[Speculative Decoding][MTP]Support mtp in epdptp mode (#4614)

* support mtp many features

* support mtp reshard in rl mode

* fix function

* support mtp ep

* support mtp in hybird-dp-tp mode

* default open scheduler_v1 in mtp

fastdeploy/engine/args_utils.py

@@ -442,8 +442,7 @@ class EngineArgs:
                 raise NotImplementedError("Only CUDA platform supports logprob.")
             if self.speculative_config is not None and self.logprobs_mode.startswith("processed"):
                 raise NotImplementedError("processed_logprobs not support in speculative.")
-        if self.speculative_config is not None:
-            envs.ENABLE_V1_KVCACHE_SCHEDULER = 0
+
         if self.splitwise_role != "mixed" and self.cache_transfer_protocol != "rdma":
             envs.ENABLE_V1_KVCACHE_SCHEDULER = 0
         if not current_platform.is_cuda() and not current_platform.is_xpu():

fastdeploy/model_executor/layers/attention/append_attn_backend.py

@@ -92,6 +92,8 @@ class AppendAttentionBackend(AttentionBackend):
         self.rope_3d: bool = getattr(fd_config.model_config, "rope_3d", False) or getattr(
             fd_config.model_config, "use_3d_rope", False
         )
+        if fd_config.speculative_config.model_type != "main":
+            self.rope_3d = False
         self.causal: bool = getattr(fd_config.model_config, "causal", True)
         self.speculative_method: str = fd_config.speculative_config.method
         self.use_speculate: bool = self.speculative_method is not None
@@ -364,7 +366,7 @@ class AppendAttentionBackend(AttentionBackend):
                 getattr(layer, "cache_v_zp", None),
                 layer.linear_shift,
                 layer.linear_smooth,
-                forward_meta.attn_mask_offsets,
+                None if self.use_speculate else forward_meta.attn_mask_offsets,
                 metadata.kv_signal_data_list[layer.layer_id],
                 getattr(layer, "q_norm_weight", None),
                 getattr(layer, "k_norm_weight", None),
@@ -383,7 +385,7 @@ class AppendAttentionBackend(AttentionBackend):
                 metadata.max_partition_size,
                 metadata.encoder_max_partition_size,
                 self.speculate_max_draft_token_num + 1,
-                self.causal,
+                self.causal or self.use_speculate,
                 self.speculative_method is not None,
                 sliding_window,
             )

fastdeploy/model_executor/layers/mtp_linear.py

@@ -18,7 +18,7 @@ import paddle
 from paddle import nn
 from paddle.distributed import fleet
 
-from fastdeploy.model_executor.utils import set_weight_attrs
+from fastdeploy.model_executor.utils import default_weight_loader, set_weight_attrs
 
 from .utils import get_tensor
 
@@ -53,44 +53,61 @@ class ParallelEHProjection(nn.Layer):
             self.bias_key = prefix + ".bias"
         else:
             self.bias_key = None
-        self.use_ep = fd_config.parallel_config.use_ep
+        self.fd_config = fd_config
+        self.tp_group = fd_config.parallel_config.tp_group
         self.column_cut = True
+        self.nranks = fd_config.parallel_config.tensor_parallel_size
 
         ColumnParallelLinear = fleet.meta_parallel.ColumnParallelLinear
         RowParallelLinear = fleet.meta_parallel.RowParallelLinear
 
-        if self.use_ep:
-            self.weight = self.create_parameter(
-                shape=[embedding_dim, num_embeddings],
-                dtype=paddle.get_default_dtype(),
-                is_bias=False,
+        if self.column_cut:
+            need_gather = True
+            self.linear = ColumnParallelLinear(
+                embedding_dim,
+                num_embeddings,
+                mp_group=self.tp_group,
+                weight_attr=None,
+                has_bias=True if self.bias_key is not None else False,
+                gather_output=need_gather,
+                fuse_matmul_bias=False,  # False diff更小
             )
-        else:
-            if self.column_cut:
-                need_gather = True
-                self.linear = ColumnParallelLinear(
-                    embedding_dim,
-                    num_embeddings,
-                    mp_group=fleet.get_hybrid_communicate_group().get_model_parallel_group(),
-                    weight_attr=None,
-                    has_bias=True if self.bias_key is not None else False,
-                    gather_output=need_gather,
-                    fuse_matmul_bias=False,  # False diff更小
+            set_weight_attrs(
+                self.linear.weight,
+                {
+                    "weight_loader": default_weight_loader(self.fd_config),
+                    "model_format": self.fd_config.model_config.model_format,
+                },
+            )
+            if self.bias_key is not None:
+                set_weight_attrs(
+                    self.linear.bias,
+                    {"rl_need_attr": {"rl_tp_degree": fd_config.parallel_config.tensor_parallel_size}},
                 )
+            if self.nranks > 1:
                 set_weight_attrs(self.linear.weight, {"output_dim": True})
-                if self.bias_key is not None:
-                    set_weight_attrs(self.linear.bias, {"output_dim": True})
-            else:
-                self.linear = RowParallelLinear(
-                    embedding_dim,
-                    num_embeddings,
-                    mp_group=fleet.get_hybrid_communicate_group().get_model_parallel_group(),
-                    weight_attr=None,
-                    has_bias=True if self.bias_key is not None else False,
-                    input_is_parallel=False,
-                    fuse_matmul_bias=False,  # False diff更小
-                )
-                set_weight_attrs(self.linear.weight, {"output_dim": False})
+        else:
+            self.linear = RowParallelLinear(
+                embedding_dim,
+                num_embeddings,
+                mp_group=self.tp_group,
+                weight_attr=None,
+                has_bias=True if self.bias_key is not None else False,
+                input_is_parallel=False,
+                fuse_matmul_bias=False,  # False diff更小
+            )
+            set_weight_attrs(
+                self.linear.weight,
+                {
+                    "weight_loader": default_weight_loader(self.fd_config),
+                    "model_format": self.fd_config.model_config.model_format,
+                },
+            )
+            if self.nranks > 1:
+                set_weight_attrs(self.linear.weight, {"output_dim": True})
+        set_weight_attrs(
+            self.linear.weight, {"rl_need_attr": {"rl_tp_degree": fd_config.parallel_config.tensor_parallel_size}}
+        )
 
     def load_state_dict(self, state_dict):
         """
@@ -100,17 +117,14 @@ class ParallelEHProjection(nn.Layer):
             state_dict (dict): A dictionary containing the checkpoint weights and biases.
         """
 
-        if self.use_ep:
-            self.weight.set_value(get_tensor(state_dict.pop(self.weight_key)).astype(paddle.get_default_dtype()))
-        else:
-            weight_tensor = get_tensor(state_dict.pop(self.weight_key)).astype(paddle.get_default_dtype())
-            if self.linear.weight.shape != weight_tensor.shape:
-                weight_tensor = weight_tensor.transpose([1, 0])
-            self.linear.weight.set_value(weight_tensor)
+        weight_tensor = get_tensor(state_dict.pop(self.weight_key)).astype(paddle.get_default_dtype())
+        if self.linear.weight.shape != weight_tensor.shape:
+            weight_tensor = weight_tensor.transpose([1, 0])
+        self.linear.weight.set_value(weight_tensor)
 
