feat(attention): 添加注意力优化器以增强提示词多样性和防止注意力退化
refactor(prompt): 使用 asyncio.gather 替代 as_completed 以提升并发性能 refactor(config): 添加注意力优化配置选项 refactor(prompt_params): 增加注意力优化开关
This commit is contained in:
@@ -89,12 +89,10 @@ def init_prompt():
|
||||
- {schedule_block}
|
||||
|
||||
## 历史记录
|
||||
### 📜 已读历史消息
|
||||
{read_history_prompt}
|
||||
|
||||
{cross_context_block}
|
||||
|
||||
### 📬 未读历史消息
|
||||
{unread_history_prompt}
|
||||
|
||||
{notice_block}
|
||||
@@ -175,12 +173,10 @@ If you need to use the search tool, please directly call the function "lpmm_sear
|
||||
{schedule_block}
|
||||
|
||||
## 历史记录
|
||||
### 📜 已读历史消息
|
||||
{read_history_prompt}
|
||||
|
||||
{cross_context_block}
|
||||
|
||||
### 📬 未读历史消息
|
||||
{unread_history_prompt}
|
||||
|
||||
{notice_block}
|
||||
@@ -858,7 +854,6 @@ class DefaultReplyer:
|
||||
# 添加标题和格式化
|
||||
notice_lines = []
|
||||
notice_lines.append("## 📢 最近的系统通知")
|
||||
notice_lines.append("")
|
||||
notice_lines.append(notice_text)
|
||||
notice_lines.append("")
|
||||
|
||||
@@ -989,7 +984,7 @@ class DefaultReplyer:
|
||||
else:
|
||||
unread_history_prompt = "暂无未读历史消息"
|
||||
|
||||
return read_history_prompt, unread_history_prompt
|
||||
return f"### 📜 已读历史消息\n{read_history_prompt}", f"### 📬 未读历史消息\n{unread_history_prompt}"
|
||||
else:
|
||||
# 回退到传统方法
|
||||
return await self._fallback_build_chat_history_prompts(message_list_before_now, target_user_id, sender)
|
||||
@@ -1091,7 +1086,7 @@ class DefaultReplyer:
|
||||
else:
|
||||
unread_history_prompt = "暂无未读历史消息"
|
||||
|
||||
return read_history_prompt, unread_history_prompt
|
||||
return f"### 📜 已读历史消息\n{read_history_prompt}", f"### 📬 未读历史消息\n{unread_history_prompt}"
|
||||
|
||||
async def build_prompt_reply_context(
|
||||
self,
|
||||
|
||||
356
src/chat/utils/attention_optimizer.py
Normal file
356
src/chat/utils/attention_optimizer.py
Normal file
@@ -0,0 +1,356 @@
|
||||
"""
|
||||
注意力优化器 - 防止提示词过度相似导致LLM注意力机制退化
|
||||
|
||||
通过轻量级随机化技术,在保持语义不变的前提下增加提示词结构多样性,
|
||||
避免短时间内重复发送高度相似的提示词导致模型回复趋同。
|
||||
|
||||
优化策略:
|
||||
1. 轻量级噪声:随机调整空白字符、换行数量
|
||||
2. 块重排:定义可交换的block组,随机调整顺序
|
||||
3. 语义变体:使用同义措辞替换固定模板文本
|
||||
"""
|
||||
|
||||
import hashlib
|
||||
import random
|
||||
import re
|
||||
from typing import Any, Literal
|
||||
|
||||
from src.common.logger import get_logger
|
||||
from src.config.config import global_config
|
||||
|
||||
logger = get_logger("attention_optimizer")
|
||||
|
||||
|
||||
class AttentionOptimizer:
|
||||
"""提示词注意力优化器"""
|
||||
|
||||
# 可交换的block组定义(组内block可以随机排序)
|
||||
# 每个组是一个列表,包含可以互换位置的block名称
|
||||
SWAPPABLE_BLOCK_GROUPS = [
|
||||
# 用户相关信息组(记忆、关系、表达习惯)
|
||||
["memory_block", "relation_info_block", "expression_habits_block"],
|
||||
# 上下文增强组(工具、知识、跨群)
|
||||
["tool_info_block", "knowledge_prompt", "cross_context_block"],
|
||||
# 元信息组(时间、身份、日程)
|
||||
["time_block", "identity_block", "schedule_block"],
|
||||
]
|
||||
|
||||
# 语义等价的文本替换模板
|
||||
# 格式: {原始文本: [替换选项1, 替换选项2, ...]}
|
||||
SEMANTIC_VARIANTS = {
|
||||
"当前时间": ["当前时间", "现在是", "此时此刻", "时间"],
|
||||
"最近的系统通知": ["最近的系统通知", "系统通知", "通知消息", "最新通知"],
|
||||
"聊天历史": ["聊天历史", "对话记录", "历史消息", "之前的对话"],
|
||||
"你的任务是": ["你的任务是", "请", "你需要", "你应当"],
|
||||
"请注意": ["请注意", "注意", "请留意", "需要注意"],
|
||||
}
|
||||
|
||||
def __init__(
|
||||
self,
|
||||
enable_noise: bool = True,
|
||||
enable_semantic_variants: bool = False,
|
||||
noise_strength: Literal["light", "medium", "heavy"] = "light",
|
||||
cache_key_suffix: str = "",
|
||||
):
|
||||
"""
|
||||
初始化注意力优化器
|
||||
|
||||
Args:
|
||||
enable_noise: 是否启用轻量级噪声注入(空白字符调整)
|
||||
enable_semantic_variants: 是否启用语义变体替换(实验性)
|
||||
noise_strength: 噪声强度 (light/medium/heavy)
|
||||
cache_key_suffix: 缓存键后缀,用于区分不同的优化配置
|
||||
"""
|
||||
self.enable_noise = enable_noise
|
||||
self.enable_semantic_variants = enable_semantic_variants
|
||||
self.noise_strength = noise_strength
|
||||
self.cache_key_suffix = cache_key_suffix
|
||||
|
||||
# 噪声强度配置
|
||||
self.noise_config = {
|
||||
"light": {"newline_range": (1, 2), "space_range": (0, 2), "indent_adjust": False},
|
||||
"medium": {"newline_range": (1, 3), "space_range": (0, 4), "indent_adjust": True},
|
||||
"heavy": {"newline_range": (1, 4), "space_range": (0, 6), "indent_adjust": True},
|
||||
}
|
||||
|
||||
|
||||
|
||||
def optimize_prompt(self, prompt_text: str, context_data: dict[str, Any]) -> str:
|
||||
"""
|
||||
优化提示词,增加结构多样性
|
||||
|
||||
Args:
|
||||
prompt_text: 原始提示词文本
|
||||
context_data: 上下文数据字典,包含各个block的内容
|
||||
|
||||
Returns:
|
||||
优化后的提示词文本
|
||||
"""
|
||||
try:
|
||||
optimized = prompt_text
|
||||
|
||||
# 步骤2: 语义变体替换(如果启用)
|
||||
if self.enable_semantic_variants:
|
||||
optimized = self._apply_semantic_variants(optimized)
|
||||
|
||||
# 步骤3: 轻量级噪声注入(如果启用)
|
||||
if self.enable_noise:
|
||||
optimized = self._inject_noise(optimized)
|
||||
|
||||
# 计算变化率
|
||||
change_rate = self._calculate_change_rate(prompt_text, optimized)
|
||||
logger.debug(f"提示词优化完成,变化率: {change_rate:.2%}")
|
||||
|
||||
return optimized
|
||||
|
||||
except Exception as e:
|
||||
logger.error(f"提示词优化失败: {e}", exc_info=True)
|
||||
return prompt_text # 失败时返回原始文本
|
||||
|
||||
def _shuffle_blocks(self, prompt_text: str, context_data: dict[str, Any]) -> str:
|
||||
"""
|
||||
重排可交换的block组
|
||||
|
||||
Args:
|
||||
prompt_text: 原始提示词
|
||||
context_data: 包含各block内容的字典
|
||||
|
||||
Returns:
|
||||
重排后的提示词
|
||||
"""
|
||||
try:
|
||||
# 对每个可交换组进行随机排序
|
||||
shuffled_context = context_data.copy()
|
||||
|
||||
for group in self.SWAPPABLE_BLOCK_GROUPS:
|
||||
# 过滤出实际存在且非空的block
|
||||
existing_blocks = [
|
||||
block for block in group if block in context_data and context_data[block]
|
||||
]
|
||||
|
||||
if len(existing_blocks) > 1:
|
||||
# 随机打乱顺序
|
||||
shuffled = existing_blocks.copy()
|
||||
random.shuffle(shuffled)
|
||||
|
||||
# 如果打乱后的顺序与原顺序不同,记录日志
|
||||
if shuffled != existing_blocks:
