[git] fixed rebased files

colossalai/inference/core/request_handler.py

@@ -227,4 +227,4 @@ class RequestHandler:
 
         self.done_list.extend(finish_seqs)
 
-        return finish_seqs
+        return finish_seqs

colossalai/inference/modeling/layers/attention.py

@@ -58,9 +58,6 @@ def convert_kvcache(cache, lengths, block_tables, pad_id=0):
     seq_len = max(lengths)
     padded_cache = []
     for i in range(bsz):
-        cache1 = cache[block_tables[i][: needed_blocks[i] - 1]].permute((0, 3, 1, 2)).reshape(-1, num_heads, head_size)
-        cache2 = cache[block_tables[i][needed_blocks[i] - 1], :, :, : num_remaing_tokens[i]].permute(2, 0, 1)
-
         _cache = torch.cat(
             (
                 cache[block_tables[i][: needed_blocks[i] - 1]].permute((0, 3, 1, 2)).reshape(-1, num_heads, head_size),
@@ -317,4 +314,4 @@ class PagedAttention:
     ):
         return self.pad_decoding_forward(
             q.unsqueeze(1), k.unsqueeze(1), v.unsqueeze(1), k_cache, v_cache, lengths, block_tables
-        )
+        )

tests/test_infer/test_inference_engine.py

@@ -81,4 +81,4 @@ def test_inference_engine():
 
 
 if __name__ == "__main__":
-    test_inference_engine()
+    test_inference_engine()