-            if self.bias_key is not None:
-                bias = get_tensor(state_dict.pop(self.bias_key)).astype(paddle.get_default_dtype())
-                self.linear.bias.set_value(bias)
+        if self.bias_key is not None:
+            bias = get_tensor(state_dict.pop(self.bias_key)).astype(paddle.get_default_dtype())
+            self.linear.bias.set_value(bias)
 
     def forward(self, input):
         """
@@ -123,8 +137,5 @@ class ParallelEHProjection(nn.Layer):
             Tensor: The output tensor after processing through the layer.
         """
         logits = input
-        if self.use_ep:
-            logits = paddle.matmul(logits, self.weight)
-        else:
-            logits = self.linear(logits)
+        logits = self.linear(logits)
         return logits

fastdeploy/model_executor/model_loader/default_loader.py

@@ -72,6 +72,11 @@ class DefaultModelLoader(BaseModelLoader):
             # register rl model
             import fastdeploy.rl  # noqa
 
+            if fd_config.speculative_config.model_type != "mtp":
+                architectures = architectures.replace("Ernie5ForCausalLM", "Ernie5MoeForCausalLM")
+            else:
+                architectures = architectures.replace("Ernie5ForCausalLM", "Ernie5MTPForCausalLM")
+
             architectures = architectures + "RL"
             context = paddle.LazyGuard()
         else:

fastdeploy/model_executor/model_loader/default_loader_v1.py

@@ -65,6 +65,11 @@ class DefaultModelLoaderV1(BaseModelLoader):
             # register rl model
             import fastdeploy.rl  # noqa
 
+            if fd_config.speculative_config.model_type != "mtp":
+                architectures = architectures.replace("Ernie5ForCausalLM", "Ernie5MoeForCausalLM")
+            else:
+                architectures = architectures.replace("Ernie5ForCausalLM", "Ernie5MTPForCausalLM")
+
             architectures = architectures + "RL"
 
         enable_cache, _, weight_cache_context = is_weight_cache_enabled(fd_config)

fastdeploy/output/token_processor.py

@@ -502,7 +502,7 @@ class TokenProcessor:
 
     def _compute_speculative_status(self):
         # TODO(liuzichang): Supplement more statistics
-        interval = 10
+        interval = 1
         if self.speculative_stats_step % interval == 0:
             accept_ratio = 1 - self.total_step * 1.0 / self.number_of_output_tokens
             spec_logger.info(
@@ -593,6 +593,9 @@ class TokenProcessor:
                         + accept_num[i]
                     ].tolist()
                     if (not recovery_stop) and (len(token_ids) == 0 or token_ids[-1] <= 0):
+                        if envs.ENABLE_V1_KVCACHE_SCHEDULER:
+                            if task_id in self.resource_manager.to_be_rescheduled_request_id_set:
+                                self.resource_manager.reschedule_preempt_task(task_id)
                         continue
             else:
                 token_id = int(tokens[i, 0])

fastdeploy/rl/dynamic_weight_manager.py

@@ -17,11 +17,10 @@
 import os
 import time
 from multiprocessing.shared_memory import SharedMemory
-from typing import Any, Dict
+from typing import Any, Dict, List
 
 import numpy as np
 import paddle
-from paddle import nn
 from paddleformers.utils.log import logger
 
 from fastdeploy.config import FDConfig
@@ -31,7 +30,7 @@ from fastdeploy.inter_communicator import ModelWeightsStatus
 class DynamicWeightManager:
     """Manages model weights loading, updating and shared state across processes."""
 
-    def __init__(self, fd_config: FDConfig, model: nn.Layer):
+    def __init__(self, fd_config: FDConfig, models):
         """Initialize with config and model instances."""
         self.fd_config = fd_config
         self.load_config = fd_config.load_config
@@ -42,7 +41,10 @@ class DynamicWeightManager:
         self.meta_src_id = self._get_gpu_id()
         self.first_load = True
         self.ipc_path = f"/shared_ipc_meta/ipc_metas_{self.meta_src_id}"
-        self.model: nn.Layer = model
+        if not isinstance(models, List):
+            self.model_list = [models]
+        else:
+            self.model_list = models
         self._capture_model_state()
         self.update_parameters()
         self.finalize_update()
@@ -55,9 +57,10 @@ class DynamicWeightManager:
     @paddle.no_grad()
     def _capture_model_state(self):
         """Capture and store initial model parameters state."""
-        for name, param in self.model.state_dict().items():
-            logger.debug(f"Model param: {name}, shape={param.shape}, dtype={param.dtype}")
-            self.state_dict[name] = param
+        for model in self.model_list:
+            for name, param in model.state_dict().items():
+                logger.info(f"Model param: {name}, shape={param.shape}, dtype={param.dtype}")
+                self.state_dict[name] = param
 
