Terry Kong

Hello! Just wanted to add that I've come across this issue as well, but using the CLIPProcessor which uses https://github.com/huggingface/transformers/blob/2f320661f364557c821c285729dab3881e977363/src/transformers/image_transforms.py#L304 Basically I pass in images that always have the dimension...

Thanks for the suggestion @aleksmirosh. Unfortunately, that kwarg only affects the output format: https://github.com/huggingface/transformers/blob/2f320661f364557c821c285729dab3881e977363/src/transformers/image_transforms.py#L322-L323 The input format is not configurable and is inferred: https://github.com/huggingface/transformers/blob/2f320661f364557c821c285729dab3881e977363/src/transformers/image_transforms.py#L340 It would be nice if the...

FWIW, I am seeing this as well using the local launcher: `hydra/launcher=submitit_local`

Hi @elliotgunton So in my use case, I was authoring configuration that used pydantic's `BaseModel`, but tried to add in hera types as fields, and ran into this issue. So...

@JimmyZhang12 I'm not sure I will be able to test this, but I'd like to try to manually test the NeMo-Aligner TRTLLM integration with this PR. It looks like this...

RL would also benefit from this since when we update our submodule of automodel we can query some file to figure out everything automodel supports. Could be JSON or python....

in RL we are seeing that updating from torch 2.7.1 -> torch 2.8 causes a failure like this: ``` File "/opt/ray_venvs/nemo_rl.models.policy.dtensor_policy_worker.DTensorPolicyWorker/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl return self._call_impl(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File...

our old 16N config worked with SP at one point in time, so suspicion is maybe it's adding input/output embedding to parallelize plan

@adil-a what's the current status of this PR?

Hi @jedcheng . This error is new to me, looks like a cert issue. Does this issue help? https://github.com/astral-sh/uv/issues/11728