refactor(core): 修改注意力掩码处理函数并重命名参数

khaosz/core/transformer.py

@@ -71,7 +71,7 @@ def apply_rotary_emb(x: Tensor, freqs_cis: Tensor) -> Tensor:
     
     return x_out.to(dtype)
 
-def create_attention_mask(
+def process_attention_mask(
         seq_mask: Tensor, 
         start_pos: int = 0,
         seq_len: int = 0,
@@ -98,6 +98,9 @@ def create_attention_mask(
     if seq_mask is None:
         return None
     
+    if seq_mask.dim() > 2:
+        return seq_mask
+    
     batch_size = seq_mask.size(0)
     seq_mask = seq_mask[:, :start_pos + seq_len].to(device=device, dtype=torch.bool)
     # (bsz, start_pos + seq_len)
@@ -306,7 +309,7 @@ class Transformer(nn.Module):
     def forward(
         self, 
         input_ids: Tensor, 
-        seq_mask: Optional[Tensor]=None,
+        input_mask: Optional[Tensor]=None,
         persistent_key_values: Optional[List[Tuple[Tensor, Tensor]]]=None,
         start_pos: int = 0
     ) -> Tensor:
@@ -318,8 +321,8 @@ class Transformer(nn.Module):
         freqs_cis = self.freq_cis[start_pos:start_pos+seq_len]
         has_kvcache = persistent_key_values is not None
         
-        attn_mask = create_attention_mask(
-            seq_mask, 
+        attn_mask = process_attention_mask(
+            input_mask, 
             start_pos=start_pos,
             seq_len=seq_len,
             is_causal=has_kvcache,