- Notifications
You must be signed in to change notification settings - Fork13
Update CoreML recipe and add LLM test#102
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to ourterms of service andprivacy statement. We’ll occasionally send you account related emails.
Already on GitHub?Sign in to your account
base:main
Are you sure you want to change the base?
Uh oh!
There was an error while loading.Please reload this page.
Conversation
HuggingFaceDocBuilderDev commentedJul 11, 2025
The docs for this PR livehere. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
| @@ -5,7 +5,7 @@ | |||
| def install_torch_nightly_deps(): | |||
| """Install torch related dependencies from pinned nightly""" | |||
| EXECUTORCH_NIGHTLY_VERSION = "dev20250625" | |||
| EXECUTORCH_NIGHTLY_VERSION = "dev20250627" | |||
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others.Learn more.
Just curious, what are the new features/fixes from ET?
| @@ -5,7 +5,7 @@ | |||
| def install_torch_nightly_deps(): | |||
| """Install torch related dependencies from pinned nightly""" | |||
| EXECUTORCH_NIGHTLY_VERSION = "dev20250625" | |||
| EXECUTORCH_NIGHTLY_VERSION = "dev20250627" | |||
| TORCHAO_NIGHTLY_VERSION = "dev20250620" | |||
| # Torch nightly is aligned with pinned nightly in https://github.com/pytorch/executorch/blob/main/install_requirements.py#L74 | |||
| TORCH_NIGHTLY_VERSION = "dev20250601" | |||
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others.Learn more.
Should update the pinned PT accordingly. Source of truth is the one pinned on ET trunk. To avoid testing multiple version combinations of different libraries
| ), | ||
| take_over_mutable_buffer=take_over_mutable_buffer, | ||
| ) | ||
| ], | ||
| compile_config=EdgeCompileConfig( | ||
| _check_ir_validity=False, | ||
| _skip_dim_order=False, | ||
| # In ET 0.7, we can set _skip_dim_order=False |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others.Learn more.
Are you saying there is a regression in the pinned ET? 0.7 has been cut already, new nightly should be created for 0.7. Do we still need to toggle this on/off?
| @@ -61,6 +61,7 @@ def load_causal_lm_model(model_name_or_path: str, **kwargs) -> CausalLMExportabl | |||
| use_custom_sdpa = use_custom_sdpa or attn_implementation == "custom_sdpa" | |||
| max_length = kwargs.get("max_length", 2048) | |||
| config = kwargs.get("config") or AutoConfig.from_pretrained(model_name_or_path) | |||
| disable_dynamic_shapes = kwargs.get("disable_dynamic_shapes", False) | |||
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others.Learn more.
nit: Maybe change the name touse_dynamic_shapes and default toTrue as it will be the mainstream for both XNNPACK and CoreML.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others.Learn more.
Looks good to me. Need to fix the CI before merging
This updates CoreML recipes to be export friendly, and adds test for SmolLM.
It also adds option to disable dynamic shapes