Not actually open?

#11
by etoml - opened

Hi,
I've downloaded this model and I'm trying to run it locally, but there's no way to actually run it locally because some of the python files reference closed code. Is there any way to run this model or another one of the molmo models entirely locally or remove the Vision encoder part of the process to get it to run locally?

Ai2 org

Oh that shouldn't be the case; can you point us to which python files?

at the beginning of multimodal_preprocessor.py:

from .mm_data import seqio_tokenizer
from .data_utils import pad_to_bounding_box,
get_3d_subsegments, _append_to_innermost_axis, resize_and_pad,
apply_with_random_selector, get_special_token_ids, make_autoregressive_inputs,
trim_and_pad_dataset, assert_not_truncated
from .prompts import apply_keyword_prompt, STYLE_TO_GENERAL_PROMPT, GENERAL_PROMPTS_V1
import .constants as config

Ai2 org

That file is not used, it was just accidentally committed into the repo. I have removed it.

chrisc36 changed discussion status to closed

Sign up or log in to comment