models : support qwen3.5 series (#19468)

* support qwen3.5 series * remove deepstack for now, and some code clean * code clean * add FULL_ATTENTION_INTERVAL metadata * code clean * reorder v heads for linear attention to avoid expensive interleaved repeat
2026-02-11 00:00:26 +08:00
parent 9a96352729
commit fc0fe40049
17 changed files with 2096 additions and 10 deletions
--- a/convert_hf_to_gguf_update.py
+++ b/convert_hf_to_gguf_update.py
@@ -148,6 +148,7 @@ models = [
    {"name": "youtu",            "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/tencent/Youtu-LLM-2B", },
    {"name": "solar-open",       "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/upstage/Solar-Open-100B", },
    {"name": "exaone-moe",       "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/LGAI-EXAONE/K-EXAONE-236B-A23B", },
+    {"name": "qwen35",           "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/Qwen/Qwen3.5-9B-Instruct", }
 ]

 # some models are known to be broken upstream, so we will skip them as exceptions