TensorRT-LLMs/tensorrt_llm/_torch/speculative/eagle3.py
2025-07-10 12:16:57 +09:00

556 lines
23 KiB
Python

from dataclasses import dataclass, field
from typing import List, Optional, Tuple
import torch
from torch import nn
from tensorrt_llm.logger import logger
from tensorrt_llm.mapping import Mapping
from ..attention_backend import AttentionMetadata
from ..pyexecutor.llm_request import LlmRequest
from ..pyexecutor.resource_manager import BaseResourceManager, SlotManager
from ..pyexecutor.sampler import SampleState, SampleStateTensors, TorchSampler
from ..pyexecutor.scheduler import ScheduledRequests
from .interface import SpecConfig, SpecMetadata, SpeculativeDecodingMode
from .mtp import MTPSampler
@dataclass
class Eagle3Config(SpecConfig):
spec_dec_name: str = "EAGLE3"
num_layers: int = 0
hidden_size: int = 0
eagle3_one_model: bool = True
def __post_init__(self):
if self.draft_model_path is None:
raise ValueError("Path to EAGLE3 weights must be specified.")
if self.eagle3_one_model:
self.spec_dec_mode = SpeculativeDecodingMode.EAGLE3_ONE_MODEL
self.num_extra_kv_tokens = self.max_draft_tokens - 1
else:
self.spec_dec_mode = SpeculativeDecodingMode.from_string(
self.spec_dec_name)
self.num_extra_kv_tokens = 0
logger.info(f"EAGLE3 Config: {self}")
def update_from_model_config(self, model_config):
self.num_layers = model_config.num_hidden_layers
self.hidden_size = model_config.hidden_size
self.dtype = model_config.torch_dtype
def get_draft_model_prompt(self,
input_tokens: torch.Tensor) -> torch.Tensor:
"""
Eagle3 always throws away the first token when processing draft inputs
"""
return input_tokens[1:]
class Eagle3ResourceManager(BaseResourceManager):
"""
Eagle3 needs to save the hidden states for the draft model. When using
Eagle3TwoModel, there will be two model engines, one for the target model
and one for the draft model. Use this class to manage the hidden states.
"""
def __init__(self, config: Eagle3Config, dtype: torch.dtype,
hidden_size: int, max_num_requests: int, max_seq_len: int,
max_num_tokens: int):
self.dtype = dtype
self.max_draft_tokens = config.max_draft_tokens
self.hidden_size = hidden_size
self.max_num_requests = max_num_requests
self.max_seq_len = max_seq_len
self.slot_manager = SlotManager(max_num_requests)
# empty hidden states tensor
max_num_tokens = min(max_num_tokens,
max_num_requests * self.max_seq_len)
self.hidden_states = torch.empty((max_num_tokens, self.hidden_size * 3),
dtype=self.dtype,
device='cuda')
# sequence length, only used for metadata preparation
self.seq_lens = {i: 0 for i in range(max_num_requests)}
# start indices of each slot
self.start_indices = {i: 0 for i in range(max_num_requests)}
# whether the next draft forward is the first
self.is_first_draft = True
def prepare_resources(self, scheduled_batch: ScheduledRequests):
context_batch = scheduled_batch.context_requests
# allocate hidden state tensors and update slot ids
self.slot_ids = []
for req in context_batch:
if req.is_first_context_chunk:
slot_id = self.slot_manager.add_slot(req.request_id)
self.slot_ids.append(slot_id)
# reset the flag before model forward
self.is_first_draft = True
def update_resources(self, scheduled_batch: ScheduledRequests):
pass
def free_resources(self, request: LlmRequest):
self.slot_manager.remove_slot(request.request_id)
def add_dummy_requests(self, request_ids: List[int]):
for rid in request_ids:
self.slot_manager.add_slot(rid)
def shutdown(self):
pass
def get_max_resource_count(self) -> int:
return self.max_num_requests
def get_needed_resource_to_completion(self, request: LlmRequest):
return 0
