feat: 277 feature update modelcard to include a snippet and base model by davidberenstein1957 · Pull Request #282 · PrunaAI/pruna

davidberenstein1957 · 2025-07-27T16:58:57Z

Description

Added multiple updates for better HF/transformers compatibility.

Renames hub methods to work better, HF-aligned naming
- we can now push_to_hub instead of save_to_hub
- we can now from_pretrained instead of from_hub
We updated the ModelCard to copy base model card info from the original loaded model and added nuance on running inference.

RENAMING

from transformers import AutoModelForCausalLM
from pruna.config.smash_config import SmashConfig
from pruna.engine.pruna_model import PrunaModel
from pruna import smash

download_repo_id = "hf-internal-testing/tiny-random-llama4"
upload_repo_id = "PrunaAI/test-save-tiny-random-llama4-smashed"
model = AutoModelForCausalLM.from_pretrained(download_repo_id)
smash_config = SmashConfig(device="cpu")
pruna_model = smash(
    model=model,
    smash_config=smash_config,
)
# we can now push_to_hub instead of save_to_hub
pruna_model.push_to_hub(upload_repo_id)
# we can now use from_pretrained instead of from_hub
pruna_model = PrunaModel.from_pretrained(upload_repo_id)

UPDATED MODEL CARD

Related Issue

Fixes #277
Fixes #278

Type of Change

Bug fix (non-breaking change which fixes an issue)
New feature (non-breaking change which adds functionality)
Breaking change (fix or feature that would cause existing functionality to not work as expected)
This change requires a documentation update

How Has This Been Tested?

Checklist

My code follows the style guidelines of this project
I have performed a self-review of my own code
I have commented my code, particularly in hard-to-understand areas
I have made corresponding changes to the documentation
I have added tests that prove my fix is effective or that my feature works
New and existing unit tests pass locally with my changes

Additional Notes

src/pruna/engine/hf_hub_utils/model_card_template.md

src/pruna/engine/pruna_model.py

sharpenb

Faire renaming of functions. I left some comments. We should make sure that it does not increase maintenant needs.

src/pruna/engine/model_checks.py

src/pruna/engine/pruna_model.py

src/pruna/engine/save.py

tests/engine/test_load.py

tests/engine/test_save.py

johannaSommer

Thanks for addressing my comments David, LGTM!

sharpenb

Thanks for the answers to my comment. I reopen the unaddressed comments and gave some more concrete suggestions to help solving them :)

src/pruna/engine/pruna_model.py

src/pruna/engine/save.py

- Updated `get_diffusers_unet_models` to use dynamic import for better module handling. - Added `push_to_hub` method in `PrunaModel` for streamlined model repository uploads. - Enhanced `from_pretrained` method to support loading from both local paths and Hugging Face Hub, with deprecation warning for `model_path`. - Improved README generation in `save_pruna_model_to_hub` by utilizing `ModelCard` for better metadata handling. - Updated model card template for improved documentation and flexibility in model usage instructions.

- Changed the tags in the model card to include "safetensors" alongside the dynamically generated library tag for improved clarity and categorization.

- Updated test functions to accept a parameter for the model loading method, allowing for more flexible testing of `from_hub` and `from_pretrained`. - Adjusted assertions to utilize the selected method dynamically, improving code maintainability and readability.

- Introduced a `token` parameter in `push_to_hub` and `save_pruna_model_to_hub` methods to facilitate authentication when pushing models to the Hugging Face Hub. - Updated model card generation to include the token for loading existing model cards. - Enhanced the model card template to reference the base repository ID for improved documentation clarity.

- Modified the test case in `test_load.py` to include the `from_pretrained` method for the non-existent model, ensuring consistency in testing model loading methods.

…l_to_hub` - Updated the `save_pruna_model_to_hub` function to ensure `model_name_or_path` is validated before checking repository existence. - Improved the model card template to conditionally display inference instructions based on the presence of `base_repo_id` and `library_name`, enhancing documentation clarity.

- Changed the documentation link in the model card template to point to the updated Pruna documentation site for improved accessibility and accuracy.

- Updated the `PrunaModel` and `save_pruna_model_to_hub` functions to replace the `token` parameter with `hf_token`, enhancing clarity regarding its purpose for Hugging Face authentication. - Adjusted related documentation to reflect the parameter name change, ensuring consistency across the codebase.

- Renamed `save_to_hub` to `push_to_hub` in the `PrunaModel` class and related functions to standardize terminology across the codebase. - Updated documentation and tutorial notebooks to reflect the new method names, ensuring clarity in model saving and loading processes. - Adjusted test cases to utilize the updated method names, enhancing consistency in testing model interactions with the Hugging Face Hub.

…utorials - Consolidated device selection logic in `image_generation.ipynb` and `llms.ipynb` to a single line for clarity. - Streamlined the calculation of percentage differences in both notebooks, enhancing code readability. - Ensured consistent formatting and structure across tutorial notebooks.

