[Torch FX] Post Quantize Weights Compression #2984
Merged
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
…ion_transformation
…ion_transformation
…ion_transformation
github-actions
bot
added
NNCF PT
…/github.com/anzr299/nncf into fx_post_quantize_compression_transformation
…ion_transformation
…/github.com/anzr299/nncf into fx_post_quantize_compression_transformation
2. change variable names
…ion_transformation
…/github.com/anzr299/nncf into fx_post_quantize_compression_transformation
Co-authored-by: Daniil Lyakhov <[email protected]>
alexsu52
reviewed
Oct 18, 2024
daniil-lyakhov
approved these changes
Oct 18, 2024
alexsu52
pushed a commit
that referenced
this pull request
Oct 30, 2024
### Changes * ~~Constant folding is applied to all TorchFX models before the quantization~~ * Some torchvision models (swin_v2_s, vit_16_b) are exported by `torch.export.export` before ov conversation * Moc transformations are applied to openvino compressed models after the compression After the #2984 * Fixed `_compress_qdq_constant_transformation` for per tensor case ### Reason for changes * To align TorchFX/OV quantized models ### Related tickets #2766 ### Tests post_training_quantization/504/ is finished successfully
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Outdated suggestions cannot be applied.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.
Changes
Transformation for removing fake quantize nodes and saving all weights to disk in int8 format after quantization. It works as follows:
Reason for changes
To compress the model after quantization
Tests
Add
test_post_quantization_compression()intests/torch/fx/test_model_transformer.pywhich checks the data type of all weights in the model after applying quantization and also checks the value after the decompression step (element-wise multiplication operation).Tickets
#2766