diff --git a/unsloth/models/_utils.py b/unsloth/models/_utils.py index aa4523621d..b20ce25320 100644 --- a/unsloth/models/_utils.py +++ b/unsloth/models/_utils.py @@ -12,7 +12,7 @@ # See the License for the specific language governing permissions and # limitations under the License. -__version__ = "2025.12.2" +__version__ = "2025.12.3" __all__ = [ "SUPPORTS_BFLOAT16", diff --git a/unsloth/trainer.py b/unsloth/trainer.py index dc4e814578..f3e09e990b 100644 --- a/unsloth/trainer.py +++ b/unsloth/trainer.py @@ -60,7 +60,7 @@ _AUTO_PADDING_FREE_ENV_DISABLED = os.environ.get( # [TODO] # Below cannot work with padding-free -_PADDING_FREE_BLOCK_LIST = { +PADDING_FREE_BLOCKLIST = { "gemma2", # - gemma2: Uses slow_attention_softcapping which has torch.compile issues "gpt_oss", # - gpt_oss: Uses Flex Attention which doesn't handle padding_free correctly "mistral", # - mistral: Unfortunately I think sliding window attention doesn't work correctly?