@dataclass
class Eagle3SpecMetadata(SpecMetadata):
hidden_states: List[torch.Tensor] = field(default_factory=list)
num_capture_layers: int = 3
layers_to_capture: Tuple[int, ...] = field(init=False)
target_model_embed_tokens: Optional[torch.nn.Module] = None
hidden_size: int = 0
max_num_tokens: int = 0
dtype: torch.dtype = torch.bfloat16
is_draft_model: bool = False
is_first_draft: bool = False
eagle3_resource_manager: Optional[Eagle3ResourceManager] = None
def __post_init__(self):
if self.num_layers == 1:
self.layers_to_capture = (0, )
else:
if self.num_layers <= 5:
raise ValueError("Not enough hidden layers for EAGLE")
self.layers_to_capture = (1, self.num_layers // 2 - 1,
self.num_layers - 4)
# Initialize to 0 to avoid reading uninitialized memory during warmup
self.hidden_states_read_indices = torch.zeros([self.max_num_tokens],
dtype=torch.long,
device='cuda')
self.hidden_states_write_indices = torch.zeros([self.max_num_tokens],
dtype=torch.long,
device='cuda')
self.hidden_states_read_indices_host = None
self.hidden_states_write_indices_host = None
def prepare(self):
is_first_draft = self.eagle3_resource_manager.is_first_draft
# Update start indices
# Here, we assume the sequence lengths (seq_lens) during the draft model
# forward will not exceed those of the target model. So pre-allocate
# hidden state space before the target model forward.
start_idx = 0
if not self.is_draft_model:
for req_id, seq_len in zip(self.request_ids, self.seq_lens):
slot_id = self.eagle3_resource_manager.slot_manager.get_slot(
req_id)
self.eagle3_resource_manager.start_indices[slot_id] = start_idx
start_idx += seq_len
# Prepare hidden states gather ids
hidden_states_read_indices = []
hidden_states_write_indices = []
for req_id, seq_len in zip(self.request_ids, self.seq_lens):
slot_id = self.eagle3_resource_manager.slot_manager.get_slot(req_id)
start_idx = self.eagle3_resource_manager.start_indices[slot_id]
# If this is the first draft or the target model forward, we need to
# read/write all of the hidden states, otherwise, only read the last token
if is_first_draft or not self.is_draft_model:
hidden_states_read_indices.extend(
list(range(start_idx, start_idx + seq_len)))
hidden_states_write_indices.extend(
list(range(start_idx, start_idx + seq_len)))
else:
old_seq_len = self.eagle3_resource_manager.seq_lens[slot_id]
hidden_states_read_indices.append(start_idx + old_seq_len - 1)
hidden_states_write_indices.append(start_idx + seq_len - 1)
self.eagle3_resource_manager.seq_lens[slot_id] = seq_len
# Prepare hidden states gather ids
self.hidden_states_read_indices_host = torch.tensor(
hidden_states_read_indices, dtype=torch.long, pin_memory=True)
self.hidden_states_write_indices_host = torch.tensor(
hidden_states_write_indices, dtype=torch.long, pin_memory=True)
self.is_first_draft = is_first_draft and self.is_draft_model
if self.is_draft_model:
self.eagle3_resource_manager.is_first_draft = False
self.hidden_states_read_indices[:self.num_tokens].copy_(
self.hidden_states_read_indices_host, non_blocking=True)
self.hidden_states_write_indices[:self.num_tokens].copy_(
self.hidden_states_write_indices_host, non_blocking=True)
def is_layer_capture(self, layer_id: int):
return layer_id in self.layers_to_capture
def maybe_capture_hidden_states(
self,
layer_id: int,
hidden_states: torch.Tensor,
residual: Optional[torch.Tensor] = None) -> None:
token_idx = self.hidden_states_write_indices[:self.num_tokens]
eagle3_hidden_states = self.eagle3_resource_manager.hidden_states
for i, captured_layer_id in enumerate(self.layers_to_capture):
if captured_layer_id == layer_id:
