[BugFix] Fix a bug of running chunked-prefill with torchair. (#1378)

This PR fixes a bug of running chunked-prefill with torchair.

Signed-off-by: whx-sjtu <2952154980@qq.com>

Author: whx-sjtu

vllm_ascend/attention/mla_v1.py

@@ -1054,11 +1054,10 @@ def forward(
         ]
         num_actual_toks = attn_metadata.num_actual_tokens
         if k_pe is None and not self.running_in_graph:
-            if not self.torchair_graph_enabled:
-                kv_c, k_pe = self.kv_a_proj_with_mqa(
-                    hidden_states_or_kv_c_normed)[0].split(
-                        [self.kv_lora_rank, self.qk_rope_head_dim], dim=-1)
-                kv_c_normed = self.kv_a_layernorm(kv_c.contiguous())
+            kv_c, k_pe = self.kv_a_proj_with_mqa(
+                hidden_states_or_kv_c_normed)[0].split(
+                    [self.kv_lora_rank, self.qk_rope_head_dim], dim=-1)
+            kv_c_normed = self.kv_a_layernorm(kv_c.contiguous())
         else:
             kv_c_normed = hidden_states_or_kv_c_normed
         assert attn_metadata.num_decodes is not None and \
@@ -1077,12 +1076,13 @@ def forward(
         if not self.running_in_graph:
             hidden_states_or_q_c = hidden_states_or_q_c[:num_actual_toks, ...]
             prefill_hs_or_q_c = hidden_states_or_q_c[num_decode_tokens:]
-            if not self.torchair_graph_enabled:
-                decode_hs_or_q_c = hidden_states_or_q_c[:num_decode_tokens]
-                k_pe = k_pe[:num_actual_toks, ...]
-                k_pe = k_pe.unsqueeze(1)
-                decode_k_pe = k_pe[:num_decode_tokens]
-                prefill_k_pe = k_pe[num_decode_tokens:]
+            decode_hs_or_q_c = hidden_states_or_q_c[:num_decode_tokens]
+            prefill_hs = hidden_states_or_kv_c_normed[num_decode_tokens:]
+            # if not self.torchair_graph_enabled:
+            k_pe = k_pe[:num_actual_toks, ...]
+            k_pe = k_pe.unsqueeze(1)
+            decode_k_pe = k_pe[:num_decode_tokens]
+            prefill_k_pe = k_pe[num_decode_tokens:]
         else:
             decode_hs_or_q_c = hidden_states_or_q_c
         if has_decode:
@@ -1146,11 +1146,11 @@ def forward(
 
                 prefill_q_pe = self.rope_single(prefill_q_pe, cos, sin)
                 prefill_k_pe, prefill_k_nope = self.exec_kv_prefill(
-                    hidden_states_or_kv_c_normed, cos, sin, kv_cache,
-                    attn_metadata.slot_mapping)
+                    prefill_hs, cos, sin, kv_cache,
+                    attn_metadata.slot_mapping[num_decode_tokens:])
 
                 kv_c_normed = prefill_k_nope[:num_actual_toks, ...]
-                prefill_k_c_normed = prefill_k_nope[num_decode_tokens:]
+                prefill_k_c_normed = prefill_k_nope
                 prefill_k_pe = prefill_k_pe.view(num_tokens, self.num_kv_heads,
                                                  -1)
                 prefill_q = torch.cat([prefill_q_nope, prefill_q_pe], dim=-1)