[Speculative Decoding] refactor MTP and optimize spec-decoding postprocess (#6973)

* support new mtp

* refactor(speculate_decoding and mtp): optimize mtp sturcture logic. Update spec-branch status-process

* fix cuda-graph for spec-decoding

* fix xpu mtp and fix some note

* fix unittest and optmize note

* fix model status update in eos-branch

fastdeploy/model_executor/pre_and_post_process.py

@@ -442,8 +442,6 @@ def post_process_specualate(
     think_end_id: int = -1,
     splitwise_role_is_decode: bool = False,
     enable_entropy: bool = False,
-    is_naive_mode: bool = False,
-    prefill_one_step_stop: bool = False,
     routing_replay_manager: RoutingReplayManager = None,
 ):
     if think_end_id > 0:
@@ -502,16 +500,16 @@ def post_process_specualate(
         )
 
     # Unified state update: merges speculate_update + speculate_set_value_by_flags_and_idx
-    # into a single kernel launch. For MTP/ngram paths, verify_draft_tokens has already
-    # handled EOS/max_dec_len detection (replacing tokens + updating step_idx), so
-    # unified_update_model_status acts as a no-op for those checks. For naive mode
-    # (which skips verify), this kernel handles EOS/max_dec_len detection.
+    # into a single kernel launch. Handles EOS detection, max_dec_len truncation, step_idx
+    # advancement, token_ids_all history write, and stop_flags/not_need_stop update for all
+    # paths (MTP, ngram, naive). Note: verify_draft_tokens intentionally does NOT write back
+    # step_idx (it is read-only in that kernel); step_idx is always updated here.
+
     unified_update_model_status(
         model_output.seq_lens_encoder,  # seq_lens_encoder
         model_output.seq_lens_decoder,  # seq_lens_decoder
         model_output.not_need_stop,  # has_running_seqs
         model_output.draft_tokens,  # step_input_ids
-        model_output.actual_draft_token_num,  # adaptive_step_input_len
         model_output.accept_tokens,  # step_output_ids (read-write)
         model_output.accept_num,  # step_output_len (read-write)
         model_output.stop_flags,  # stop_flags (read-write)
@@ -523,8 +521,6 @@ def post_process_specualate(
         model_output.step_idx,  # step_idx (read-write)
         model_output.eos_token_id,  # end_tokens
         model_output.max_dec_len,  # max_dec_len
-        is_naive_mode,  # is_naive_mode
-        prefill_one_step_stop,  # prefill_one_step_stop
     )
 
     if not skip_save_output:
@@ -592,8 +588,6 @@ def post_process(
     think_end_id: int = -1,
     splitwise_role_is_decode: bool = False,
     enable_entropy: bool = False,
-    is_naive_mode: bool = False,
-    prefill_one_step_stop: bool = False,
     routing_replay_manager: RoutingReplayManager = None,
 ) -> None:
     """Post-processing steps after completing a single token generation."""
@@ -621,8 +615,6 @@ def post_process(
                 think_end_id,
                 splitwise_role_is_decode,
                 enable_entropy,
-                is_naive_mode,
-                prefill_one_step_stop,
                 routing_replay_manager,
             )
         else: