Update modeling_molmo.py
#41
by
ctranslate2-4you
- opened
required to comply with transformers 4.49+, which removed _extract_past_from_model_output
Thanks, it worked!!
This is how I loaded the patch, if anyone else is having the same issue:
from transformers import AutoModelForCausalLM, AutoProcessor, GenerationConfig
from PIL import Image
import requests
Define the commit hash
commit_hash = "03214def8ce840adaf1f4f569b20e45245af8fab"
Load processor with specific commit
processor = AutoProcessor.from_pretrained(
'allenai/Molmo-7B-D-0924',
revision=commit_hash,
trust_remote_code=True,
torch_dtype='auto',
device_map='auto'
)
Load model with specific commit
model = AutoModelForCausalLM.from_pretrained(
'allenai/Molmo-7B-D-0924',
revision=commit_hash,
trust_remote_code=True,
torch_dtype='auto',
device_map='auto'
)