[V1][Spec Decode] Always use argmax for sampling draft tokens (#16899)

Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>

vllm/v1/sample/rejection_sampler.py

@@ -226,7 +226,7 @@ def rejection_sample(
         is_greedy,
         max_spec_len,
         vocab_size,
-        IS_NGRAM=draft_probs is None,
+        NO_DRAFT_PROBS=draft_probs is None,
         num_warps=1,
     )
     return output_token_ids
@@ -423,7 +423,7 @@ def sample_recovered_tokens(
         q,
         vocab_size,
         triton.next_power_of_2(vocab_size),
-        IS_NGRAM=draft_probs is None,
+        NO_DRAFT_PROBS=draft_probs is None,
     )
     return recovered_token_ids
 
@@ -490,7 +490,7 @@ def rejection_random_sample_kernel(
     is_greedy_ptr,  # [batch_size]
     max_spec_len,
     vocab_size,
-    IS_NGRAM: tl.constexpr,
+    NO_DRAFT_PROBS: tl.constexpr,
 ):
     req_idx = tl.program_id(0)
     is_greedy = tl.load(is_greedy_ptr + req_idx)
@@ -509,7 +509,7 @@ def rejection_random_sample_kernel(
     for pos in range(num_draft_tokens):
         if not rejected:
             draft_token_id = tl.load(draft_token_ids_ptr + start_idx + pos)
-            if IS_NGRAM:
+            if NO_DRAFT_PROBS:
                 draft_prob = 1
             else:
                 draft_prob = tl.load(draft_probs_ptr +
@@ -575,7 +575,7 @@ def sample_recovered_tokens_kernel(
     q_ptr,  # [batch_size, vocab_size]
     vocab_size,
     PADDED_VOCAB_SIZE: tl.constexpr,
-    IS_NGRAM: tl.constexpr,
+    NO_DRAFT_PROBS: tl.constexpr,
 ):
     req_idx = tl.program_id(0)
     if req_idx == 0:
@@ -591,7 +591,7 @@ def sample_recovered_tokens_kernel(
         return
 
     vocab_offset = tl.arange(0, PADDED_VOCAB_SIZE)
-    if IS_NGRAM:
+    if NO_DRAFT_PROBS:
         draft_token_id = tl.load(draft_token_ids_ptr + start_idx + pos)
         orig_prob = tl.load(target_probs_ptr + (start_idx + pos) * vocab_size +
                             draft_token_id)
@@ -624,7 +624,7 @@ def sample_recovered_tokens_kernel(
     recovered_id = tl.argmax(prob / q, axis=-1)
     tl.store(output_token_ids_ptr + start_idx + pos, recovered_id)
 
-    if IS_NGRAM:
+    if NO_DRAFT_PROBS:
         # Restore the original probability.
         tl.store(
             target_probs_ptr + (start_idx + pos) * vocab_size + draft_token_id,

vllm/v1/spec_decode/eagle.py

@@ -51,7 +51,7 @@ class EagleProposer:
         # [batch_size, max_num_blocks_per_req]
         block_table: torch.Tensor,
         sampling_metadata: SamplingMetadata,
-    ) -> tuple[torch.Tensor, torch.Tensor]:
+    ) -> torch.Tensor:
         num_tokens = target_token_ids.shape[0]
         batch_size = next_token_ids.shape[0]
         last_token_indices = cu_num_tokens[1:] - 1
@@ -94,17 +94,15 @@ class EagleProposer:
             )
         sample_hidden_states = hidden_states[last_token_indices]
         logits = self.model.compute_logits(sample_hidden_states, None)
-        draft_token_ids, draft_probs = compute_probs_and_sample_next_token(
+        draft_token_ids = logits.argmax(dim=-1)
-            logits, sampling_metadata)
 
         # Early exit if there is only one draft token to be generated.
         if self.num_speculative_tokens == 1:
-            # [batch_size, 1] and [batch_size, 1, vocab_size]
+            # [batch_size, 1]
-            return draft_token_ids.view(-1, 1), draft_probs.unsqueeze(dim=1)
+            return draft_token_ids.view(-1, 1)
 
         # Generate the remaining draft tokens.
         draft_token_ids_list = [draft_token_ids]
-        draft_probs_list = [draft_probs]
 
         positions = target_positions[last_token_indices]
         hidden_states = sample_hidden_states
@@ -159,16 +157,12 @@ class EagleProposer:
                     positions=clamped_positions,
                 )
             logits = self.model.compute_logits(hidden_states, None)
-            draft_token_ids, probs = compute_probs_and_sample_next_token(
+            draft_token_ids = logits.argmax(dim=-1)
-                logits, sampling_metadata)
             draft_token_ids_list.append(draft_token_ids)
-            draft_probs_list.append(probs)
 
         # [batch_size, num_speculative_tokens]
         draft_token_ids = torch.stack(draft_token_ids_list, dim=1)
-        # [batch_size, num_speculative_tokens, vocab_size]
+        return draft_token_ids
-        draft_probs = torch.stack(draft_probs_list, dim=1)
-        return draft_token_ids, draft_probs
 
     @staticmethod
     def prepare_inputs(
@@ -238,6 +232,10 @@ class EagleProposer:
         self.model.lm_head = target_model.lm_head
 
 
+# NOTE(woosuk): Currently, the below code is not used and we always use argmax
+# to sample the draft tokens. We will use this after we find a way to manage
+# the draft prob tensor.
+# Refer to https://github.com/vllm-project/vllm/pull/16899 for the details.
 # FIXME(woosuk): The logic here is duplicated with the main sampling code.
 # We should refactor this to reuse the same sampling implementation.
 def compute_probs_and_sample_next_token(

vllm/v1/worker/gpu_model_runner.py

@@ -1230,7 +1230,7 @@ class GPUModelRunner(LoRAModelRunnerMixin):
                 target_hidden_states = hidden_states[token_indices]
                 target_slot_mapping = attn_metadata.slot_mapping[token_indices]
 
-            draft_token_ids, draft_probs = self.drafter.propose(
+            draft_token_ids = self.drafter.propose(
                 target_token_ids=target_token_ids,
                 target_positions=target_positions,
                 target_hidden_states=target_hidden_states,
@@ -1241,9 +1241,6 @@ class GPUModelRunner(LoRAModelRunnerMixin):
                 sampling_metadata=sampling_metadata,
             )
             spec_token_ids = draft_token_ids.tolist()
-            # TODO(woosuk): Cache draft_probs and use it for rejection sampling
-            # in the next step.
-            del draft_probs
 
         # Clear KVConnector state after all KVs are generated.
         if has_kv_transfer_group():