     def update_parameters(self, pid: int = 0) -> None:
         """Core method to update model parameters based on strategy."""
@@ -137,8 +140,9 @@ class DynamicWeightManager:
 
         paddle.device.cuda.empty_cache()
         # step2: release model weight
-        for param in self.model.state_dict().values():
-            param._clear_data()
+        for model in self.model_list:
+            for param in model.state_dict().values():
+                param._clear_data()
 
         self._verify_parameters("clearance")
 

fastdeploy/spec_decode/base.py

@@ -38,13 +38,20 @@ class Proposer(ABC):
         Init Speculative proposer
         """
         fd_config.parallel_config.tp_group = None
+        fd_config.parallel_config.ep_group = None
         self.fd_config = deepcopy(fd_config)
         fd_config.parallel_config.tp_group = dist.get_group(
             fd_config.parallel_config.data_parallel_rank + envs.FD_TP_GROUP_GID_OFFSET
         )
+        fd_config.parallel_config.ep_group = dist.get_group(
+            fd_config.parallel_config.data_parallel_size + envs.FD_TP_GROUP_GID_OFFSET
+        )
         self.fd_config.parallel_config.tp_group = dist.get_group(
             fd_config.parallel_config.data_parallel_rank + envs.FD_TP_GROUP_GID_OFFSET
         )
+        self.fd_config.parallel_config.ep_group = dist.get_group(
+            fd_config.parallel_config.data_parallel_size + envs.FD_TP_GROUP_GID_OFFSET
+        )
         self.parallel_config = self.fd_config.parallel_config
         self.model_config = self.fd_config.model_config
         self.speculative_config = self.fd_config.speculative_config

