Neural Insights: PyTorch support frontend #1209
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
Signed-off-by: aradys-intel <[email protected]>
Signed-off-by: aradys-intel <[email protected]>
Signed-off-by: aradys-intel <[email protected]>
Signed-off-by: aradys-intel <[email protected]>
Signed-off-by: aradys-intel <[email protected]>
bmyrcha
approved these changes
Sep 1, 2023
chensuyue
pushed a commit
to chensuyue/lpot
that referenced
this pull request
Feb 21, 2024
* refine onnxrt adaptor * fix dynamic quant * Strategy refine (intel#1210) * refine(strategy): refine tuning strategy * refine(strategy): add fallback tuning sampler * refine(strategy): integrate to main logic * refine(strategy): add support for dynamic/static * refine(strategy): add default strategy * refine(strategy): update the best cfg * refine(strategy): merge with user cfg * fix(strategy): remove pdb command Co-authored-by: sys-lpot-val <[email protected]> * fix onnxrt adaptor bug * refine(strategy): add optype-wise tuning * refine(strategy): replace default with basic * refine(strategy): fix auto mixed-precision * fix(strategy): fix line too long * fix(strategy): fix line too long * fix(strategy): add init * fix(strategy): add init & fix best update * fix(strategy): fix line too long * fix(strategy): workaround before adaptor ready * modify `_merge_op_wise_cfg` to support config with regex. * fix(strategy): fix parse capability * fix onnxrt ut * fix(strategy): fix calib_iter setting * refine(strategy): add mse * fix(strategy): fix typo * fix(strategy): keep other strategies * fixed the bug that ptq quant always crashes. * fix(strategy): fixed tf recover model ut * fix ort ut * fix qdq auto_quant mode * fix(strategy): refactor exhaustive strategy * fix(strategy): fixed resume from history * fix(strategy): refactor random strategy * fix(strategy): fixed the merge logic * fix(startegy): fixed test_multi_metrics * refine(strategy): update model-wise tuning * fix(strategy): fixed typos * fix(strategy): fixed mixed precision ut * fix(strategy): fix `_tune_cfg_converter` and `_create_calib_dataloader`. * Revert "fix(strategy): fix `_tune_cfg_converter` and `_create_calib_dataloader`." This reverts commit 9804db605eb20ca2faa2559c23ed60c65f57db51. * fix(strategy): fix config for qat * refactor(strategy): refactor the strategy * refactor(strategy): refactor tpe,sigopt, bayesian * refactor(strategy): refactor tpe,sigopt, bayesian (intel#1261) * fix(strategy): fixed typos * fix(strategy): enable dataloader for ptq auto quant * refine(strategy): refine the interfaces of query result * update data strcture for onnx backend * fix(strategy): update the interfaces * refine onnxrt adaptor * fix dynamic quant * Strategy refine (intel#1210) * refine(strategy): refine tuning strategy * refine(strategy): add fallback tuning sampler * refine(strategy): integrate to main logic * refine(strategy): add support for dynamic/static * refine(strategy): add default strategy * refine(strategy): update the best cfg * refine(strategy): merge with user cfg * fix(strategy): remove pdb command Co-authored-by: sys-lpot-val <[email protected]> * fix(strategy): solve merge conflict * refine(strategy): add optype-wise tuning * refine(strategy): replace default with basic * refine(strategy): fix auto mixed-precision * fix(strategy): fix line too long * fix(strategy): fix line too long * fix(strategy): add init * fix(strategy): add init & fix best update * fix(strategy): fix line too long * fix(strategy): workaround before adaptor ready * fix(strategy): fix parse capability * modify `_merge_op_wise_cfg` to support config with regex. * fix(strategy): fix calib_iter setting * fix onnxrt ut * refine(strategy): add mse * fix(strategy): fix typo * fixed the bug that ptq quant always crashes. * fix(strategy): keep other strategies * fix(strategy): fixed tf recover model ut * fix ort ut * fix qdq auto_quant mode * fix(strategy): refactor exhaustive strategy * fix(strategy): fixed resume from history * fix(strategy): refactor random strategy * fix(strategy): fixed the merge logic * fix(startegy): fixed test_multi_metrics * refine(strategy): update model-wise tuning * fix(strategy): fixed typos * fix(strategy): fixed mixed precision ut * fix(strategy): fix `_tune_cfg_converter` and `_create_calib_dataloader`. * Revert "fix(strategy): fix `_tune_cfg_converter` and `_create_calib_dataloader`." This reverts commit 9804db605eb20ca2faa2559c23ed60c65f57db51. * fix(strategy): fix config for qat * refactor(strategy): refactor the strategy * refactor(strategy): refactor tpe,sigopt, bayesian * fix(strategy): fixed typos * refactor(strategy): refactor tpe,sigopt, bayesian (intel#1261) * fix(strategy): enable dataloader for ptq auto quant * refine(strategy): refine the interfaces of query result * update data strcture for onnx backend * fix(strategy): update the interfaces * fix(stratgy): fixed the conflicts after rebase * fix(strategy): fixed the conflicts * fix onnx bug * remove list * fix(strategy): update the merge logic * fix(strategy): WA for TF/ONNX/PT * fix(strategy): fix typo * Adapt TF to the refined strategy * Remove TF WA * update pytorch adaptor * update config structure for mxnet. * fix pytorch adaptor bug * remove code * fix(strategy): remove temporary convertion since adaptor is ready. * exclude const node and placeholder from fp32 list * remove fp32 from capability * fix ut issue * fix(test_adaptor_pytorch): fix assertions to fit new fw_capability format. * fix adaptor bug * remove some unnecessary debug info * fixed tf qat * add LSTM * Strategy/mengni (intel#1279) * remove some unnecessary debug info * fixed tf qat * fix(PT): LSTM won't be quantized in static ptq * fix(PT): quantized LSTM in qat * fix ut issue * fix(util): enable loading both old and new format of tuning cfgs. * fix(strategy): keep the order of options * fix(strategy): fixed the cfg initialization for qat * fix(strategy): remove some debug info and fix qat cfg * fix(strategy): remove unused code * fix(strategy): fixed the cfg initialization for bayesian * fix(startegy): fixed the model-wise sampler * ut(strategy): add more uts * fix(strategy): fixed the fallback order Co-authored-by: Ray <[email protected]> Co-authored-by: sys-lpot-val <[email protected]> Co-authored-by: yiliu30 <[email protected]> Co-authored-by: Zhang Yi5 <[email protected]> Co-authored-by: lvliang-intel <[email protected]> Co-authored-by: Lv, Kaokao <[email protected]>
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.
Type of Change
feature or bug fix or documentation or validation or others
API changed or not
Description
detail description
Expected Behavior & Potential Risk
the expected behavior that triggered by this PR
How has this PR been tested?
how to reproduce the test (including hardware information)
Dependency Change?
any library dependency introduced or removed