Skip to content

[QEff. finetune_experimnetal] cherry picking PR-787, 791,813,795#872

Merged
quic-akuruvil merged 20 commits intoquic:ft_experimental_v1from
smedhe:smedhe_rebase_branch_v1
Mar 25, 2026
Merged

[QEff. finetune_experimnetal] cherry picking PR-787, 791,813,795#872
quic-akuruvil merged 20 commits intoquic:ft_experimental_v1from
smedhe:smedhe_rebase_branch_v1

Conversation

@smedhe
Copy link
Copy Markdown
Contributor

@smedhe smedhe commented Mar 18, 2026

we are only cherry-picking PR-787, 791,813,795, skipping rebasing PR 785, cherry-picking experimental related branches from PR 692,747

@smedhe smedhe force-pushed the smedhe_rebase_branch_v1 branch 3 times, most recently from 8a03d4d to 30a40a3 Compare March 23, 2026 06:22
@smedhe smedhe changed the title [QEff. finetune_experimnetal] cherry picking PR-787, 791 [QEff. finetune_experimnetal] cherry picking PR-787, 791,813,795 Mar 23, 2026
@smedhe smedhe force-pushed the smedhe_rebase_branch_v1 branch 2 times, most recently from f100915 to 42e7b36 Compare March 23, 2026 10:18
quic-rishinr and others added 20 commits March 24, 2026 12:33
Added code for generating PR report.

---------

Signed-off-by: Rishin Raj <rishinr@qti.qualcomm.com>
…on compile coverage (quic#873)

This PR tightens the ONNX export path in modeling_qeff and updates
quickcheck coverage.

  **Changes:**
- Validate cached ONNX before reuse and re-export if the cached file is
invalid.
- Use instance-level transform names for hashing/metadata so export
cache reflects the
    active transform set.
- Only pass onnx_base_dir when ONNX transforms actually need external
tensor data
(FP16ClipTransform / SplitTensorsTransform), avoiding unnecessary tensor
    materialization in the default path.
- Keep the lightweight onnx_transforms guard so external data is only
loaded when a
    base dir is provided.

  **Tests**:

  - Removed the use_dynamo quickcheck case.
- Added .compile(..., use_onnx_subfunctions=True) mocked compile-path
coverage for all
    causal-LM quickcheck models.
  - Full quickcheck validation passed:
      - python -m pytest -q tests/test_model_quickcheck.py -n auto

---------

Signed-off-by: vbaddi <vbaddi@qti.qualcomm.com>
Signed-off-by: Rishin Raj <rishinr@qti.qualcomm.com>
Co-authored-by: vbaddi <vbaddi@qti.qualcomm.com>
Co-authored-by: Rishin Raj <rishinr@qti.qualcomm.com>
Updated the PR report to include charts

---------

Signed-off-by: Rishin Raj <rishinr@qti.qualcomm.com>
Co-authored-by: Hem Agnihotri <hemagnih@qti.qualcomm.com>
…t file (quic#787)

1) Adding text field required by TRL's scripts.
2) Passing config_name in the load_dataset_builder
3) Updated test_dataset accordingly.

Signed-off-by: Swati Allabadi <sallabad@qti.qualcomm.com>
….py) and related code (quic#791)

1) Added FinetuningPipeline (finetune_experiemental.py) which integrates
all the components added for HF-trainer and enable running fine tuning
through it.
2) Added files to handle PEFT and training config.
3) Made changes in the config_manager and callbacks files.
4) Added unit tests for the FinetuningPipeline (test_finetune.py)
5) Updated tests in test_callback and test_config_manager based on above
changes.

