[Spec] Rename accepted_drafts -> correct_drafts for unambiguous naming (#24081)

python/sglang/srt/managers/tokenizer_manager.py

@@ -2125,9 +2125,13 @@ class TokenizerManager(TokenizerControlMixin, TokenizerManagerScoreMixin):
                     recv_obj.completion_tokens[i] / recv_obj.spec_verify_ct[i]
                 )
 
+                meta_info["spec_num_correct_drafts"] = num_correct_drafts
+                meta_info["spec_num_proposed_drafts"] = num_proposed_drafts
+                meta_info["spec_verify_ct"] = recv_obj.spec_verify_ct[i]
+
+                # FIXME: backward-compat aliases, remove in next release.
                 meta_info["spec_accepted_drafts"] = num_correct_drafts
                 meta_info["spec_proposed_drafts"] = num_proposed_drafts
-                meta_info["spec_verify_ct"] = recv_obj.spec_verify_ct[i]
 
             # Acceptance histogram: tracks how many decoding steps accepted a certain number of draft tokens.
             if (
@@ -2135,6 +2139,10 @@ class TokenizerManager(TokenizerControlMixin, TokenizerManagerScoreMixin):
                 and len(recv_obj.spec_correct_drafts_histogram) > i
                 and recv_obj.spec_correct_drafts_histogram[i]
             ):
+                meta_info["spec_correct_drafts_histogram"] = (
+                    recv_obj.spec_correct_drafts_histogram[i]
+                )
+                # FIXME: backward-compat alias, remove in next release.
                 meta_info["spec_accept_histogram"] = (
                     recv_obj.spec_correct_drafts_histogram[i]
                 )

python/sglang/srt/observability/req_time_stats.py

@@ -609,7 +609,15 @@ class SchedulerReqTimeStats(ReqTimeStatsBase):
         ts = ts or time.perf_counter()
         self.spec_verify_start_time = ts
 
-    def set_spec_verify_end_time(self, ts=None, accepted_tokens: int = 0):
+    def set_spec_verify_end_time(
+        self,
+        ts=None,
+        num_correct_drafts: int = 0,
+        # FIXME: backward-compat alias, remove in next release.
+        accepted_tokens: Optional[int] = None,
+    ):
+        if accepted_tokens is not None:
+            num_correct_drafts = accepted_tokens
         ts = ts or time.perf_counter()
 
         if self.trace_ctx.tracing_enable:
@@ -618,7 +626,11 @@ class SchedulerReqTimeStats(ReqTimeStatsBase):
                 stage,
                 self.spec_verify_start_time,
                 ts,
-                {"accepted_tokens": accepted_tokens},
+                {
+                    "num_correct_drafts": num_correct_drafts,
+                    # FIXME: backward-compat alias, remove in next release.
+                    "accepted_tokens": num_correct_drafts,
+                },
             )
 
     def set_spec_draft_extend_start_time(self, ts=None):

python/sglang/srt/speculative/eagle_info.py

@@ -426,11 +426,11 @@ class EagleVerifyInput(SpecInput, EagleVerifyInputV2Mixin):
         # Iterate every accepted token and check if req has finished after append the token
         # should be checked BEFORE free kv cache slots
         for i, (req, accept_index_row) in enumerate(zip(batch.reqs, accept_index_cpu)):
-            num_accepted = 0
+            num_accept_tokens = 0
             for j, idx in enumerate(accept_index_row):
                 if idx == -1:
                     break
-                num_accepted += 1
+                num_accept_tokens += 1
                 id = predict_cpu[idx]
                 req.output_ids.append(id)
                 if req.require_reasoning and think_end_id is not None:
@@ -451,7 +451,7 @@ class EagleVerifyInput(SpecInput, EagleVerifyInputV2Mixin):
                     accept_index[i, j + 1 :] = -1
                     break
             # Update KV cache tracking for the accepted tokens
-            req.kv_committed_len += num_accepted
+            req.kv_committed_len += num_accept_tokens
             req.kv_allocated_len = req.kv_committed_len
             if not req.finished():
                 unfinished_index.append(i)

python/sglang/srt/speculative/eagle_worker.py

@@ -491,8 +491,12 @@ class EAGLEWorker(TpModelWorker):
 
             if get_global_tracing_enabled():
                 for idx, req in enumerate(batch.reqs):
-                    accepted = verify_output.num_correct_drafts_per_req_cpu[idx]
-                    req.time_stats.set_spec_verify_end_time(accepted_tokens=accepted)
+                    num_correct_drafts = verify_output.num_correct_drafts_per_req_cpu[
+                        idx
+                    ]
+                    req.time_stats.set_spec_verify_end_time(
+                        num_correct_drafts=num_correct_drafts
+                    )
 
             set_time_batch(
                 batch.reqs, "set_spec_draft_extend_start_time", trace_only=True

python/sglang/srt/speculative/frozen_kv_mtp_worker.py

@@ -452,8 +452,10 @@ class FrozenKVMTPWorker(TpModelWorker):
 
         if get_global_tracing_enabled():
             for idx, req in enumerate(batch.reqs):
-                accepted = verify_output.num_correct_drafts_per_req_cpu[idx]
-                req.time_stats.set_spec_verify_end_time(accepted_tokens=accepted)
+                num_correct_drafts = verify_output.num_correct_drafts_per_req_cpu[idx]
+                req.time_stats.set_spec_verify_end_time(
+                    num_correct_drafts=num_correct_drafts
+                )
 
         set_time_batch(batch.reqs, "set_spec_draft_extend_start_time", trace_only=True)
         with self.draft_tp_context(

python/sglang/srt/speculative/ngram_worker.py

@@ -321,12 +321,14 @@ class NGRAMWorker:
 
             if get_global_tracing_enabled():
                 for idx, req in enumerate(batch.reqs):
-                    accepted = (
+                    num_correct_drafts = (
                         verify_input.num_correct_drafts[idx].item()
                         if verify_input.num_correct_drafts is not None
                         else 0
                     )
-                    req.time_stats.set_spec_verify_end_time(accepted_tokens=accepted)
+                    req.time_stats.set_spec_verify_end_time(
+                        num_correct_drafts=num_correct_drafts
+                    )
 
             # Store accept_lens (with bonus) for per-request metrics; downstream
             # subtracts 1 to recover drafts-only counts.

python/sglang/srt/speculative/spec_utils.py

@@ -618,19 +618,19 @@ def traverse_tree(
         if curr == 0:
             # the first token generated by the target model, and thus it is always
             # accepted from the previous iteration
-            accepted = True
+            is_accepted = True
         else:
             parent_bitmask = allocate_token_bitmask[parent_pos]
             curr_token_id = draft_tokens[curr]
             if vocab_size and curr_token_id >= vocab_size:
-                accepted = False
+                is_accepted = False
             else:
                 # 32 boolean bitmask values are packed into 32-bit integers
-                accepted = (
+                is_accepted = (
                     parent_bitmask[curr_token_id // 32] & (1 << (curr_token_id % 32))
                 ) != 0
 
-        if accepted:
+        if is_accepted:
             if curr != 0:
                 # Accept the current token
                 grammar.accept_token(int(draft_tokens[curr]))