Fix incorrect spec_num_draft_tokens in draft_extend (#7757)

python/sglang/srt/layers/dp_attention.py

@@ -237,6 +237,10 @@ def _dp_gather(
         assert (
             local_tokens.untyped_storage() is not global_tokens.untyped_storage()
         ), "aliasing between global_tokens and local_tokens not allowed"
+
+        # NOTE: During draft extend, the gathered_buffer is padded to num_tokens * (speculative_num_steps + 1).
+        # But the size of local_tokens is total accepted tokens. We need to reduce the local_num_tokens to the
+        # actual size of the accepted tokens.
         if forward_batch.forward_mode.is_draft_extend():
             shape_tensor = local_num_tokens.new_full((), local_tokens.shape[0])
             local_num_tokens = torch.minimum(local_num_tokens, shape_tensor)
@@ -291,6 +295,10 @@ def dp_scatter(
         assert (
             local_tokens.untyped_storage() is not global_tokens.untyped_storage()
         ), "aliasing between local_tokens and global_tokens not allowed"
+
+        # NOTE: During draft extend, the gathered_buffer is padded to num_tokens * (speculative_num_steps + 1).
+        # But the size of local_tokens is total accepted tokens. We need to reduce the local_num_tokens to the
+        # actual size of the accepted tokens.
         if forward_batch.forward_mode.is_draft_extend():
             shape_tensor = local_num_tokens.new_full((), local_tokens.shape[0])
             local_num_tokens = torch.minimum(local_num_tokens, shape_tensor)

python/sglang/srt/speculative/eagle_worker.py

@@ -844,7 +844,7 @@ class EAGLEWorker(TpModelWorker):
                 )
         batch.return_hidden_states = False
         model_worker_batch = batch.get_model_worker_batch()
-        model_worker_batch.spec_num_draft_tokens = self.speculative_num_draft_tokens
+        model_worker_batch.spec_num_draft_tokens = self.speculative_num_steps + 1
         assert model_worker_batch.capture_hidden_mode == CaptureHiddenMode.LAST
         forward_batch = ForwardBatch.init_new(
             model_worker_batch, self.draft_model_runner