strisunshinewentingwang/pytorch

Pull Requests Commits

benchmark: higher tolerance for RobertaForQuestionAnswering (#107376)

blzheng committed 1 year ago

b9befc53

benchmark: convert output of fp64 to torch.float64 (#107375)

blzheng committed 1 year ago

1ea83f04

[Compiled Autograd] Improve nyi error messages (#106176)

jansel committed 1 year ago

d77e95c3

[inductor] Improve handling of index_expr with floating point dtypes (#105021)

peterbell10 committed 1 year ago

59c54246

Fix wrong error messages with torch.nn.AdaptiveMaxPool1d (#107450)

FFFrog committed 1 year ago

3b160ecc

Revert "Fakify leaf of FunctionalTensor (#107062)"

pytorchmergebot committed 1 year ago

96c5be8b

Enable a number inductor of tests on CPU (#107465)

lezcano committed 1 year ago

c1cc74c7

[cpu] add sdpa choice and UT (#105131)

Valentine233 committed 1 year ago

71632d4d

[CPU] Enable fused_attention pattern matcher (#107128)

Valentine233 committed 1 year ago

a46217d2

[cpu] enable bfloat16 and refactor for flash attention (#104863)

Valentine233 committed 1 year ago

6d647762

[cpu] implement flash attention backward (#104693)

Valentine233 committed 1 year ago

3fc321f3

[cpu] implement scaled dot product flash attention (#103826)

Valentine233 committed 1 year ago

5516fe12

expand functional map for reduced floating points on CPU (#104584)

Valentine233 committed 1 year ago

02dfacb1

Simplify verbose error guard printing (#107516)

ezyang committed 1 year ago

68b9bf96

Revamp guard debug logging (#107505)

ezyang committed 1 year ago

d6d485fa

fix symint meta val (#107491)

avikchaudhuri committed 1 year ago

db3a199b

disable multi_linear_share_same_input for dynamic shape case (#107123)

zhuhaozhe committed 1 year ago

4d0e7908

[BE]: Update cudnn_frontend submodule to v0.9.2. (#107525)

Skylion007 committed 1 year ago

e21ca06f

do not raise constraint violation on trivial guards (#107470)

avikchaudhuri committed 1 year ago

2c3d2fa2

[BE]: Apply PYI autofixes to various types (#107521)

Skylion007 committed 1 year ago

b1e8e01e

[EASY] Use runtime_var_to_range for guards (#107329)

ezyang committed 1 year ago

24f0b552

[BE]: Update ruff to 0.285 (#107519)

Skylion007 committed 1 year ago

88ab3e43

Add support for GET_YIELD_FROM_ITER, YIELD_FROM, SEND (#106986)

voznesenskym committed 1 year ago

02c2b750

[ATen] Update pre-compiled header (#106915)

peterbell10 committed 1 year ago

4f3284e3

Remove some unnecessary <iostream> includes from headers (#106914)

peterbell10 committed 1 year ago

60936e4c

Raise TypeError for calling moduletype in dynamo (#107393)

wconstab committed 1 year ago

eee2f572

Fakify leaf of FunctionalTensor (#107062)

ydwu4 committed 1 year ago

33497257

[dynamo] fix disable_saved_tensors_hooks - graph break (#106875)

kshitij12345 committed 1 year ago

11602ac5

Trace through Tensor slots (#107159)

lezcano committed 1 year ago

4eac43d0

[functorch] vmap-dynamo: run vmap_impl under fake_mode (#107462)

kshitij12345 committed 1 year ago

8df298bc

Newer Older