Update modeling_Hixtral.py

modeling_Hixtral.py

@@ -788,7 +788,7 @@ class HixtralDecoderLayer(nn.Module):
         super().__init__()
         self.hidden_size = config.hidden_size
 
-        self.self_attn = LLAMOE_ATTENTION_CLASSES[config._attn_implementation](config, layer_idx)
+        self.self_attn = HIXTRAL_ATTENTION_CLASSES[config._attn_implementation](config, layer_idx)
 
         self.block_sparse_moe = HixtralSparseMoeBlock(config)
         self.input_layernorm = HixtralRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
@@ -1020,7 +1020,7 @@ class HixtralModel(HixtralPreTrainedModel):
         self.embed_tokens = value
 
     # Ignore copy
-    @add_start_docstrings_to_model_forward(LLAMOE_INPUTS_DOCSTRING)
+    @add_start_docstrings_to_model_forward(HIXTRAL_INPUTS_DOCSTRING)
     def forward(
         self,
         input_ids: torch.LongTensor = None,