You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We fine tune models and store them on HuggingFace hub. The model names are built from datasetid__experiment name such as:
Company/client1_fielddata_[date]__favoritemodel_finetuned_300epochs
We have noticed random-appearing crashes when trying to create the endpoint, and narrowed it down to the model name being too long. The issue is twofold:
The error message is unclear and does not lead you to recreate a model under a different name.
The name restriction of 70 (60?) characters together with namespace is restricting. Having long descriptive model names is helpful during the experimentation phase.
Traceback (most recent call last):
File "/Applications/PyCharm.app/Contents/plugins/python/helpers-pro/pydevd_asyncio/pydevd_asyncio_utils.py", line 117, in _exec_async_code
result = func()
^^^^^^
File "<input>", line 1, in<module>
File "/Users/bdzyubak/Library/Caches/pypoetry/virtualenvs/vlm-finetuning-X38kV7cE-py3.11/lib/python3.11/site-packages/huggingface_hub/hf_api.py", line 7264, in create_inference_endpoint
hf_raise_for_status(response)
File "/Users/bdzyubak/Library/Caches/pypoetry/virtualenvs/vlm-finetuning-X38kV7cE-py3.11/lib/python3.11/site-packages/huggingface_hub/utils/_errors.py", line 358, in hf_raise_for_status
raise BadRequestError(message, response=response) from e
huggingface_hub.utils._errors.BadRequestError: (Request ID: m4c7P7)
Bad request:
Bad Request: Invalid name
Describe the bug
We fine tune models and store them on HuggingFace hub. The model names are built from datasetid__experiment name such as:
Company/client1_fielddata_[date]__favoritemodel_finetuned_300epochs
We have noticed random-appearing crashes when trying to create the endpoint, and narrowed it down to the model name being too long. The issue is twofold:
Reproduction
endpoint = create_inference_endpoint(
name='client1_fielddata_favoritemodel_finetuned',
repository='Company/client1_fielddata_[date]__favoritemodel_finetuned_300epochs,
framework="pytorch",
task="custom",
accelerator="gpu",
vendor=vendor,
region=region,
type="protected",
instance_size="x1",
instance_type="nvidia-l4",
custom_image={
"health_route": "/health",
"env": env_vars,
"url": VLLM_HF_IMAGE_URL,
},
token=os.getenv("HF_TOKEN"),
namespace="Company"
)
Logs
System info
The text was updated successfully, but these errors were encountered: