ceos/unsloth
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Fix Qwen 3 mapping

Browse source
This commit is contained in:
Daniel Han 2025-05-02 09:44:01 -07:00
commit d957caeae7
2 changed files with 4 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

2
unsloth/models/_utils.py
View file

@ -12,7 +12,7 @@
# See the License for the specific language governing permissions and
# limitations under the License.
__version__ = "2025.4.5"
__version__ = "2025.4.6"
__all__ = [
"SUPPORTS_BFLOAT16",

6
unsloth/models/loader_utils.py
View file

@ -21,9 +21,9 @@ SUPPORTS_FOURBIT = transformers_version >= Version("4.37")
BAD_MAPPINGS = \
{
"unsloth/qwen3-32B-unsloth-bnb-4bit" : "unsloth/Qwen3-32B-bnb-4bit", # 32B dynamic quant is way too big
"unsloth/qwen3-30B-A3B-unsloth-bnb-4bit" : "unsloth/qwen3-30B-A3B", # HF loads MoEs too slowly
"unsloth/qwen3-30B-A3B-bnb-4bit" : "unsloth/qwen3-30B-A3B", # We rather do it on the fly
"unsloth/qwen3-32B-unsloth-bnb-4bit".lower() : "unsloth/Qwen3-32B-bnb-4bit".lower(), # 32B dynamic quant is way too big
"unsloth/qwen3-30B-A3B-unsloth-bnb-4bit".lower() : "unsloth/qwen3-30B-A3B".lower(), # HF loads MoEs too slowly
"unsloth/qwen3-30B-A3B-bnb-4bit".lower() : "unsloth/qwen3-30B-A3B".lower(), # We rather do it on the fly
}
def __get_model_name(