Beta spec-overlap for EAGLE (#11398)

Co-authored-by: Lianmin Zheng <15100009+merrymercy@users.noreply.github.com>
Co-authored-by: Hanming Lu <69857889+hanming-lu@users.noreply.github.com>
This commit is contained in:
Liangsheng Yin
2025-10-12 11:02:22 +08:00
committed by GitHub
parent 47c606d3dc
commit 20a6c0a63d
21 changed files with 1567 additions and 108 deletions

View File

@@ -4,7 +4,7 @@ import copy
import dataclasses
import logging
from dataclasses import replace
from typing import TYPE_CHECKING, Dict, List, Optional, Sequence, Union
from typing import TYPE_CHECKING, Dict, List, Optional, Sequence
import torch
@@ -30,12 +30,12 @@ from sglang.srt.model_executor.forward_batch_info import (
)
from sglang.srt.operations import execute_operations, execute_overlapped_operations
from sglang.srt.operations_strategy import OperationsStrategy
from sglang.srt.speculative.eagle_info import EagleDraftInput, EagleVerifyInput
from sglang.srt.speculative.spec_info import SpecInput
from sglang.srt.utils import BumpAllocator, empty_context, get_bool_env_var, is_hip
if TYPE_CHECKING:
from sglang.srt.layers.moe.token_dispatcher import DispatchOutput
from sglang.srt.speculative.eagle_info import EagleVerifyInput
_is_hip = is_hip()