to_save = hidden_states + residual if residual is not None else hidden_states
to_save = to_save.to(dtype=eagle3_hidden_states.dtype)
eagle3_hidden_states[:, i * self.hidden_size:(i + 1) *
self.hidden_size].index_copy_(
0, token_idx, to_save)
break
def get_hidden_states(self):
hidden_states = self.eagle3_resource_manager.hidden_states[
self.hidden_states_read_indices[:self.num_tokens], :]
if not self.is_first_draft:
hidden_states = hidden_states[:, :self.hidden_size]
return hidden_states
class Eagle3Sampler(TorchSampler):
def _batch_sample(self, scheduled_requests, model_outputs) -> SampleState:
logits = model_outputs["logits"]
new_tokens_device = torch.argmax(logits, dim=-1)
if "d2t" in model_outputs:
d2t = model_outputs["d2t"]
new_tokens_device = d2t[new_tokens_device] + new_tokens_device
device = SampleStateTensors(new_tokens=new_tokens_device)
host = SampleStateTensors(
new_tokens=new_tokens_device.to('cpu', non_blocking=True))
sampler_event = torch.cuda.Event()
sampler_event.record()
return SampleState(scheduled_requests=scheduled_requests,
logits=logits,
device=device,
host=host,
sampler_event=sampler_event)
@dataclass
class Eagle3OneModelSpecMetadata(SpecMetadata):
# The hidden states
hidden_states: Optional[torch.Tensor] = None
# The layers to be captured
layers_to_capture: Tuple[int, ...] = field(init=False)
# The hidden size of the hidden states
hidden_size: int = 0
# The max number of tokens
max_num_tokens: int = 0
# The dtype of the hidden states
dtype: torch.dtype = torch.bfloat16
# The index of the batche inputs
batch_indices_cuda: Optional[torch.Tensor] = None
def __post_init__(self):
if self.num_layers == 1:
self.layers_to_capture = (1, )
else:
if self.num_layers <= 5:
raise ValueError("Not enough hidden layers for EAGLE")
self.layers_to_capture = (1, self.num_layers // 2 - 1,
self.num_layers - 4)
self.hidden_states = torch.empty(
(self.max_num_tokens,
self.hidden_size * len(self.layers_to_capture)),
dtype=self.dtype,
device='cuda')
self.batch_indices_cuda = torch.empty(
[self.max_num_requests],
dtype=torch.int,
device='cuda',
)
def is_layer_capture(self, layer_id: int):
return layer_id in self.layers_to_capture
def prepare(self):
assert self.request_ids is not None
# update batch indeices
num_seqs = len(self.request_ids)
batch_indices = torch.arange(num_seqs,
dtype=torch.int,
device='cpu',
pin_memory=True)
self.batch_indices_cuda[:num_seqs].copy_(batch_indices,
non_blocking=True)
self.num_tokens -= (self.num_generations) * self.max_draft_tokens
def maybe_capture_hidden_states(
self,
layer_id: int,
hidden_states: torch.Tensor,
residual: Optional[torch.Tensor] = None) -> None:
for i, captured_layer_id in enumerate(self.layers_to_capture):
if captured_layer_id == layer_id:
num_tokens = hidden_states.shape[0]
to_save = hidden_states + residual if residual is not None else hidden_states
self.hidden_states[:num_tokens, i * self.hidden_size:(i + 1) *
self.hidden_size].copy_(to_save,
non_blocking=True)
break
class Eagle3Decoder(TorchSampler):
def _batch_sample(self, scheduled_requests, model_outputs) -> SampleState:
logits = model_outputs["logits"]
new_tokens_device = torch.argmax(logits, dim=-1)
if "d2t" in model_outputs:
d2t = model_outputs["d2t"]
new_tokens_device = d2t[new_tokens_device] + new_tokens_device
new_tokens_host = new_tokens_device.to('cpu', non_blocking=True)
new_tensors_device = {"new_tokens_device": new_tokens_device}
new_tensors_host = {"new_tokens_host": new_tokens_host}
decoder_event = torch.cuda.Event()
decoder_event.record()
return SampleState(scheduled_requests=scheduled_requests,
logits=logits,
new_tensors_device=new_tensors_device,
new_tensors_host=new_tensors_host,
decoder_event=decoder_event)
class Eagle3OneModelDecoder(MTPSampler):
def __init__(self, max_seq_len: int, config: Eagle3Config):
super().__init__(max_seq_len, None)
self.draft_len = config.max_draft_tokens
class Eagle3OneModelWorker(nn.Module):
def __init__(self, spec_config: Eagle3Config, mapping: Mapping):
super().__init__()
self.spec_config = spec_config
self.max_draft_tokens = self.spec_config.max_draft_tokens
self.mapping = mapping
@torch.compile(options={"max-autotune": True})
def forward(self, input_ids, position_ids, hidden_states, logits,
attn_metadata, spec_metadata, draft_model):
batch_size = attn_metadata.num_seqs
num_contexts = attn_metadata.num_contexts
num_gens = batch_size - num_contexts
raw_logits = logits
# Sample and accept tokens
accepted_tokens, num_accepted_tokens = self.sample_and_accept_draft_tokens(
logits, attn_metadata, spec_metadata)
# Save the old attn_metadata and spec_metadata
if attn_metadata.is_cuda_graph:
seq_len = attn_metadata._seq_lens[:batch_size].clone()
seq_len_cuda = attn_metadata._seq_lens_cuda[:batch_size].clone()
# Prepare inputs for the 1st draft model forward
position_ids = position_ids.squeeze(0)
last_tokens_idx = torch.cumsum(
attn_metadata.seq_lens_cuda, dim=0, dtype=torch.long) - 1
inputs = self.prepare_1st_drafter_inputs(
input_ids=input_ids,
position_ids=position_ids,
last_tokens_idx=last_tokens_idx,
hidden_states=hidden_states,
accepted_tokens=accepted_tokens,
attn_metadata=attn_metadata,
spec_metadata=spec_metadata,
draft_model=draft_model)
# Predict draft tokens
next_draft_tokens = []
for i in range(self.max_draft_tokens):
hidden_states, hidden_states_to_save = draft_model.model(**inputs)
if i == 0:
start_ids_gen = (spec_metadata.batch_indices_cuda[:num_gens] *
(self.max_draft_tokens + 1)).long()
gather_ids_gen = (start_ids_gen +
num_accepted_tokens[num_contexts:] - 1 +
attn_metadata.num_ctx_tokens)
gather_ids = torch.concat(
[last_tokens_idx[:num_contexts], gather_ids_gen], dim=0)
else:
# All of the seq_len are 1, use batch_indices_cuda as gather_ids
gather_ids = spec_metadata.batch_indices_cuda[:batch_size]
logits = draft_model.logits_processor(hidden_states[gather_ids],
draft_model.lm_head,
attn_metadata, True)
new_draft_token = self.draft_decoder(logits, draft_model)
next_draft_tokens.append(new_draft_token)
# update inputs
hidden_states = hidden_states_to_save[gather_ids]
position_ids = inputs["position_ids"][gather_ids] + 1
# update attn_metadata
if i == 0:
attn_metadata._seq_lens[:batch_size].fill_(1)
attn_metadata._seq_lens_cuda[:batch_size].fill_(1)
attn_metadata.on_update()
# cannot run generation if their is no kv cache
if inputs["attn_metadata"].kv_cache_manager is not None:
attn_metadata.host_request_types[:attn_metadata.
num_contexts].fill_(1)
attn_metadata.num_contexts = 0
# update kv_lens_cuda
if hasattr(attn_metadata, 'kv_lens_cuda'):
attn_metadata.kv_lens_cuda[num_contexts:batch_size] -= (
self.max_draft_tokens -
num_accepted_tokens[num_contexts:])
attn_metadata.kv_lens_cuda[:num_contexts] += 1
elif hasattr(attn_metadata, 'kv_lens_cuda'):
attn_metadata.kv_lens_cuda[:batch_size] += 1
# support attention dp
if spec_metadata.all_rank_num_tokens is not None:
spec_metadata.all_rank_num_tokens = spec_metadata.all_rank_num_seqs
inputs = {
"input_ids": new_draft_token,
"position_ids": position_ids,
"hidden_states": hidden_states,
"attn_metadata": attn_metadata,
"spec_metadata": spec_metadata,
}
next_draft_tokens = torch.stack(next_draft_tokens, dim=1)
