Instructions to use gatilin/north-micro-vision-encoder with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use gatilin/north-micro-vision-encoder with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("feature-extraction", model="gatilin/north-micro-vision-encoder")# pip install -U transformers accelerate # Load model directly from transformers import AutoProcessor, AutoModelForMultimodalLM processor = AutoProcessor.from_pretrained("gatilin/north-micro-vision-encoder") model = AutoModelForMultimodalLM.from_pretrained("gatilin/north-micro-vision-encoder", device_map="auto") - Notebooks
- Google Colab
- Kaggle
File size: 1,077 Bytes
9820884 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 | """North Micro Vision's visual tower and all deepstack mergers."""
import json
from pathlib import Path
from accelerate import init_empty_weights
from safetensors.torch import load_file
from transformers.models.cohere_compass.configuration_cohere_compass import CohereCompassVisionConfig
from transformers.models.cohere_compass.modeling_cohere_compass import CohereCompassVisionModel
from transformers.models.cohere_compass.image_processing_cohere_compass import CohereCompassImageProcessor
def load_model(root):
root=Path(root);c=json.loads((root/'config.json').read_text())['vision_config'];config=CohereCompassVisionConfig(**c);config._attn_implementation='eager'
with init_empty_weights():model=CohereCompassVisionModel(config)
state=load_file(str(root/'model.safetensors'),device='cpu');model.load_state_dict({k.removeprefix('model.visual.'):v for k,v in state.items()},strict=True,assign=True)
return model.eval()
def load_processor(root):
config=json.loads((Path(root)/'preprocessor_config.json').read_text());return CohereCompassImageProcessor.from_dict(config)
|