LiquidAI

/

LFM2.5-VL-1.6B

like 245

Follow

Liquid AI 2.69k

I’m seeing a weight-loading issue for this model and wanted to check whether the uploaded weights are incomplete or if I’m loading it incorrectly.

Load the model with "transformers==5.0.0" without missing-key warnings/errors, and run text generation normally.
When loading, I get a message about missing weights:

• Missing key: "lm_head.weight" (and possibly related keys)

from transformers import AutoProcessor, AutoModelForImageTextToText, AutoModelForCausalLM

model_id = "LiquidAI/LFM2.5-VL-1.6B"
model = AutoModelForImageTextToText.from_pretrained(
    model_id,
    device_map="auto",
    dtype="bfloat16"
)

Questions

1. Is "lm_head.weight" expected to be missing for this model (tied embeddings), or should it be present in the repo weights?
2. If it should be present, could there be a missing shard / incorrect upload / conversion issue?
3. If this is expected, what is the recommended way to load so it ties embeddings correctly and avoids missing-key warnings?

Thanks for taking a look.