Zero-Shot Classification
Transformers
Safetensors
qwen3_5
feature-extraction
decision-model
classification
system-one
multimodal
vision
video
custom_code
Instructions to use vllm-sr/d3-mini with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use vllm-sr/d3-mini with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("zero-shot-classification", model="vllm-sr/d3-mini", trust_remote_code=True)# pip install -U transformers accelerate # Load model directly from transformers import AutoProcessor, AutoModel processor = AutoProcessor.from_pretrained("vllm-sr/d3-mini", trust_remote_code=True) model = AutoModel.from_pretrained("vllm-sr/d3-mini", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
The package is broken
#4
by mazurkin - opened
You just updated README.md without updating the manifest
File ~/.cache/huggingface/modules/transformers_modules/vllm_hyphen_sr/d3_hyphen_mini/c5aa751c277d26905c28747706023709ea56324b/d3_runtime.py:883, in verify_package(root, mode)
881 problems.append(f"{name}: SHA-256 differs from {MANIFEST}")
882 if problems:
--> 883 raise ValueError("package verification failed: " + "; ".join(problems[:8]))
884 return manifest
ValueError: package verification failed: README.md: 17203 bytes, manifest says 16441
if you read this use revision='6d8971f464d6f9d4c5af7d9a8d75f3aa03be4f59' in transformers.AutoModel.from_pretrained() call
Thanks for the report and for sharing the workaround, @mazurkin .
Loading main failed because the runtime's load-time check also covered the model card (README.md), so a card-only update broke verification.
This is fixed in v3.1.3, which is now main: loading checks only the files the runtime reads, so card updates can no longer stop the model from loading. Weights, model_sha256 and answers are unchanged, and pinning a revision is no longer needed.