add eager mode check for NaN and Inf #1015
Closed
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Summary: This diff includes some debug tool improvements In IG_CTR MC proposal debug, we noted that some new snapshots generated NaN in results. We want to figure out the root cause. With this diff, we can run with `--run-accuracy-check` which will run the generate merge + load merge through pybind. But it does not check eager mode run. In this diff, I added this feature. The random inputs are created the same way as we did in load merge. Attach the results P1494263214 ``` CUDA_VISIBLE_DEVICES=5 TORCH_COMPILE_DEBUG=1 TORCHINDUCTOR_MAX_AUTOTUNE=1 buck2 run mode/opt-split-dwarf mode/inplace -c fbcode.platform010_cuda_version=12 -c fbcode.nvcc_arch=h100 caffe2/torch/fb/model_transform/experimental/benchmark:mts_gpu_benchmark -- --model-path=/data/local/models/581303767/85/gpu_lowering/input.predictor.disagg.gpu.merge --lower-backend="AOT_INDUCTOR" --run-accuracy-check --debug_operator_range="1397,1397" --generate_sample_inputs=False --min_acc_module_size=0 --disable-multiple-batch-run 2>&1 | tee aot.log ``` If we want to enable the layer print, add "--dispatch-print" and it will print out each layer's output and check if NaN or INF is contained. Reviewed By: hl475, chenyang78 Differential Revision: D60150435
facebook-github-bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D60150435 |
|
This pull request has been merged in bbb9b84. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary:
This diff includes some debug tool improvements
In IG_CTR MC proposal debug, we noted that some new snapshots generated NaN in results. We want to figure out the root cause.
With this diff, we can run with
--run-accuracy-checkwhich will run the generate merge + load merge through pybind. But it does not check eager mode run. In this diff, I added this feature. The random inputs are created the same way as we did in load merge. Attach the results P1494263214If we want to enable the layer print, add "--dispatch-print" and it will print out each layer's output and check if NaN or INF is contained.
Reviewed By: hl475, chenyang78
Differential Revision: D60150435