shunting314

[inductor] do benchmark in sub processes for max autotuning

20

Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * __->__ #96410 This PR implements the support to benchmark max-autotune choices in subprocesses. This way crash like https://github.com/openai/triton/issues/1298 will only abort the autotuning...

This issue tracks the PT2 dashboard inference pass rate (excluding dynamic shape and aot-inductor). The pass rate is already pretty high so the issue just list the failed models and...

This issue tracks the PT2 dashboard training pass rate (excluding dynamic shape). The pass rate is already pretty high so the issue just list the failed models and track the...

A group of models fail accuracy test in inference with freezing is enabled

1

There are a batch of models that fail the accuracy test in inference when freezing is enabled. - TB: mnasnet1_0 , moco, shufflenet_v2_x1_0 - TIMM: adv_inception_v3, botnet26t_256, gluon_inception_v3, selecsls42b, swsl_resnext101_32x16d...

tts_angular training fail assertion in cudagraph_trees

1

### 🐛 Describe the bug It works if cudagraphs is disabled: ``` time python benchmarks/dynamo/torchbench.py --backend inductor --amp --performance --only tts_angular --training --disable-cudagraphs ``` But it fails when cudagraphs is...

[Inductor] log fusion failure due to index mismatch

2

Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * __->__ #124986 The scheduler searches for fusion opportunities by looking for common memory access. Two memory access are considered common not only when...

[inductor][easy] add buffer layout to SchedulerNode.debug_str

1

Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * #124986 * #125091 * __->__ #125090 cc @ezyang @msaroufim @bdhirsh @anijain2305 @chauhang @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @peterbell10...

[inductor] add triton code to SchedulerNode.debug_str

1

Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * #124986 * __->__ #125091 * #125090 Here is an example print: https://gist.github.com/shunting314/75c161368a833a535bd0d240b8099d7e cc @ezyang @msaroufim @bdhirsh @anijain2305 @chauhang @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen...

type conversion before tl.dot fails compilation

21

Context: we are trying to upgrade trition pin in pytorch and start encountering test failure for `python test/inductor/test_pattern_matcher.py -k test_mixed_mm` Doing a type conversion for an input tensor to tl.dot...

Revert "Skip test_memory_format_nn_BatchNorm2d in inductor (#125970)"

3

Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * __->__ #126594 * #126593 * #126586 This reverts commit 0a9c6e92f8d1a35f33042c8dab39f23b7f39d6e7. enable the test since it's fixed.