strisunshinewentingwang/pytorch

Pull Requests Commits

[dont review] assortment of broken AC partitioner

anijain2305 committed 1 year ago

0236b4c8

[dynamo][activation checkpointing] Tagging based compiler activation checkpointing

anijain2305 committed 1 year ago

df67b3a0

[partitioner] Min cut partitioner based on tags on Fx nodes

anijain2305 committed 1 year ago

b81b4892

[benchmarks] Use train mode for accuracy checks for HF models

anijain2305 committed 1 year ago

4586ea0e

[inductor][pattern matcher] Retain meta tags

anijain2305 committed 1 year ago

9d9e59a8

[dynamo][higher order op] Bugfixes to pass graph.lint

anijain2305 committed 1 year ago

a9f7dea8

[benchmark] Flag to switch on activation checkpointing for HF models

anijain2305 committed 1 year ago

2185d04d

[fx] Fix repr when arg is an OpOverload

anijain2305 committed 1 year ago

6fb224cf

[minifier] add missing import

anijain2305 committed 1 year ago

9224c79e

Add padding check for use_nnpack (#92238)

CaoE committed 1 year ago

999bae0f

[profiler] Global function for controlling fwd-bwd connection behavior (#102492)

davidberard98 committed 1 year ago

00992ffa

[vision hash update] update the pinned vision hash (#102495)

pytorchupdatebot committed 1 year ago

0e72ada9

squash xblock for persistent inner reduction (#102444)

Natalia Gimelshein committed 1 year ago

2cc6ae19

Revert "apply constexpr and if constexpr when possible (#102471)"

pytorchmergebot committed 1 year ago

3c2519ab

apply constexpr and if constexpr when possible (#102471)

cyyever committed 1 year ago

461c03a9

[inductor] Replaced refs.op by torch.op in _refs/* (#102176)

vfdev-5 committed 1 year ago

319a1cb4

[inductor] fix issue with ops.lookup_seed (#102485)

jansel committed 1 year ago

fc0fed36

[inductor] Handle floordiv and remainder in IndexPropagation (#102277)

peterbell10 committed 1 year ago

c6d9a0b9

[inductor] Inline ComputedBuffer computation when there are no reads (#102000)

peterbell10 committed 1 year ago

e4e151d6

[inductor] erfinv: CPU/CUDA lowering (#101863)

kshitij12345 committed 1 year ago

b1bc8aec

Revert "Replace int64_t with a size type in python_list.h when applicable (#101922)"

pytorchmergebot committed 1 year ago

0803b918

Improve precision and performance for BFloat16 upsampling (#91169)

CaoE committed 1 year ago

af1d4376

[dynamo] Some torchrec_dlrm related fixes (#101953)

anijain2305 committed 1 year ago

040d2cc9

Enable CuDNN v8 frontend in RL (#102284)

malfet committed 1 year ago

53d1d301

Revert "[FSDP]Add device_mesh to FSDPstate (#102317)"

pytorchmergebot committed 1 year ago

81ac076b

[PT2][Quant] Enable test_qnnpack_quantizer_conv_linear test (#102399)

kimishpatel committed 1 year ago

af70fe9f

[PT2][Quant] Move observer sharing ops to use module partitions (#102398)

kimishpatel committed 1 year ago

0d876f7d

[PT2][Quant] Move add/add relu pattern via module partitioner (#102397)

kimishpatel committed 1 year ago

9fac5afb

[PT2][Quant] Move maxpool_2d quant to use module partitioners (#102396)

kimishpatel committed 1 year ago

3d8f4050

[PT2][Quant] Use module partitions for conv2d and conv2d + relu (#102395)

kimishpatel committed 1 year ago

d997e3aa

Older