Dobromir Popov 471893c9d5 Skip multimodal/MTP checkpoint tensors absent from the text-only causal LM
Qwen3.5/3.6-MoE checkpoints ship vision (model.visual.*) and multi-token-
prediction (mtp.*) weights; the partial shard loader assigned them into the
text-only Qwen3_5MoeForCausalLM and crashed with AttributeError 'mtp'.
Filter selected tensors against the built model's state_dict keys, matching
transformers' _keys_to_ignore_on_load_unexpected behavior.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-07 19:16:19 +02:00
2026-06-28 23:49:11 +03:00
2026-07-07 16:06:05 +02:00
2026-07-07 15:56:38 +03:00
2026-07-02 22:03:37 +03:00
2026-07-07 17:37:38 +03:00
2026-07-03 00:55:06 +02:00
2026-07-04 23:58:07 +03:00
env
2026-07-06 18:47:21 +03:00
db
2026-07-07 19:31:29 +03:00
2026-07-01 14:40:42 +02:00
db
2026-07-07 19:31:29 +03:00
2026-07-07 15:27:33 +03:00
db
2026-07-07 19:31:29 +03:00
2026-06-28 23:49:11 +03:00
Description
No description provided
640 MiB
Languages
Python 94.6%
HTML 5.1%
JavaScript 0.2%