fastdeploy/spec_decode/mtp.py

@@ -96,7 +96,8 @@ class MTPProposer(Proposer):
         """
         Update config for MTP from global config
         """
-        self.model_config.architectures[0] = "Ernie4_5_MTPForCausalLM"
+        self.forward_meta: ForwardMeta = None
+        self.model_config.architectures[0] = self.model_config.architectures[0].replace("Moe", "MTP")
         self.speculative_config.sharing_model = main_model
         self.model_config.num_hidden_layers = 1
         self.model_config.model = self.speculative_config.model
@@ -169,6 +170,9 @@ class MTPProposer(Proposer):
         kv_cache_shape = self.attn_backends[0].get_kv_cache_shape(
             max_num_blocks=self.num_gpu_blocks, kv_cache_quant_type=kv_cache_quant_type
         )
+        if kv_cache_quant_type == "block_wise_fp8":
+            kv_cache_scale_shape = [kv_cache_shape[0], kv_cache_shape[1], kv_cache_shape[2]]
+        local_rank = self.local_rank % self.parallel_config.tensor_parallel_size
         if not profile and (
             self.cache_config.enable_prefix_caching or self.scheduler_config.splitwise_role != "mixed"
         ):
@@ -178,8 +182,8 @@ class MTPProposer(Proposer):
                 self.num_main_model_layers + self.model_config.num_hidden_layers,
             ):
                 key_cache = paddle.empty(shape=[], dtype=cache_type)
-                key_cache_name = f"key_caches_{i}_rank{self.local_rank}.device{self.device_id}"
-                val_cache_name = f"value_caches_{i}_rank{self.local_rank}.device{self.device_id}"
+                key_cache_name = f"key_caches_{i}_rank{local_rank}.device{self.device_id}"
+                val_cache_name = f"value_caches_{i}_rank{local_rank}.device{self.device_id}"
                 key_cache = share_external_data(key_cache, key_cache_name, kv_cache_shape)
                 cache_kvs_list.append(key_cache)
                 value_cache = paddle.empty(shape=[], dtype=cache_type)
@@ -199,6 +203,17 @@ class MTPProposer(Proposer):
                     fill_value=0,
                     dtype=cache_type,
                 )
+                if kv_cache_quant_type == "block_wise_fp8":
+                    self.cache_kvs[f"key_cache_scales_{i}"] = paddle.full(
+                        shape=kv_cache_scale_shape,
+                        fill_value=0,
+                        dtype=paddle.get_default_dtype(),
+                    )
+                    self.cache_kvs[f"value_cache_scales_{i}"] = paddle.full(
+                        shape=kv_cache_scale_shape,
+                        fill_value=0,
+                        dtype=paddle.get_default_dtype(),
+                    )
             self.model_inputs["caches"] = list(self.cache_kvs.values())
             for value in self.cache_kvs.values():
                 del value
@@ -430,11 +445,10 @@ class MTPProposer(Proposer):
         if "caches" not in self.model_inputs:
             self.initialize_kv_cache()
         req_len = len(req_dicts)
-        # has_prefill_task = False
-        # has_decode_task = False
+
         for i in range(req_len):
             request = req_dicts[i]
-            logger.info(f"{i}th request-{request.request_id}: {request}")
+            logger.debug(f"{i}th request-{request.request_id}: {request}")
             idx = request.idx
             if request.task_type.value == RequestType.PREFILL.value:  # prefill task
                 prefill_start_index = request.prefill_start_index
@@ -688,7 +702,7 @@ class MTPProposer(Proposer):
             self.max_model_len,
             self.model_inputs["substep"],
         )
-        if self.role == "prefill":
+        if self.role == "prefill" and self.parallel_config.tensor_parallel_rank == 0:
             mtp_save_first_token(
                 self.model_inputs["base_model_draft_tokens"],
                 self.model_inputs["not_need_stop"],
@@ -820,11 +834,18 @@ class MTPProposer(Proposer):
                     )
 
                 if self.parallel_config.tensor_parallel_size > 1:
-                    paddle.distributed.broadcast(sampled_token_ids, 0)
+                    paddle.distributed.broadcast(
+                        sampled_token_ids,
+                        self.parallel_config.data_parallel_rank * self.parallel_config.tensor_parallel_size,
+                        group=self.parallel_config.tp_group,
+                    )
 
                 self._post_process(sampled_token_ids)
                 if substep != self.num_model_steps - 1:
                     self._get_self_hidden_states(hidden_states)
+            else:
+                if hasattr(self.model, "empty_input_forward"):
+                    self.model.empty_input_forward()
 
     def _get_self_hidden_states(self, hidden_states):
         target_hidden_states = eagle_get_self_hidden_states(

fastdeploy/worker/worker_process.py

@@ -778,13 +778,6 @@ def initialize_fd_config(args, ranks: int = 1, local_rank: int = 0) -> FDConfig:
     logger.info(f"- Dynamic load weight: {load_config.dynamic_load_weight}")
     logger.info(f"- Load strategy: {load_config.load_strategy}")
 
-    if (
-        args.speculative_config is not None
-        and ("method" in args.speculative_config)
-        and (args.speculative_config["method"] is not None)
-    ):
-        logger.info("Set ENABLE_V1_KVCACHE_SCHEDULER to 0 due to not support speculative decoding now.")
-        envs.ENABLE_V1_KVCACHE_SCHEDULER = 0
     if args.splitwise_role != "mixed" and args.cache_transfer_protocol != "rdma":
         envs.ENABLE_V1_KVCACHE_SCHEDULER = 0
     if not current_platform.is_cuda() and not current_platform.is_xpu():