aws-samples / aws-samples/sagemaker-genai-hosting-examples

Not able to deploy Gemma3 model using TGI container

Open
#77 1 comment 0 reactions 0 assignees View on GitHub
Dominant language
Jupyter Notebook
Stars
144
Forks
103
Avg merge
1d 15h
Merged PRs (30d)
12

Description

I am trying to deploy Gemma 3 4B model using TGI container in the below notebook and running into issue where end point is not created.
https://github.com/aws-samples/sagemaker-genai-hosting-examples/blob/main/Gemma3/Gemma-3-4B-instruct-TGI.ipynb

Step to reproduce:
1. Run Gemma-3-4B-instruct-TGI.ipynb notebook
2. Follow steps out lined in the notebook.
3. Confirmed I am using HF token and I have access to this Gated model.

Error
```
│ ╰──────────────────────────────────────────────────────────────────────────╯ │

2025-05-08T15:46:38.762Z

│ │

2025-05-08T15:46:38.762Z

│ /usr/src/server/text_generation_server/utils/hub.py:111 in weight_hub_files │

2025-05-08T15:46:38.762Z

│ │

2025-05-08T15:46:38.762Z

│ 108 │ │ filenames = _weight_hub_files_from_model_info(info, extension) │

2025-05-08T15:46:38.762Z

│ 109 │ │

2025-05-08T15:46:38.762Z

│ 110 │ if not filenames: │

2025-05-08T15:46:38.762Z

│ ❱ 111 │ │ raise EntryNotFoundError( │

2025-05-08T15:46:38.762Z

│ 112 │ │ │ f"No {extension} weights found for model {model_id} and re │

2025-05-08T15:46:38.762Z

│ 113 │ │ │ None, │

2025-05-08T15:46:38.762Z

│ 114 │ │ ) │

2025-05-08T15:46:38.762Z

│ │

2025-05-08T15:46:38.762Z

│ ╭───────────────────────────────── locals ─────────────────────────────────╮ │

2025-05-08T15:46:38.762Z

│ │ api = │ │

2025-05-08T15:46:38.762Z

│ │ extension = '.bin' │ │

2025-05-08T15:46:38.762Z

│ │ filenames = [] │ │

2025-05-08T15:46:38.762Z

│ │ info = ModelInfo( │ │

2025-05-08T15:46:38.762Z

│ │ │ id='google/gemma-3-4b-it', │ │

2025-05-08T15:46:38.762Z

│ │ │ author='google', │ │

2025-05-08T15:46:38.762Z

│ │ │ sha='093f9f388b31de276ce2de164bdc2081324b9767', │ │

2025-05-08T15:46:38.762Z

│ │ │ created_at=datetime.datetime(2025, 2, 20, 21, 20, 7, │ │

2025-05-08T15:46:38.762Z

│ │ tzinfo=datetime.timezone.utc), │ │

2025-05-08T15:46:38.762Z

│ │ │ last_modified=datetime.datetime(2025, 3, 21, 20, 20, 53, │ │

2025-05-08T15:46:38.762Z

│ │ tzinfo=datetime.timezone.utc), │ │

2025-05-08T15:46:38.762Z

│ │ │ private=False, │ │

2025-05-08T15:46:38.762Z

│ │ │ disabled=False, │ │

2025-05-08T15:46:38.762Z

│ │ │ downloads=611914, │ │

2025-05-08T15:46:38.762Z

│ │ │ downloads_all_time=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ gated='manual', │ │

2025-05-08T15:46:38.762Z

│ │ │ gguf=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ inference=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ inference_provider_mapping=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ likes=501, │ │

2025-05-08T15:46:38.762Z

│ │ │ library_name='transformers', │ │

2025-05-08T15:46:38.762Z

│ │ │ tags=[ │ │

2025-05-08T15:46:38.762Z

│ │ │ │ 'transformers', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ 'safetensors', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ 'gemma3', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ 'image-text-to-text', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ 'conversational', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ 'arxiv:1905.07830', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ 'arxiv:1905.10044', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ 'arxiv:1911.11641', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ 'arxiv:1904.09728', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ 'arxiv:1705.03551', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ ... +29 │ │

2025-05-08T15:46:38.762Z

│ │ │ ], │ │

2025-05-08T15:46:38.762Z

│ │ │ pipeline_tag='image-text-to-text', │ │

2025-05-08T15:46:38.762Z

│ │ │ mask_token=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ card_data={'base_model': 'google/gemma-3-4b-pt', │ │

2025-05-08T15:46:38.762Z

│ │ 'datasets': None, 'eval_results': None, 'language': None, │ │

2025-05-08T15:46:38.762Z

│ │ 'library_name': 'transformers', 'license': 'gemma', │ │

2025-05-08T15:46:38.762Z

│ │ 'license_name': None, 'license_link': None, 'metrics': None, │ │

2025-05-08T15:46:38.762Z

│ │ 'model_name': None, 'pipeline_tag': 'image-text-to-text', │ │

2025-05-08T15:46:38.762Z

│ │ 'tags': None, 'extra_gated_heading': 'Access Gemma on │ │

2025-05-08T15:46:38.762Z

│ │ Hugging Face', 'extra_gated_prompt': 'To access Gemma on │ │

2025-05-08T15:46:38.762Z

│ │ Hugging Face, you’re required to review and agree to │ │

2025-05-08T15:46:38.762Z

│ │ Google’s usage license. To do this, please ensure you’re │ │

2025-05-08T15:46:38.762Z

│ │ logged in to Hugging Face and click below. Requests are │ │

2025-05-08T15:46:38.762Z

│ │ processed immediately.', 'extra_gated_button_content': │ │

2025-05-08T15:46:38.762Z

│ │ 'Acknowledge license'}, │ │

2025-05-08T15:46:38.762Z

│ │ │ widget_data=[ │ │

2025-05-08T15:46:38.762Z

│ │ │ │ {'text': 'Hi, what can you help me with?'}, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ {'text': 'What is 84 * 3 / 2?'}, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ { │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ 'text': 'Tell me an interesting fact about the │ │

2025-05-08T15:46:38.762Z

│ │ universe!' │ │

2025-05-08T15:46:38.762Z

│ │ │ │ }, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ { │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ 'text': 'Explain quantum computing in simple │ │

2025-05-08T15:46:38.762Z

│ │ terms.' │ │

2025-05-08T15:46:38.762Z

│ │ │ │ } │ │

2025-05-08T15:46:38.762Z

│ │ │ ], │ │

2025-05-08T15:46:38.762Z

│ │ │ model_index=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ config={ │ │

2025-05-08T15:46:38.762Z

│ │ │ │ 'architectures': [ │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ 'Gemma3ForConditionalGeneration' │ │

2025-05-08T15:46:38.762Z

│ │ │ │ ], │ │

2025-05-08T15:46:38.762Z

│ │ │ │ 'model_type': 'gemma3', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ 'processor_config': { │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ 'chat_template': "{{ bos_token }}\n{%- if │ │

2025-05-08T15:46:38.762Z

│ │ messages[0]['role'] == 'system' -%}\n {%- if │ │

2025-05-08T15:46:38.762Z

│ │ messages[0"+1452 │ │

2025-05-08T15:46:38.762Z

│ │ │ │ }, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ 'tokenizer_config': { │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ 'bos_token': '', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ 'chat_template': "{{ bos_token }}\n{%- if │ │

2025-05-08T15:46:38.762Z

│ │ messages[0]['role'] == 'system' -%}\n {%- if │ │

2025-05-08T15:46:38.762Z

│ │ messages[0"+1452, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ 'eos_token': '', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ 'pad_token': '', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ 'unk_token': '', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ 'use_default_system_prompt': False │ │

2025-05-08T15:46:38.762Z

│ │ │ │ } │ │

2025-05-08T15:46:38.762Z

│ │ │ }, │ │

2025-05-08T15:46:38.762Z

│ │ │ transformers_info=TransformersInfo( │ │

2025-05-08T15:46:38.762Z

│ │ │ │ auto_model='AutoModelForImageTextToText', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ custom_class=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ pipeline_tag='image-text-to-text', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ processor='AutoProcessor' │ │

2025-05-08T15:46:38.762Z

│ │ │ ), │ │

2025-05-08T15:46:38.762Z

│ │ │ trending_score=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ siblings=[ │ │

2025-05-08T15:46:38.762Z

│ │ │ │ RepoSibling( │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ rfilename='.gitattributes', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ size=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ blob_id=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ lfs=None │ │

2025-05-08T15:46:38.762Z

│ │ │ │ ), │ │

2025-05-08T15:46:38.762Z

│ │ │ │ RepoSibling( │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ rfilename='README.md', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ size=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ blob_id=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ lfs=None │ │

2025-05-08T15:46:38.762Z

│ │ │ │ ), │ │

2025-05-08T15:46:38.762Z

│ │ │ │ RepoSibling( │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ rfilename='added_tokens.json', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ size=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ blob_id=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ lfs=None │ │

2025-05-08T15:46:38.762Z

│ │ │ │ ), │ │

2025-05-08T15:46:38.762Z

│ │ │ │ RepoSibling( │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ rfilename='chat_template.json', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ size=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ blob_id=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ lfs=None │ │

2025-05-08T15:46:38.762Z

│ │ │ │ ), │ │

2025-05-08T15:46:38.762Z