Finetuning on openai/gsm8k for 5 epochs on single SOC gave the following
numbers:

{"eval_loss":1.0224987268447876,"eval_runtime":484.8933,"eval_samples_per_second":2.72,"eval_steps_per_second":2.72,"eval_entropy":0.9871161538059735,"eval_num_tokens":6525025.0,"eval_mean_token_accuracy":0.7452040632806826,"epoch":5.0,"num_input_tokens_seen":6525025,"global_step":37365}

{"train_runtime":32856.1501,"train_samples_per_second":1.137,"train_steps_per_second":1.137,"total_flos":3.8132170931712e+16,"train_loss":1.0178058738101043,"epoch":5.0,"num_input_tokens_seen":6525025,"global_step":37365}

Training loss at the start of training :1.5146,

Signed-off-by: Swati Allabadi <sallabad@qti.qualcomm.com>
Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com>

Author:    Swati Allabadi <sallabad@qti.qualcomm.com>
Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com>
Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com>
* Added PP support in HF trainer stack. 
* Updated the documentation for the same. 
* Sample command to test PP : QAIC_VISIBLE_DEVICES=0,1 python -m
QEfficient.cloud.finetune_experimental
QEfficient/finetune/experimental/configs/sample_pp_config.yaml

---------

Signed-off-by: Swati Allabadi <sallabad@qti.qualcomm.com>
Co-authored-by: Swati Allabadi <sallabad@qti.qualcomm.com>
Co-authored-by: Swati Allabadi <sallabad@qti.qualcomm>
To run integrated_test for DDP use following command:
QAIC_VISIBLE_DEVICES=0,1 torchrun --nproc-per-node=2 -m pytest -q
QEfficient/finetune/experimental/tests/test_integrated.py

---------

Signed-off-by: Tanisha Chawada <tchawada@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Modified test_finetune.py
Changed optimizer names

---------

Signed-off-by: Tanisha Chawada <tchawada@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
…set registration (quic#835)

Added example script for registering seq_completion dataset_type and
also updated the hf_finetune.md.

---------

Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com>
1) Fixed Data Parallel issue.
2) Removed sample config for PP as the changes of config manager are now
merged.
3) Updated the PP documentation accordingly.

---------

Signed-off-by: Swati Allabadi <sallabad@qti.qualcomm.com>
Co-authored-by: Swati Allabadi <sallabad@qti.qualcomm.com>
Updated logger.py and test_logger

---------

Signed-off-by: Tanisha Chawada <tchawada@qti.qualcomm.com>
Updated terminal logs

Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Edit and format docs

---------

Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
@smedhe smedhe force-pushed the smedhe_rebase_branch_v1 branch from 42e7b36 to c56b9d1 Compare March 24, 2026 12:37
@quic-akuruvil
Copy link
Copy Markdown
Contributor

looks good

@quic-akuruvil quic-akuruvil merged commit 4be2260 into quic:ft_experimental_v1 Mar 25, 2026
4 checks passed
quic-abhamidi pushed a commit to quic-abhamidi/efficient-transformers that referenced this pull request Mar 25, 2026
…c#872)

we are only cherry-picking PR-787, 791,813,795, skipping rebasing PR
785, cherry-picking experimental related branches from PR 692,747

---------

Signed-off-by: Swati Allabadi <sallabad@qti.qualcomm.com>
Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com>
Signed-off-by: Tanisha Chawada <tchawada@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Co-authored-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
quic-akuruvil pushed a commit to quic-akuruvil/efficient_transformers that referenced this pull request Mar 25, 2026
…c#872)

we are only cherry-picking PR-787, 791,813,795, skipping rebasing PR
785, cherry-picking experimental related branches from PR 692,747

---------

Signed-off-by: Swati Allabadi <sallabad@qti.qualcomm.com>
Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com>
Signed-off-by: Tanisha Chawada <tchawada@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Co-authored-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
quic-akuruvil pushed a commit that referenced this pull request Mar 25, 2026
we are only cherry-picking PR-787, 791,813,795, skipping rebasing PR
785, cherry-picking experimental related branches from PR 692,747

---------

Signed-off-by: Swati Allabadi <sallabad@qti.qualcomm.com>
Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com>
Signed-off-by: Tanisha Chawada <tchawada@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Co-authored-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
quic-akuruvil pushed a commit that referenced this pull request Mar 25, 2026
we are only cherry-picking PR-787, 791,813,795, skipping rebasing PR
785, cherry-picking experimental related branches from PR 692,747

---------

Signed-off-by: Swati Allabadi <sallabad@qti.qualcomm.com>
Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com>
Signed-off-by: Tanisha Chawada <tchawada@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Co-authored-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
quic-akuruvil pushed a commit to quic-akuruvil/efficient_transformers that referenced this pull request Mar 26, 2026
…c#872)

we are only cherry-picking PR-787, 791,813,795, skipping rebasing PR
785, cherry-picking experimental related branches from PR 692,747

---------

Signed-off-by: Swati Allabadi <sallabad@qti.qualcomm.com>
Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com>
Signed-off-by: Tanisha Chawada <tchawada@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Co-authored-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
quic-akuruvil pushed a commit to quic-akuruvil/efficient_transformers that referenced this pull request Mar 27, 2026
…c#872)

we are only cherry-picking PR-787, 791,813,795, skipping rebasing PR
785, cherry-picking experimental related branches from PR 692,747

---------

Signed-off-by: Swati Allabadi <sallabad@qti.qualcomm.com>
Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com>
Signed-off-by: Tanisha Chawada <tchawada@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Co-authored-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
quic-akuruvil pushed a commit that referenced this pull request Mar 27, 2026
we are only cherry-picking PR-787, 791,813,795, skipping rebasing PR
785, cherry-picking experimental related branches from PR 692,747

---------

Signed-off-by: Swati Allabadi <sallabad@qti.qualcomm.com>
Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com>
Signed-off-by: Tanisha Chawada <tchawada@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Co-authored-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
quic-akuruvil pushed a commit that referenced this pull request Mar 30, 2026
we are only cherry-picking PR-787, 791,813,795, skipping rebasing PR
785, cherry-picking experimental related branches from PR 692,747

---------

Signed-off-by: Swati Allabadi <sallabad@qti.qualcomm.com>
Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com>
Signed-off-by: Tanisha Chawada <tchawada@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Co-authored-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
smedhe added a commit to smedhe/QEff_Sharvari that referenced this pull request Apr 6, 2026
…c#872)

we are only cherry-picking PR-787, 791,813,795, skipping rebasing PR
785, cherry-picking experimental related branches from PR 692,747

---------

Signed-off-by: Swati Allabadi <sallabad@qti.qualcomm.com>
Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com>
Signed-off-by: Tanisha Chawada <tchawada@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Co-authored-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
quic-akuruvil pushed a commit to quic-akuruvil/efficient_transformers that referenced this pull request Apr 15, 2026
…c#872)

we are only cherry-picking PR-787, 791,813,795, skipping rebasing PR
785, cherry-picking experimental related branches from PR 692,747

---------

Signed-off-by: Swati Allabadi <sallabad@qti.qualcomm.com>
Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com>
Signed-off-by: Tanisha Chawada <tchawada@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Co-authored-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
quic-akuruvil pushed a commit that referenced this pull request Apr 16, 2026
we are only cherry-picking PR-787, 791,813,795, skipping rebasing PR
785, cherry-picking experimental related branches from PR 692,747

---------

Signed-off-by: Swati Allabadi <sallabad@qti.qualcomm.com>
Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com>
Signed-off-by: Tanisha Chawada <tchawada@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Co-authored-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
quic-akuruvil pushed a commit to quic-akuruvil/efficient_transformers that referenced this pull request Apr 16, 2026
…c#872)

we are only cherry-picking PR-787, 791,813,795, skipping rebasing PR
785, cherry-picking experimental related branches from PR 692,747

---------

Signed-off-by: Swati Allabadi <sallabad@qti.qualcomm.com>
Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com>
Signed-off-by: Tanisha Chawada <tchawada@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Co-authored-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
quic-akuruvil pushed a commit to quic-akuruvil/efficient_transformers that referenced this pull request Apr 22, 2026
…c#872)

we are only cherry-picking PR-787, 791,813,795, skipping rebasing PR
785, cherry-picking experimental related branches from PR 692,747

---------

Signed-off-by: Swati Allabadi <sallabad@qti.qualcomm.com>
Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com>
Signed-off-by: Tanisha Chawada <tchawada@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Co-authored-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
quic-akuruvil pushed a commit that referenced this pull request Apr 27, 2026
we are only cherry-picking PR-787, 791,813,795, skipping rebasing PR
785, cherry-picking experimental related branches from PR 692,747

---------

Signed-off-by: Swati Allabadi <sallabad@qti.qualcomm.com>
Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com>
Signed-off-by: Tanisha Chawada <tchawada@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Co-authored-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

6 participants