# restore attn_metadata to support cuda graph
if attn_metadata.is_cuda_graph:
attn_metadata._seq_lens[:batch_size].copy_(seq_len)
attn_metadata._seq_lens_cuda[:batch_size].copy_(seq_len_cuda)
attn_metadata.on_update()
# prepare next new tokens to support overlap scheduler
next_new_tokens = accepted_tokens[
spec_metadata.batch_indices_cuda[:batch_size],
num_accepted_tokens - 1].unsqueeze(1)
next_new_tokens = torch.concat([next_new_tokens, next_draft_tokens],
dim=1)
return {
'logits': raw_logits,
'new_tokens': accepted_tokens,
'new_tokens_lens': num_accepted_tokens,
'next_draft_tokens': next_draft_tokens,
'next_new_tokens': next_new_tokens,
}
def sample_and_accept_draft_tokens(
self,
logits: torch.Tensor,
attn_metadata: AttentionMetadata,
spec_metadata: Eagle3OneModelSpecMetadata,
):
batch_size = attn_metadata.num_seqs
num_contexts = attn_metadata.num_contexts
num_gens = batch_size - num_contexts
if logits.dim() == 1:
logits = logits.unsqueeze(0)
# The return buffer
accepted_tokens = torch.empty((batch_size, (self.max_draft_tokens + 1)),
dtype=torch.int,
device=logits.device)
num_accepted_tokens = torch.ones(batch_size,
dtype=torch.int,
device=logits.device)
# Do greedy sampling for the input logits
target_tokens = torch.argmax(logits, dim=-1)
# context
accepted_tokens[:num_contexts, 0] = target_tokens[:num_contexts]
# generation
gen_target_tokens = target_tokens[num_contexts:].reshape(
num_gens, self.max_draft_tokens + 1)
accepted_tokens[num_contexts:, :] = gen_target_tokens
draft_tokens = spec_metadata.draft_tokens.reshape(
num_gens, self.max_draft_tokens)
num_accepted_tokens[num_contexts:] += torch.cumprod((
draft_tokens == gen_target_tokens[:, :self.max_draft_tokens]).int(),
dim=-1).sum(1)
return accepted_tokens, num_accepted_tokens
def draft_decoder(
self,
logits: torch.Tensor,
draft_model: nn.Module,
):
'''
Sampling draft tokens.
Args:
logits: torch.Tensor
[num_tokens, vocab_size]
Logits produced by the draft model.
draft_model: nn.Module
The draft model.
Returns:
draft_tokens: torch.Tensor
[batch_size * max_draft_tokens]
Draft token ids. Flattened.
'''
draft_tokens = torch.argmax(logits, dim=-1).type(torch.int32)
# Apply d2t (offsets between draft model dictionary and main model dictionary).
if hasattr(draft_model.model,
"d2t") and draft_model.model.d2t is not None:
draft_tokens = draft_model.model.d2t[draft_tokens] + draft_tokens
return draft_tokens
def prepare_1st_drafter_inputs(
self,
input_ids: torch.LongTensor,
position_ids: torch.LongTensor,
last_tokens_idx: torch.LongTensor,
hidden_states: torch.Tensor,
accepted_tokens: torch.Tensor,
attn_metadata: AttentionMetadata,
spec_metadata: Eagle3OneModelSpecMetadata,
draft_model: nn.Module,
):
num_contexts = attn_metadata.num_contexts
num_tokens = input_ids.shape[0]
# prepare hidden states
hidden_size_up = spec_metadata.hidden_size * len(
spec_metadata.layers_to_capture)
hidden_states = spec_metadata.hidden_states[:num_tokens, :
hidden_size_up]
hidden_states = draft_model.apply_eagle3_fc(hidden_states)
# context
input_ctx_ids = input_ids[:attn_metadata.num_ctx_tokens]
input_ids_ctx = torch.empty_like(input_ctx_ids,
dtype=torch.int32,
device="cuda")
input_ids_ctx[:-1].copy_(input_ctx_ids[1:])
input_ids_ctx[
last_tokens_idx[:num_contexts]] = accepted_tokens[:num_contexts, 0]
# generation
input_ids_gen = accepted_tokens[num_contexts:, :].flatten()
# get draft inputs
input_ids = torch.concat([input_ids_ctx, input_ids_gen], dim=0)
return {
"input_ids": input_ids,
"position_ids": position_ids,
"hidden_states": hidden_states,
"attn_metadata": attn_metadata,
"spec_metadata": spec_metadata,
}