│ │ │ │ RepoSibling( │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ rfilename='config.json', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ size=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ blob_id=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ lfs=None │ │

2025-05-08T15:46:38.762Z

│ │ │ │ ), │ │

2025-05-08T15:46:38.762Z

│ │ │ │ RepoSibling( │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ rfilename='generation_config.json', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ size=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ blob_id=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ lfs=None │ │

2025-05-08T15:46:38.762Z

│ │ │ │ ), │ │

2025-05-08T15:46:38.762Z

│ │ │ │ RepoSibling( │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ rfilename='model-00001-of-00002.safetensors', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ size=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ blob_id=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ lfs=None │ │

2025-05-08T15:46:38.762Z

│ │ │ │ ), │ │

2025-05-08T15:46:38.762Z

│ │ │ │ RepoSibling( │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ rfilename='model-00002-of-00002.safetensors', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ size=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ blob_id=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ lfs=None │ │

2025-05-08T15:46:38.762Z

│ │ │ │ ), │ │

2025-05-08T15:46:38.762Z

│ │ │ │ RepoSibling( │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ rfilename='model.safetensors.index.json', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ size=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ blob_id=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ lfs=None │ │

2025-05-08T15:46:38.762Z

│ │ │ │ ), │ │

2025-05-08T15:46:38.762Z

│ │ │ │ RepoSibling( │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ rfilename='preprocessor_config.json', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ size=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ blob_id=None, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ │ lfs=None │ │

2025-05-08T15:46:38.762Z

│ │ │ │ ), │ │

2025-05-08T15:46:38.762Z

│ │ │ │ ... +5 │ │

2025-05-08T15:46:38.762Z

│ │ │ ], │ │

2025-05-08T15:46:38.762Z

│ │ │ spaces=[ │ │

2025-05-08T15:46:38.762Z

│ │ │ │ 'hadadrjt/ai', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ 'TIGER-Lab/MEGA-Bench', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ 'whitecircle-ai/circle-guard-bench', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ 'VIDraft/Gemma-3-R1984-4B', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ 'merterbak/gemma-3', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ 'CalvinU/writing-prototypes', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ 'Tomoniai/gemma-3-chat', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ 'hanzla/PlaygroundGemma3', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ 'AC2513/gemma-translate', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ 'stiv14/pdf-multilanguage-qa-role', │ │

2025-05-08T15:46:38.762Z

│ │ │ │ ... +46 │ │

2025-05-08T15:46:38.762Z

│ │ │ ], │ │

2025-05-08T15:46:38.762Z

│ │ │ safetensors=SafeTensorsInfo( │ │

2025-05-08T15:46:38.762Z

│ │ │ │ parameters={'BF16': 4300079472}, │ │

2025-05-08T15:46:38.762Z

│ │ │ │ total=4300079472 │ │

2025-05-08T15:46:38.762Z

│ │ │ ), │ │

2025-05-08T15:46:38.762Z

│ │ │ security_repo_status=None │ │

2025-05-08T15:46:38.762Z

│ │ ) │ │

2025-05-08T15:46:38.762Z

│ │ model_id = 'google/gemma-3-4b-it' │ │

2025-05-08T15:46:38.762Z

│ │ revision = None │ │

2025-05-08T15:46:38.762Z

│ ╰──────────────────────────────────────────────────────────────────────────╯ │

2025-05-08T15:46:38.762Z

╰──────────────────────────────────────────────────────────────────────────────╯

2025-05-08T15:46:38.762Z

EntryNotFoundError: No .bin weights found for model google/gemma-3-4b-it and

2025-05-08T15:46:39.262Z

revision None.
```

Contributor guide

Open the contributing guide

Research direction

Start with Gemma-3-4B-instruct-TGI.ipynb and reproduce the deployment using the stated Hugging Face token and model access. Inspect the TGI container error at /usr/src/server/text_generation_server/utils/hub.py:111, where no .bin weights are found despite safetensors files being listed. Done means the notebook creates the SageMaker endpoint successfully.

Written by the indexing model from the issue text.

Assessment

Tech stack
aws, jupyter-notebook
Domain
cloud, machine-learning
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
30/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.