branch: master
Commits on master
- 2a2d323 add test that the compiler isn't used (#3025) 2 years ago
- 520406c add Tensor.unflatten and Tensor.flatten(end_dim) (#3023) 2 years ago
- f432ec9 Bitcast hip fix + fix mixtral (#3022) 2 years ago
- eda4376 use Scalar = Union[float, int, bool] in tensor.py (#3021) 2 years ago
- 60abc62 fast hip read (#3014) 2 years ago
- 4465ef2 add test_softmax to test_ops (#3020) 2 years ago
- 7c80b78 cleanup gpt2 build function (#3018) 2 years ago
- 55e52ab minor cleanup of matvec in hand_coded_optimizations (#3015) 2 years ago
- f88506e move gpt2/llama sampling inside the model call (#3013) 2 years ago
- c2a044e disk_read_speed example 2 years ago
- 8a63f26 make LR scheduler work with multigpu (#3011) 2 years ago
- 8524493 minor gpt2 cleanup (#3012) 2 years ago
- 2b6670d separate entry for HALF hlb_cifar10 in benchmark (#3010) 2 years ago
- 5337211 llvm CMPEQ 2 years ago
- b8c30eb no midcast MULACC for llvm 2 years ago
- 91665ef rewrite MUL CAST SUM to CAST MULACC 2 years ago
- ab7dfd6 use float for acc dtype for half tensor sum 2 years ago
- 6fa285b touchup onnx xor and not (#3008) 2 years ago
- 5781702 removed redundant dtype hacks in onnx_ops (#2939) 2 years ago
- 9f39165 correct (dtype, device) in test_dtype.is_dtype_supported (#3007) 2 years ago
- ae112c9 fix some long lines in tests (#3006) 2 years ago
- 7e191fb hotfix: don't jitcache with 1 kernel. improvements to hip sniffer 2 years ago
- bcc1aa2 make disk simpler (#3002) 2 years ago
- 9699c8c don't alloc for InterpretedASTRunner (#2999) 2 years ago
- bca0b95 bump shapetracker simplify message to DEBUG >= 5 (#2998) 2 years ago
- 74a3043 replace `d[a] if a in d else b` with `d.get(a, b)` (#2997) 2 years ago
- 74cc6fd remove AndNode.__floordiv__ special case (#2996) 2 years ago
- a0c7cb2 hotfix: create weights dir in local tg checkout 2 years ago
- fc36a7d tinygrad weights 2 years ago
- 1ac4d27 remove VariableOrNum from Node.substitute arg (#2995) 2 years ago