# /// script | |
# requires-python = ">=3.12" | |
# dependencies = [ | |
# "transformers", | |
# "torch", | |
# ] | |
# /// | |
# Use a pipeline as a high-level helper | |
from transformers import pipeline | |
pipe = pipeline("image-text-to-text", model="bharatgenai/patram-7b-instruct", trust_remote_code=True) | |
messages = [ | |
{ | |
"role": "user", | |
"content": [ | |
{"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, | |
{"type": "text", "text": "What animal is on the candy?"} | |
] | |
}, | |
] | |
pipe(text=messages) | |
# Load model directly | |
from transformers import AutoModelForCausalLM | |
model = AutoModelForCausalLM.from_pretrained("bharatgenai/patram-7b-instruct", trust_remote_code=True) |