|
||||
logger.debug(f"重排block组: {existing_blocks} -> {shuffled}")
|
||||
|
||||
# 注意:实际的重排需要在模板格式化之前进行
|
||||
# 这里只是演示逻辑,真正的实现需要在 _format_with_context 中处理
|
||||
|
||||
# 由于block重排需要在模板构建阶段进行,这里只返回原文本
|
||||
# 真正的重排逻辑需要集成到 Prompt 类的 _format_with_context 方法中
|
||||
return prompt_text
|
||||
|
||||
except Exception as e:
|
||||
logger.error(f"Block重排失败: {e}", exc_info=True)
|
||||
return prompt_text
|
||||
|
||||
def _apply_semantic_variants(self, text: str) -> str:
|
||||
"""
|
||||
应用语义等价的文本替换
|
||||
|
||||
Args:
|
||||
text: 原始文本
|
||||
|
||||
Returns:
|
||||
替换后的文本
|
||||
"""
|
||||
try:
|
||||
result = text
|
||||
|
||||
for original, variants in self.SEMANTIC_VARIANTS.items():
|
||||
if original in result:
|
||||
# 随机选择一个变体(包括原始文本)
|
||||
replacement = random.choice(variants)
|
||||
result = result.replace(original, replacement, 1) # 只替换第一次出现
|
||||
|
||||
return result
|
||||
|
||||
except Exception as e:
|
||||
logger.error(f"语义变体替换失败: {e}", exc_info=True)
|
||||
return text
|
||||
|
||||
def _inject_noise(self, text: str) -> str:
|
||||
"""
|
||||
注入轻量级噪声(空白字符调整)
|
||||
|
||||
Args:
|
||||
text: 原始文本
|
||||
|
||||
Returns:
|
||||
注入噪声后的文本
|
||||
"""
|
||||
try:
|
||||
config = self.noise_config[self.noise_strength]
|
||||
result = text
|
||||
|
||||
# 1. 调整block之间的换行数量
|
||||
result = self._adjust_newlines(result, config["newline_range"])
|
||||
|
||||
# 2. 在某些位置添加随机空格(保持可读性)
|
||||
result = self._adjust_spaces(result, config["space_range"])
|
||||
|
||||
# 3. 调整缩进(仅在medium/heavy模式下)
|
||||
if config["indent_adjust"]:
|
||||
result = self._adjust_indentation(result)
|
||||
|
||||
return result
|
||||
|
||||
except Exception as e:
|
||||
logger.error(f"噪声注入失败: {e}", exc_info=True)
|
||||
return text
|
||||
|
||||
def _adjust_newlines(self, text: str, newline_range: tuple[int, int]) -> str:
|
||||
"""
|
||||
调整连续换行的数量
|
||||
|
||||
Args:
|
||||
text: 原始文本
|
||||
newline_range: 换行数量范围 (min, max)
|
||||
|
||||
Returns:
|
||||
调整后的文本
|
||||
"""
|
||||
# 匹配连续的换行符
|
||||
pattern = r"\n{2,}"
|
||||
|
||||
def replace_newlines(match):
|
||||
# 随机选择新的换行数量
|
||||
count = random.randint(*newline_range)
|
||||
return "\n" * count
|
||||
|
||||
return re.sub(pattern, replace_newlines, text)
|
||||
|
||||
def _adjust_spaces(self, text: str, space_range: tuple[int, int]) -> str:
|
||||
"""
|
||||
在某些位置添加随机空格
|
||||
|
||||
Args:
|
||||
text: 原始文本
|
||||
space_range: 空格数量范围 (min, max)
|
||||
|
||||
Returns:
|
||||
调整后的文本
|
||||
"""
|
||||
# 在行尾随机添加空格(不可见但会改变文本哈希)
|
||||
lines = text.split("\n")
|
||||
result_lines = []
|
||||
|
||||
for line in lines:
|
||||
if line.strip() and random.random() < 0.3: # 30%概率添加空格
|
||||
spaces = " " * random.randint(*space_range)
|
||||
result_lines.append(line + spaces)
|
||||
else:
|
||||
result_lines.append(line)
|
||||
|
||||
return "\n".join(result_lines)
|
||||
|
||||
def _adjust_indentation(self, text: str) -> str:
|
||||
"""
|
||||
微调某些行的缩进(保持语义)
|
