Image Segmentation
Transformers
Safetensors
English
prima
text-generation
computer-vision
multi-image-reasoning
pixel-grounding
vision-language
custom-code
custom_code
Instructions to use kanguyen-vn/PRIMA with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use kanguyen-vn/PRIMA with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("image-segmentation", model="kanguyen-vn/PRIMA", trust_remote_code=True)# pip install -U transformers accelerate # Load model directly from transformers import AutoModelForCausalLM model = AutoModelForCausalLM.from_pretrained("kanguyen-vn/PRIMA", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
Include clean PRIMA code and evaluator; document metrics, initialization, and MIT contributions
91c1e9e verified Download code/tools/batch.py from kanguyen-vn/PRIMA: direct link, hf CLI and curl.
- Browser
- Download file 736 Bytes
-
https://huggingface.co/kanguyen-vn/PRIMA/resolve/main/code/tools/batch.py
- Command line
-
hf download hf://kanguyen-vn/PRIMA/code/tools/batch.py
-
curl -L -o batch.py https://huggingface.co/kanguyen-vn/PRIMA/resolve/main/code/tools/batch.py
736 Bytes
| """Move M4SEG batches without changing mask or index dtypes.""" | |
| import torch | |
| def move_batch(batch, device): | |
| def move(value): | |
| if torch.is_tensor(value): | |
| return value.to(device=device, non_blocking=True) | |
| if isinstance(value, dict): | |
| return {key: move(item) for key, item in value.items()} | |
| if isinstance(value, list): | |
| return [move(item) for item in value] | |
| if isinstance(value, tuple): | |
| return tuple(move(item) for item in value) | |
| return value | |
| result = move(batch) | |
| for key in ('global_enc_images', 'grounding_enc_images'): | |
| if result[key] is not None: | |
| result[key] = result[key].to(dtype=torch.bfloat16) | |
| return result | |