00
• a8ba1ac08f track size in shapetracker (#3026) George HotzandGitHub 2024-01-05 20:15:53 -08:00
• 138c17c094 enable argmax tests for METAL/WEBGPU in CI (#3027) chenyuandGitHub 2024-01-05 21:43:00 -05:00
• 2a2d3233d2 add test that the compiler isn't used (#3025) George HotzandGitHub 2024-01-05 17:24:01 -08:00
• 520406cf3a add Tensor.unflatten and Tensor.flatten(end_dim) (#3023) chenyuandGitHub 2024-01-05 17:55:29 -05:00
• f432ec9c33 Bitcast hip fix + fix mixtral (#3022) George HotzandGitHub 2024-01-05 14:51:25 -08:00
• eda43767de use Scalar = Union[float, int, bool] in tensor.py (#3021) chenyuandGitHub 2024-01-05 13:56:26 -05:00
• 60abc62a3f fast hip read (#3014) George HotzandGitHub 2024-01-05 10:33:13 -08:00
• 4465ef28c5 add test_softmax to test_ops (#3020) chenyuandGitHub 2024-01-05 11:19:49 -05:00
• 7c80b78be9 cleanup gpt2 build function (#3018) chenyuandGitHub 2024-01-04 23:14:53 -05:00
• 55e52abeba minor cleanup of matvec in hand_coded_optimizations (#3015) chenyuandGitHub 2024-01-04 19:43:49 -05:00
• f88506e630 move gpt2/llama sampling inside the model call (#3013) chenyuandGitHub 2024-01-04 17:01:50 -05:00
• c2a044ed83 disk_read_speed example geohot 2024-01-04 13:58:39 -08:00
• 8a63f26a0f make LR scheduler work with multigpu (#3011) Yixiang GaoandGitHub 2024-01-04 12:10:56 -08:00
• 8524493748 minor gpt2 cleanup (#3012) chenyuandGitHub 2024-01-04 13:53:18 -05:00
• 2b6670d2ea separate entry for HALF hlb_cifar10 in benchmark (#3010) chenyuandGitHub 2024-01-04 13:24:10 -05:00
• 5337211058 llvm CMPEQ chenyu 2024-01-04 10:01:12 -08:00
• b8c30eb358 no midcast MULACC for llvm chenyu 2024-01-04 09:53:12 -08:00
• 91665ef143 rewrite MUL CAST SUM to CAST MULACC chenyu 2024-01-04 09:42:22 -08:00
• ab7dfd637b use float for acc dtype for half tensor sum chenyu 2024-01-03 18:56:42 -08:00
• 6fa285b943 touchup onnx xor and not (#3008) chenyuandGitHub 2024-01-04 02:02:42 -05:00
• 57817028bb removed redundant dtype hacks in onnx_ops (#2939) 3 people 2024-01-04 14:45:24 +08:00
• 9f39165188 correct (dtype, device) in test_dtype.is_dtype_supported (#3007) chenyuandGitHub 2024-01-04 00:25:37 -05:00
• ae112c9dbe fix some long lines in tests (#3006) chenyuandGitHub 2024-01-03 23:53:33 -05:00
• 7e191fbb86 hotfix: don't jitcache with 1 kernel. improvements to hip sniffer geohot 2024-01-03 19:17:08 -08:00
• bcc1aa21ac make disk simpler (#3002) George HotzandGitHub 2024-01-03 17:46:21 -08:00
• 9699c8c90b don't alloc for InterpretedASTRunner (#2999) George HotzandGitHub 2024-01-03 17:05:53 -08:00
• bca0b95ee3 bump shapetracker simplify message to DEBUG >= 5 (#2998) chenyuandGitHub 2024-01-03 20:00:36 -05:00
• 74a30431b4 replace d[a] if a in d else b with d.get(a, b) (#2997) chenyuandGitHub 2024-01-03 18:10:25 -05:00
• 74cc6fd3c2 remove AndNode.__floordiv__ special case (#2996) chenyuandGitHub 2024-01-03 17:44:55 -05:00
• a0c7cb2564 hotfix: create weights dir in local tg checkout geohot 2024-01-03 14:14:33 -08:00
• fc36a7d669 tinygrad weights geohot 2024-01-03 14:09:28 -08:00
• 1ac4d27869 remove VariableOrNum from Node.substitute arg (#2995) chenyuandGitHub 2024-01-03 17:02:25 -05:00
• 65dc3700b7 hip device is default on supported platforms (#2993) George HotzandGitHub 2024-01-03 13:42:13 -08:00
• 77c98a1543 hotfix: remove weights directory geohot 2024-01-03 13:40:39 -08:00
• 0be0f2f745 remove stable diffusion test on tinymac geohot 2024-01-03 13:18:24 -08:00
• a354ec9dad Revert "hotfix: HIP is the default device on HIP platforms" geohot 2024-01-03 13:16:54 -08:00
• b748b569f5 hotfix: HIP is the default device on HIP platforms geohot 2024-01-03 13:13:52 -08:00
• 753a7ecc05 Hip driver (#2992) George HotzandGitHub 2024-01-03 12:53:47 -08:00
• f290ca3924 hotfix: save lines in graph geohot 2024-01-03 12:03:42 -08:00
• bc4b6e758b Merge pull request #2981 from g1y5x3/cifar_fp16 Yixiang GaoandGitHub 2024-01-03 11:15:42 -08:00
• d7d5a487ad hotfix: all device canonicalize should be done in Tensor geohot 2024-01-03 10:48:04 -08:00
• ea3bc2f509 remove wino benchmark for now Yixiang Gao 2024-01-03 10:46:43 -08:00
• 5663dd46b6 Merge branch 'master' of github.com:tinygrad/tinygrad into cifar_fp16 Yixiang Gao 2024-01-03 10:11:46 -08:00
• 81b97cd2c6 canonicalize device in LazyBuffer constructor (#2991) chenyuandGitHub 2024-01-03 12:55:25 -05:00
• db525cf8c2 multitensor failed test case with +1 then sum on DEVICE:0 (#2990) chenyuandGitHub 2024-01-03 12:17:11 -05:00
• 7f1802cd50 update benchmark Yixiang Gao 2024-01-03 09:09:34 -08:00
• 5dbaaa7061 hotfix: make multitensor shard contiguous geohot 2024-01-03 08:48:30 -08:00
• 590268fa03 out_tokens -> grouped in linearizer (#2989) chenyuandGitHub 2024-01-03 11:45:28 -05:00
• 8e1fd6ae9d test works Yixiang Gao 2024-01-03 07:22:01 -08:00
• 4f89f8b73a make sure the old hyp breaks the test Yixiang Gao 2024-01-03 07:13:54 -08:00