- Enhanced the formatting of code snippets in various tutorial notebooks for better readability, including consistent use of line breaks and indentation. - Consolidated multiple lines into single lines where appropriate, particularly in model loading and configuration sections. - Updated markdown cells to maintain clarity and consistency across tutorials, ensuring a smoother learning experience for users.

- Updated comments in the `save_pruna_model_to_hub` function to enhance clarity regarding the retrieval of the pruna library version, ensuring better understanding of the code's purpose and functionality.

- Bumped versions for `accelerate` (1.7.0 to 1.9.0), `aiohttp` (3.12.4 to 3.12.14), `yaspin` (3.1.0 to 2.2.0), and `zipp` (3.22.0 to 3.23.0). - Adjusted dependency URLs and hashes for updated packages. - Refined resolution markers for Python version compatibility.

…unctions - Added a ValueError in the `PrunaModel` class to ensure a valid `pretrained_model_name_or_path` is provided, enhancing robustness. - Updated the documentation in `save_load.rst` for clarity on loading the smashed model, correcting capitalization for consistency.

- Added new package `aenum` (version 3.1.16) and `kernels` (version 0.9.0) with their respective dependencies. - Updated `hqq` version from 0.2.6 to 0.2.7.post1 and adjusted its source URL. - Introduced `opencv-python` (version 4.11.0.86) with dependencies. - Refined resolution markers for Python version compatibility across various packages, enhancing support for different platforms.

src/pruna/engine/save.py

sharpenb

Thanks for incorporating all my comments! Let's go :)

davidberenstein1957 linked an issue Jul 27, 2025 that may be closed by this pull request

[FEATURE] update ModelCard to include a snippet and base model #277

Closed

davidberenstein1957 requested review from johannaSommer and sharpenb July 27, 2025 16:59

johannaSommer requested changes Jul 28, 2025

View reviewed changes

src/pruna/engine/hf_hub_utils/model_card_template.md Outdated Show resolved Hide resolved

src/pruna/engine/pruna_model.py Outdated Show resolved Hide resolved

sharpenb requested changes Aug 1, 2025

View reviewed changes

davidberenstein1957 requested review from johannaSommer and sharpenb August 7, 2025 16:51

johannaSommer approved these changes Aug 7, 2025

View reviewed changes

davidberenstein1957 requested review from SaboniAmine and sharpenb and removed request for sharpenb August 8, 2025 08:39

davidberenstein1957 force-pushed the feat/277-feature-update-modelcard-to-include-a-snippet-and-base-model branch from 552b9b2 to 7e00587 Compare August 8, 2025 09:22

sharpenb requested changes Aug 8, 2025

View reviewed changes

src/pruna/engine/pruna_model.py Outdated Show resolved Hide resolved

src/pruna/engine/pruna_model.py Outdated Show resolved Hide resolved

src/pruna/engine/save.py Outdated Show resolved Hide resolved

src/pruna/engine/save.py Outdated Show resolved Hide resolved

This comment was marked as outdated.

Sign in to view

davidberenstein1957 added 15 commits August 8, 2025 15:29

remove debugging code from save_pruna_model_to_hub function

21d570c

fix: update model card tags in save_pruna_model_to_hub function

20e0a09

- Changed the tags in the model card to include "safetensors" alongside the dynamically generated library tag for improved clarity and categorization.

fix: update test case for non-existent model loading

724764b

- Modified the test case in `test_load.py` to include the `from_pretrained` method for the non-existent model, ensuring consistency in testing model loading methods.

fix: update Pruna documentation link in model card template

c5b6609

- Changed the documentation link in the model card template to point to the updated Pruna documentation site for improved accessibility and accuracy.

refactor: improve comment clarity in save_pruna_model_to_hub function

0e65a66

- Updated comments in the `save_pruna_model_to_hub` function to enhance clarity regarding the retrieval of the pruna library version, ensuring better understanding of the code's purpose and functionality.

davidberenstein1957 force-pushed the feat/277-feature-update-modelcard-to-include-a-snippet-and-base-model branch from a88e4c4 to ece5a2d Compare August 8, 2025 13:33

davidberenstein1957 requested a review from sharpenb August 8, 2025 13:34

cursor bot reviewed Aug 8, 2025

View reviewed changes

src/pruna/engine/save.py Show resolved Hide resolved

sharpenb approved these changes Aug 8, 2025

View reviewed changes

davidberenstein1957 merged commit 2dc281b into main Aug 8, 2025
7 checks passed

Conversation

davidberenstein1957 commented Jul 27, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Related Issue

Type of Change

How Has This Been Tested?

Checklist

Additional Notes

Uh oh!

Uh oh!

Uh oh!

sharpenb left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

johannaSommer left a comment

Choose a reason for hiding this comment

Uh oh!

sharpenb left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

This comment was marked as outdated.

Uh oh!

This comment was marked as outdated.

Uh oh!

Uh oh!

sharpenb left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

davidberenstein1957 commented Jul 27, 2025 •

edited

Loading