mozempk/ggml-org-llama-cpp-mirror
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-10-13 00:10:37 +02:00
Code Issues Packages Projects Releases Wiki Activity

convert : support Qwen3.5 embedding models (#27920)

Browse Source 
* Update convert_hf_to_gguf: support Qwen3.5 embedding models

* Behavior-preserving refactor for conventions.

* conversion: simplify pooling comment
This commit is contained in:
Sam Malayek
2026-10-08 08:40:01 +03:00
committed by GitHub
parent 9b4ed0ca57
commit 75118a3a59
4 changed files with 34 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
conversion/base.py
+10
View File
@@ -2336,6 +2336,16 @@ class TextModel(ModelBase):
else:
raise NotImplementedError("Only MEAN, CLS, and LAST pooling types supported")
self.gguf_writer.add_pooling_type(pooling_type)
else:
embedding_config_path = self.dir_model / "embedding_config.json"
if embedding_config_path.is_file():
with open(embedding_config_path, encoding="utf-8") as f:
embedding_config = json.load(f)
pooling = embedding_config.get("pooling")
if pooling == "last_token":
self.gguf_writer.add_pooling_type(gguf.PoolingType.LAST)
elif pooling is not None:
raise NotImplementedError(f"unsupported embedding_config.json pooling {pooling!r}")
# pooling before a classification head (e.g. ModernBertForSequenceClassification)
if (classifier_pooling := self.hparams.get("classifier_pooling")) is not None: