jinaai
/

xlm-roberta-flash-implementation

Inference Endpoints

🇪🇺 Region: EU

Model card Files Files and versions Community

jupyterjazz commited on May 15

Commit

27d23b2

•

1 Parent(s): 4b000ec

lora-multiple-adapters (#11)

- fix: remove old var (1ba907ac288b5edddf2b71406ed211b36d8e9b10)

Files changed (1) hide show

modeling_lora.py +1 -1

modeling_lora.py CHANGED Viewed

@@ -233,7 +233,7 @@ class XLMRobertaLoRA(XLMRobertaModel):
         self._alpha = config.lora_alpha
         self._register_lora(
-            num_adaptations=self._num_adaptations,
             rank=self._rank,
             dropout_p=self._dropout_p,
             alpha=self._alpha,

         self._alpha = config.lora_alpha
         self._register_lora(
+            num_adaptations=len(self._lora_adaptations),
             rank=self._rank,
             dropout_p=self._dropout_p,
             alpha=self._alpha,