branch: master
Commits on master
- 371005c use one kvcache tensor in gpt2 instead of two separate caches (#2662) 2 years ago
- 5a7b2ff masked shapetrackers (#2657) 2 years ago
- b931a20 minor shapetracker cleanup (#2652) 2 years ago
- c704a77 green dtypes ALU tests (#2617) 2 years ago
- 71d989b adding test to cover #2644 failure (#2645) 2 years ago
- 50dcd53 Get all WEBGPU test_ops passing (#2646) 2 years ago
- 0978c24 fast gpt2 embedding with variable bs=1 (#2596) 2 years ago
- 229ada5 Gpt2 benchmark with HALF and BEAM (#2636) 2 years ago
- a735799 mlx benchmark, a lil slower than tg 2 years ago
- 7c427d7 don't apply padding on script call (#2585) 2 years ago
- 9d7ead8 hotfix: no need for model cache in examples/coder.py 2 years ago
- be09cc8 Bitcast support / fast bf16 load (#2011) 2 years ago
- 232ed2a more test cleanups (#2631) 2 years ago
- a63f48d gpt2 half for kvcache and output logits (#2630) 2 years ago
- 0be5d16 only 62 gflops (#2629) 2 years ago
- 6d58c19 binaryops xor (#2627) 2 years ago
- c53e854 cast image doesn't work on nvidia (#2626) 2 years ago
- 8c67eb1 GPT bugfixes (#2624) 2 years ago
- 8903a40 update the onnx test so cuda local run passes (#2623) 2 years ago
- ec594cf hotfix: tasteful ctrl-c in parallel beam 2 years ago
- 35b5e95 parallel beam search (#2610) 2 years ago
- 9996f1a no document prs (#2622) 2 years ago
- dd8b463 regression test for reshape fix #2616 (#2620) 2 years ago
- c257a0d minor reshape cleanups (#2619) 2 years ago
- a6b68e8 fix for false merge (#2616) 2 years ago
- fc00da5 helper functions for test_indexing.py (#2615) 2 years ago
- 7322ab8 onnx tests with different dtypes (#2612) 2 years ago
- f12bccc [ready] refactor getitem round 2 :D (#2568) 2 years ago
- 6ba6349 JIT=0 llama.py should not jit (#2609) 2 years ago
- 41d6961 hotfix: forking works okay in HIP now 2 years ago