strisunshinewentingwang/pytorch

Pull Requests Commits

[ROCm] libtorch nightly now correctly uses rocm runners

jeffdaily committed 2 years ago

d38d950f

[quant][core][improvement] Added support for data_ptr<T> for quantized tensors to return pointer to underlying int type (e.g., int8* instead of qint*)

dzdang committed 2 years ago

b1310b46

Add checks for public and private API

anjali411 committed 2 years ago

d7e23286

Adding Caching of calculated Symbolic Shapes

Gamrix committed 2 years ago

9a7bfaa9

[shard] Add basic math ops to ShardedTensor and add ReplicatedTensor inter-op

wanchaol committed 2 years ago

be354d81

patching clamp for one sided clamp

jjsjann123 committed 2 years ago

0203341b

[FSDP] Avoid using the legacy _sharded_tensor package

fegin committed 2 years ago

c2d4e201

disabling reshape

jjsjann123 committed 2 years ago

f7e7af80

updated cudnn frontend to v0.6.1

dzdang committed 2 years ago

25aa251f

[FSDP] summon full params staticmethod

rohan-varma committed 2 years ago

143f7cca

Add note about testing inconsistency

janeyx99 committed 2 years ago

8d3e3ebc

Update docker-builds to add CUDA 11.6

atalman committed 2 years ago

6d9210be

Refractor flatbuffer jit code

pavithranrao committed 2 years ago

6402e624

[ONNX] Make Non-Float Op Exportation Compatible to Avoid Invalid ONNX Models

ganler committed 2 years ago

c1f0e6e7

[PyTorch Edge] Add Quantized Softmax Op (Naive Implementation) (Re-land)

salilsdesai committed 2 years ago

ca0ef523

ci: Add lts for s3 index updating workflow

seemethere committed 2 years ago

2debdfdc

Moving Remove Tensor Type Specializations to after custom passes

Gamrix committed 2 years ago

f281d83d

Docs: Detail 3D tensor shape for transformer masks

jaketae committed 2 years ago

04224e18

Docs: Add missing zero-ing step in Rprop algorithm

jaketae committed 2 years ago

3b18bc36

Warn once about sparse CSR tensors beta support

pearu committed 2 years ago

484534f3

Fix some typos.

Yulv-git committed 2 years ago

ac2d2e3a

[FSDP] Fix `_get_param_to_unflat_param_names()` for shared params

awgu committed 2 years ago

8e365fab

ci: add repo-token for stale workflow

seemethere committed 2 years ago

20bf0d45

code-generate non-aliasing {view}_copy kernels (#73442)

bdhirsh committed 2 years ago

23b84143

Add retry to Run Actionlint shellcheck step

janeyx99 committed 2 years ago

dfcb7035

[ONNX] Initial test suite for Torch IR to ONNX

BowenBao committed 2 years ago

1793ab0b

[DataPipe] Change interface generation process to revert back to original working process

NivekT committed 2 years ago

26d22b7f

Add fastpath for stack and cat JVP computation

anjali411 committed 2 years ago

91d13409

Add ZT fastpath for sub

anjali411 committed 2 years ago

1aaf3ee6

Support masked prod on CSR tensors

pearu committed 2 years ago

e21e4b70

Newer Older