Views
No views yet
OpenGVLab/InternVL3_5-8B-HF for zeromodels. One implementation runs unmodified on TensorFlow / Torch / JAX. This is a InternVL3.5 checkpoint, served as image + text -> text via InternVLProcessor; weights are stored in bfloat16.1import os
2os.environ["KERAS_BACKEND"] = "torch" # or "jax" / "tensorflow"
3
4from PIL import Image
5from zeromodels.models.internvl import InternVLConditionalGenerate, InternVLProcessor
6
7model = InternVLConditionalGenerate.from_weights("zeromodels/internvl3.5-8b")
8processor = InternVLProcessor.from_weights("zeromodels/internvl3.5-8b")
9
10inputs = processor(conversation=[
11 {"role": "user", "content": [
12 {"type": "image", "image": Image.open("photo.jpg")},
13 {"type": "text", "text": "Describe this image in one sentence."},
14 ]}
15])
16outputs = model.generate(**inputs, max_new_tokens=64)
17print(processor.decode(outputs[0]))from_weights("zeromodels/<variant>"). Browse them all in the InternVL collection.apache-2.0 (per the upstream model card).