strisunshinewentingwang/pytorch

Pull Requests Commits

Dont hold onto references of saved tensors in backward (#88247)

eellison committed 1 year ago

ce961b34

Fix fuse_func method overwrite (#87791) (#88193)

sstsai-adl committed 1 year ago

65de9a2b

[pytorch] Expose EmbeddingPackedParamsBase::unpack to Python (#88362)

poweic committed 1 year ago

43374630

[ONNX] Remove the INT64_MAX magic numbers (#88341)

justinchuby committed 1 year ago

23a6e153

[FSDP] Default to `BACKWARD_PRE` (#88428)

awgu committed 1 year ago

6d7eee04

[profiler] Add an option initialize kineto profiler on start up (#87226) (#88020)

briancoutinho committed 1 year ago

c28022d9

[coreml] delegate multiple outputs (#88345)

mcr229 committed 1 year ago

826b4a9c

[pytorch][vulkan] Add bias storage type to template (#88324)

kimishpatel committed 1 year ago

9533fe90

[PyTorch][Vulkan] Add template based codegen for shader generation (#88323)

kimishpatel committed 1 year ago

893f8e37

Dont clone inputs if using fake tensor (#88208)

eellison committed 1 year ago

60925fcb

[Pytorch][vulkan] Generate shader with parameters (#88322)

kimishpatel committed 1 year ago

192e806c

[minor] use set_default_dtype instead of try and finally (#88295)

kshitij12345 committed 1 year ago

fe3a226d

[dashboard] Replace aot_nvfuser with nvprims_nvfuser (#88437)

anijain2305 committed 1 year ago

f8b73340

[Dynamo][Easy] Fix config.suppress_errors error log (#88402)

yanboliang committed 1 year ago

2bda2baa

Verbose exc printing fix (#88387)

mlazos committed 1 year ago

4d62ee1b

[ONNX] Default runtime type checking to raising errors (#86555)

justinchuby committed 1 year ago

0a274c4b

add parameters check for mkldnn_transpose (#85318)

XiaobingSuper committed 1 year ago

d70bc222

[dynamo] Support compare op for userfunctionvariable (#88372)

anijain2305 committed 1 year ago

c1dd13fb

Disallow module attribute mutation (#88354)

mergennachin committed 1 year ago

2c46d572

Revert "Fix primTorch compute_elementwise_output_strides (#88175)"

pytorchmergebot committed 1 year ago

2b117c84

disable the out variants in test_cumprod test for inductor (#88328)

Krovatkin committed 1 year ago

0f6304ef

add an exclude for test_constructor for inductor (#88143)

Krovatkin committed 1 year ago

529ba076

better error message for out= ops (#88367)

Krovatkin committed 1 year ago

002dad35

reduce the number of autotuning iterations, don't autotune simple til… (#88386)

ngimel committed 1 year ago

b4fcfe77

Add support for neg to NestedTensor (#88131)

cpuhrsch committed 1 year ago

5e6ceebc

[FSDP()][Easy] Make `fully_shard()` only `FULL_SHARD` (#88260)

awgu committed 1 year ago

35be73df

[FSDP()] Have `fully_shard()` abide by `@contract`! (#88235)

awgu committed 1 year ago

fc743ec0

Add a shortcut in Makefile for updating triton (#88318)

desertfire committed 1 year ago

63cd5d7e

Make Python op registration work with torchdeploy/multipy (#87162)

ezyang committed 1 year ago

f884e817

Add a reshape_copy operator. (#88314)

ezyang committed 1 year ago

2f296cfd

Newer Older