strisunshinewentingwang/pytorch

Pull Requests Commits

[inductor] dropout avoid zero division error

Valentine233 committed 2 years ago

787fa9a2

Revert "[c10d] Comment out ddp_hook_with_optimizer_parity tests (#100215)"

pytorchmergebot committed 2 years ago

9609aeef

Improved message to suppress errors in _dynamo/exc.py (#97345)

vfdev-5 committed 2 years ago

0692bdd9

[Docs] Fix docstring format (#99396)

zhouzaida committed 2 years ago

b51f92eb

Add directly referenced header files for "ceil_div.h" (#99607)

caizhi-mt committed 2 years ago

64efd888

[CI] Start to collect inference perf with cpp_wrapper ON (#100187)

desertfire committed 2 years ago

3e87fc52

per-Tensor `grad_fn` for in-place foreach functions (#96405)

crcrpar committed 2 years ago

67401890

More rigorous mixed overloads on SymInt (#100008)

ezyang committed 2 years ago

9a3e411a

Delete pytorch/caffe2/contrib/docker-ubuntu-14.04 (#100155)

kit1980 committed 2 years ago

9dcabe29

[FSDP] Remove unneeded disable of tf32 (#100179)

Andrew Gu committed 2 years ago

d1fbd33c

[FSDP] Subtest sharding strategy in test_fsdp_grad_acc.py (#100178)

Andrew Gu committed 2 years ago

1f4183e2

[c10d] Comment out ddp_hook_with_optimizer_parity tests (#100215)

kwen2501 committed 2 years ago

ae40a6c7

Add tensor to fake clone snapshot for immutable source of truth (#100128)

voznesenskym committed 2 years ago

a145a333

Not flatten states when use_orig_param is True and sharding is NO_SHARD (#100189)

zhaojuanmao committed 2 years ago

ca1cf434

Inductor cpp wrapper: support LinearBinary (#99957)

chunyuan-w committed 2 years ago

3241fbd6

Added Typechecking to input tensor in RNN (#100100)

TamirCohen committed 2 years ago

02211987

refactor test_sdpa into two test classes to account for failure modes (#100121)

drisspg committed 2 years ago

b8d7a28e

Avoid elementwise dispatch of gradient unscaling/validation ops in `_foreach_non_finite_check_and_unscale_cpu_` (#100108)

speediedan committed 2 years ago

477ca178

[ONNX] Refactor test_op_consistenct.py and test_fx_op_consistency.py (#100172)

titaiwangms committed 2 years ago

1dba53cb

Make DimConstraints create actionable message (#100103)

ydwu4 committed 2 years ago

61917a00

Check for debug mode (#92707)

ZainRizvi committed 2 years ago

d5f15d35

[Feature] storage resize_ support custom device. (#99882)

wbigat committed 2 years ago

b02aa5e7

Get SchemaCheckMode to error on ops that return inputs directly. Expose as a dynamo backend, eager_debug (#99744)

bdhirsh committed 2 years ago

9834358e

move SchemaCheckMode to torch/_subclasses (#99743)

bdhirsh committed 2 years ago

1f2d00e5

Pass torch.compile mode/options to all backends (#99645)

jansel committed 2 years ago

884c5c86

[ONNX] Add test_fx_op_consistency.py (#99465)

titaiwangms committed 2 years ago

7295ab67

Decompose arange.default to arange.start_step (#99739)

angelayi committed 2 years ago

d06b93b0

[vmap] Fix searchsorted batch rule (#99698)

qqaatw committed 2 years ago

a67fa845

Do not use `--extra-index-url` in testing wheels (#100183)

malfet committed 2 years ago

991b1c02

[quant][pt2e] remove dropout from fx quant

mcr229 committed 2 years ago

Verified 151d76cc

Older