||||
|
||||
Args:
|
||||
text: 原始文本
|
||||
|
||||
Returns:
|
||||
调整后的文本
|
||||
"""
|
||||
lines = text.split("\n")
|
||||
result_lines = []
|
||||
|
||||
for line in lines:
|
||||
# 检测列表项
|
||||
list_match = re.match(r"^(\s*)([-*•])\s", line)
|
||||
if list_match and random.random() < 0.5:
|
||||
indent = list_match.group(1)
|
||||
marker = list_match.group(2)
|
||||
# 随机调整缩进(±2个空格)
|
||||
adjust = random.choice([-2, 0, 2])
|
||||
new_indent = " " * max(0, len(indent) + adjust)
|
||||
new_line = line.replace(indent + marker, new_indent + marker, 1)
|
||||
result_lines.append(new_line)
|
||||
else:
|
||||
result_lines.append(line)
|
||||
|
||||
return "\n".join(result_lines)
|
||||
|
||||
def _calculate_change_rate(self, original: str, optimized: str) -> float:
|
||||
"""
|
||||
计算文本变化率
|
||||
|
||||
Args:
|
||||
original: 原始文本
|
||||
optimized: 优化后的文本
|
||||
|
||||
Returns:
|
||||
变化率(0-1之间的浮点数)
|
||||
"""
|
||||
if not original or not optimized:
|
||||
return 0.0
|
||||
|
||||
# 使用简单的字符差异比率
|
||||
diff_chars = sum(1 for a, b in zip(original, optimized) if a != b)
|
||||
max_len = max(len(original), len(optimized))
|
||||
|
||||
return diff_chars / max_len if max_len > 0 else 0.0
|
||||
|
||||
def get_cache_key(self, prompt_text: str) -> str:
|
||||
"""
|
||||
生成优化后提示词的缓存键
|
||||
|
||||
由于注意力优化会改变提示词内容,缓存键也需要相应调整
|
||||
|
||||
Args:
|
||||
prompt_text: 提示词文本
|
||||
|
||||
Returns:
|
||||
缓存键字符串
|
||||
"""
|
||||
# 计算文本哈希
|
||||
text_hash = hashlib.md5(prompt_text.encode()).hexdigest()[:8]
|
||||
|
||||
# 添加随机后缀,确保相似提示词有不同的缓存键
|
||||
random_suffix = random.randint(1000, 9999)
|
||||
|
||||
return f"{text_hash}_{random_suffix}_{self.cache_key_suffix}"
|
||||
|
||||
|
||||
def get_attention_optimizer_from_config() -> AttentionOptimizer:
|
||||
"""
|
||||
从全局配置创建注意力优化器实例
|
||||
|
||||
Returns:
|
||||
配置好的 AttentionOptimizer 实例
|
||||
"""
|
||||
# 从配置中读取设置(如果存在)
|
||||
config = getattr(global_config, "attention_optimization", None)
|
||||
|
||||
if not config:
|
||||
# 使用默认配置
|
||||
return AttentionOptimizer(
|
||||
enable_noise=True,
|
||||
enable_semantic_variants=False, # 实验性功能,默认关闭
|
||||
noise_strength="light",
|
||||
)
|
||||
|
||||
# config 是 Pydantic 模型对象,直接访问属性
|
||||
return AttentionOptimizer(
|
||||
enable_noise=config.enable_noise,
|
||||
enable_semantic_variants=config.enable_semantic_variants,
|
||||
noise_strength=config.noise_strength,
|
||||
)
|
||||
|
||||
|
||||
# 全局单例
|
||||
_global_optimizer: AttentionOptimizer | None = None
|
||||
|
||||
|
||||
def get_attention_optimizer() -> AttentionOptimizer:
|
||||
"""获取全局注意力优化器实例"""
|
||||
global _global_optimizer
|
||||
if _global_optimizer is None:
|
||||
_global_optimizer = get_attention_optimizer_from_config()
|
||||
return _global_optimizer
|
||||
@@ -375,6 +375,15 @@ class Prompt:
|
||||
# 这样做可以更早地组合模板,也使得`Prompt`类的职责更单一。
|
||||
result = main_formatted_prompt
|
||||
|
||||
# 步骤 4: 注意力优化(如果启用)
|
||||
# 通过轻量级随机化避免提示词过度相似导致LLM注意力退化
|
