strisunshinewentingwang/pytorch

Pull Requests Commits

Rearrange the fields in at::OperandInfo to reduce padding. (#98037)

nadavrot committed 1 year ago

09117751

[Dynamo] Code refactor: move context managers out of misc.py (#97958)

yanboliang committed 1 year ago

9be9592f

use libcusolver_lapack_static.a for CUDA>=12 (#98072)

ptrblck committed 1 year ago

3c7b2b73

Fix `aten::squeeze.dims` shape function (#98078)

vivekkhandelwal1 committed 1 year ago

5810f5ad

[pruning] Add gelu to list of supported activation functions (#95618)

jcaip committed 1 year ago

d158545b

do not need to check if element in dict input is Tensor. (#97866)

sywangyi committed 1 year ago

8564ed24

[PyTorch] Accept string_view in Pickler::pushGlobal (#96402)

swolchok committed 1 year ago

794f6e50

[FSDP] Do not `_unshard` if already prefetched (#97981)

awgu committed 1 year ago

fb7b3984

[FSDP][Easy] Minor cleanups to `_runtime_utils.py` (#97980)

awgu committed 1 year ago

195b92ab

[FSDP][Docs] Tidy up FSDP ctor docs (#97979)

awgu committed 1 year ago

adee9423

Revert "[Reland] fix some MKL detection issues of CMake (#94924)"

pytorchmergebot committed 1 year ago

3226ad21

Retry at test file level (#97506)

clee2000 committed 1 year ago

0d73cfb3

Don't use subclass when tracing and call wait_tensor immediately. (#98001)

Rodrigo Kumpera committed 1 year ago

3b188c58

[PyTorch] Add Vulkan support and tests for at::upsample_bilinear2d (#98022)

liuk22 committed 1 year ago

f2127bbf

Fix typos under c10 directory (#98079)

kiszk committed 1 year ago

64b8d20a

[dynamo 3.11] make create_instruction kwarg mandatory (#98032)

williamwen42 committed 1 year ago

762a2079

[dynamo 3.11] implement 3.11 linetable (#96509)

williamwen42 committed 1 year ago

089134bf

[dynamo 3.11] small bug fixes (#96508)

williamwen42 committed 1 year ago

14ef91ce

[dynamo 3.11] support prefix instructions MAKE_CELL, COPY_FREE_VARS, RETURN_GENERATOR, RESUME (#96506)

williamwen42 committed 1 year ago

cb4bc8e0

[dynamo 3.11] fix jump if (not) none (#96505)

williamwen42 committed 1 year ago

05641b81

Print test times for pytest in verbose mode (#98028)

clee2000 committed 1 year ago

27e06e1a

optimize the AMP func name in custom_device_mod (#98052)

heidongxianhua committed 1 year ago

d03799f9

[CI] Bump up torchbench version to fix dynamo graph breaks in transformers (#98003)

desertfire committed 1 year ago

c699ac17

Revert "[dtensor] switch mean to use reduction linear (#97996)"

pytorchmergebot committed 1 year ago

9e3b3477

[dynamo] Add guards for deterministic algos (#96695)

colesbury committed 1 year ago

87f5e929

aot_autograd: avoid using intermediate_base logic unnecessarily (#97786)

bdhirsh committed 1 year ago

864ab936

fix load_sharded_optimizer_state_dict error on multi node (#98063)

wz337 committed 1 year ago

4e26ad78

add StorageImpl::mutable_unsafe_data (#97648)

mikey dagitses committed 1 year ago

cb8c0be5

Revert "Move functional collectives to the right namespace (#97793)"

pytorchmergebot committed 1 year ago

f4f1a5b5

Fix device handling in `nn.utils.rnn.unpad_sequence` (#98042)

anton-bushuiev committed 1 year ago

fa1a8b9f

Newer Older