Update modeling_super_linear.py

modeling_super_linear.py

@@ -464,11 +464,9 @@ class superLinear(nn.Module):
         else:
             out, self.moe_loss = self.moe(x)
 
-        print(F"auto_regressive: {self.auto_regressive}")
-        print(F"max_horizon: {self.max_horizon}")
-        print(F"self.inf_pred_len: {self.inf_pred_len}")
+
         if self.auto_regressive and self.max_horizon < self.inf_pred_len:
-            print("bitch")
+            #print("bitch")
             outputs = [out]
             ar_x = torch.cat([x, out], dim=1)[:, -self.seq_len:]
             for i in range(0, self.inf_pred_len, self.max_horizon):
@@ -497,18 +495,9 @@ class SuperLinearForCausalLM(PreTrainedModel, GenerationMixin):
   
 
         # the backbone keeps its own Config dataclass, so build one on‑the‑fly:
-        #print("SuperLinearForCausalLM")
-        #print(config)
         backbone_cfg   = type("Cfg", (), config.to_dict())()
         self.args      = backbone_cfg
         self.backbone  = superLinear(backbone_cfg)
-
-        # optional final projection: map backbone output to discrete bins
-        # (delete if your model already returns logits over a vocabulary)
-        '''self.vocab_size = getattr(config, "vocab_size", None)
-        if self.vocab_size is not None:
-            self.lm_head = nn.Linear(backbone_cfg.pred_len, self.vocab_size)'''
-
         self.post_init()                             
 
     # ------------------------------------------------------------------