branch: master
Commits on master
- 6cb74bb fix using clone with shrink [pr] (#8724) 1 year ago
- af65331 update beam params for bert green [pr] (#8726) 1 year ago
- 907dfa0 image buffer realization spec [pr] (#8420) 1 year ago
- 49b914e simpler bert acc [pr] (#8714) 1 year ago
- 93fb50c allreduce: add flags (#8713) 1 year ago
- 8914368 remove buffer size check in schedule item [pr] (#8712) 1 year ago
- 2dae467 scheduler + process_replay import cleanup (#8711) 1 year ago
- e3d1464 move assign preload out of schedule item [pr] (#8710) 1 year ago
- 9a90791 envvar BERT_LAYERS [pr] (#8709) 1 year ago
- 9f6d545 bert log global_norm in training step [pr] (#8708) 1 year ago
- c5e46c5 am: recover from any boot interrupt (#8703) 1 year ago
- 1e283c3 remove realize in bert model init [pr] (#8707) 1 year ago
- 018edd9 don't use view in copy [pr] (#8704) 1 year ago
- d6bf1fe remove the "no copy" line from copy_to_device (#8702) 1 year ago
- 3628f89 fix deallocate for subbuffers (#8701) 1 year ago
- 6733a3a am: fix typo (#8700) 1 year ago
- f0d424e Tensor UOps can become a buffer or const after scheduling (#8698) 1 year ago
- e2008c9 allow symbolic shape in tensor const parents [pr] (#8699) 1 year ago
- 2b239db temp() with usernames (#8697) 1 year ago
- 66ac008 more high level contiguous tests + scheduler deletions [pr] (#8695) 1 year ago
- 08eb1f1 simplify tensors before scheduling [pr] (#8580) 1 year ago
- 02ad450 add failing assert for gradient realization [pr] (#8692) 1 year ago
- b14c984 small changes to make the tensor_map_simple diff cleaner [pr] (#8691) 1 year ago
- 1a15c0e Move define_acc down an unrolled add chain (#8404) 1 year ago
- dd82b4c make onnx runner a class (#8647) 1 year ago
- 46a8c5e delete forced_realize (#8615) 1 year ago
- 679b1ad move softmax upcast to after subtracting max (#8684) 1 year ago
- 08ca871 am: remove pm block (#8688) 1 year ago
- 9d3c406 am: fast memory manager (#8654) 1 year ago
- 9e55495 fold double contiguous [pr] (#8687) 1 year ago