Upload PAL_B_RM_opt

config.json

@@ -20,5 +20,17 @@
   "sfx_temperature": 1.0,
   "sfx_type": "softmax",
   "torch_dtype": "float32",
-  "transformers_version": "4.44.2"
+  "transformers_version": "4.44.2",
+  "uids": [
+    "KZL1qeRzHNYSfDAuOctL1iyVV8WC5N",
+    "ZzGCcAhvqF0HnKxNsUjtJFadcZdyZj",
+    "p4Oh7rUGyLe1EpilJFWr9sPDpkO016",
+    "qo6WIyEh27cwAjWpA3Q60J7NaDxzQJ",
+    "zKV8BFGy60O0q7102ALF84S6Jo5i4q",
+    "i8YiBZlrYmlkkChr5b9BUKvDO6lR1d",
+    "M3icahkfAtC9CJrtKgQ7qvyZ5SD8wC",
+    "HNzkrs9geGu1YMMfZ5Qvdt0ZaCthfB",
+    "Jxv4hxfb9zTVa5nsMDFlnjSX5LZ8MK",
+    "UhQipwcpQmiGJmScocXOGOKyCBaFUg"
+  ]
 }

configuration_pal_b_rm.py

@@ -17,6 +17,7 @@ class PAL_B_Config(PretrainedConfig):
         sfx_temperature: float = 1.0,
         is_temperature_learnable: bool = False,
         is_gumbel_hard: bool = None,
+        uids: list = None,
         **kwargs,
     ):
         self.d_hid = d_hid
@@ -31,5 +32,6 @@ class PAL_B_Config(PretrainedConfig):
         self.sfx_temperature = sfx_temperature
         self.is_temperature_learnable = is_temperature_learnable
         self.is_gumbel_hard = is_gumbel_hard
+        self.uids = uids
         super().__init__(**kwargs)
         

modeling_pal_b_rm.py

@@ -20,8 +20,10 @@ class PAL_B_RM_opt(PreTrainedModel):
             sfx_temperature=config.sfx_temperature,
             is_temperature_learnable=config.is_temperature_learnable,
             is_gumbel_hard=config.is_gumbel_hard,
+            uids=config.uids,
         )
-        # self.model.user_learner.init_weight(uids)
+        if config.uids is not None:
+            self.model.user_learner.init_weight(config.uids)
     
     def forward(self, x):
         logits = self.model(x)