||||
if self.parameters.enable_attention_optimization:
|
||||
from src.chat.utils.attention_optimizer import get_attention_optimizer
|
||||
|
||||
optimizer = get_attention_optimizer()
|
||||
result = optimizer.optimize_prompt(result, context_data)
|
||||
logger.debug("已应用注意力优化")
|
||||
|
||||
total_time = time.time() - start_time
|
||||
logger.debug(
|
||||
f"Prompt构建完成,模式: {self.parameters.prompt_mode}, 耗时: {total_time:.2f}s"
|
||||
@@ -492,11 +501,12 @@ class Prompt:
|
||||
"expression_habits": 10.0,
|
||||
}
|
||||
|
||||
# 使用 as_completed 并发执行任务,提供更好的性能和错误处理
|
||||
# 使用 asyncio.gather 实现并发执行,提供更好的错误处理和性能
|
||||
results = [None] * len(tasks) # 预分配结果列表,保持任务顺序
|
||||
task_with_meta = []
|
||||
tasks_to_run = [] # 存储带超时的任务
|
||||
task_info = [] # 存储任务信息,用于结果处理
|
||||
|
||||
# 准备任务和元数据
|
||||
# 准备任务并创建带超时的协程
|
||||
for i, task in enumerate(tasks):
|
||||
task_name = task_names[i] if i < len(task_names) else f"task_{i}"
|
||||
task_timeout = task_timeouts.get(
|
||||
@@ -505,48 +515,41 @@ class Prompt:
|
||||
|
||||
# 检查任务是否为协程,非协程任务直接使用默认值
|
||||
if asyncio.iscoroutine(task):
|
||||
task_with_meta.append(
|
||||
(
|
||||
asyncio.wait_for(task, timeout=task_timeout),
|
||||
task_name,
|
||||
i,
|
||||
task_timeout,
|
||||
)
|
||||
)
|
||||
# 创建带超时的任务
|
||||
timeout_task = asyncio.wait_for(task, timeout=task_timeout)
|
||||
tasks_to_run.append(timeout_task)
|
||||
task_info.append({"index": i, "name": task_name, "timeout": task_timeout})
|
||||
else:
|
||||
logger.warning(
|
||||
f"任务{task_name}不是协程对象,类型: {type(task)},跳过处理"
|
||||
)
|
||||
results[i] = self._get_default_result_for_task(task_name) # type: ignore
|
||||
|
||||
# 并发执行任务,使用 as_completed 获得更好的性能
|
||||
for future in asyncio.as_completed(
|
||||
[task_meta[0] for task_meta in task_with_meta]
|
||||
):
|
||||
# 找到对应的任务元数据
|
||||
task_index = None
|
||||
task_name = None
|
||||
task_timeout = None
|
||||
# 使用 gather 并发执行所有任务,return_exceptions=True 确保单个任务失败不影响其他任务
|
||||
if tasks_to_run:
|
||||
task_results = await asyncio.gather(*tasks_to_run, return_exceptions=True)
|
||||
|
||||
for idx, (task, name, index, timeout) in enumerate(task_with_meta):
|
||||
if task == future:
|
||||
task_index = index
|
||||
task_name = name
|
||||
task_timeout = timeout
|
||||
break
|
||||
# 处理任务结果
|
||||
for i, result in enumerate(task_results):
|
||||
info = task_info[i]
|
||||
task_index = info["index"]
|
||||
task_name = info["name"]
|
||||
task_timeout = info["timeout"]
|
||||
|
||||
try:
|
||||
result = await future
|
||||
results[task_index] = result # type: ignore
|
||||
logger.debug(f"构建任务{task_name}完成 ({task_timeout}s)")
|
||||
except asyncio.TimeoutError:
|
||||
if isinstance(result, asyncio.TimeoutError):
|
||||
# 处理超时错误
|
||||
logger.warning(
|
||||
f"构建任务{task_name}超时 ({task_timeout}s),使用默认值"
|
||||
)
|
||||
results[task_index] = self._get_default_result_for_task(task_name) # type: ignore
|
||||
except Exception as e:
|
||||
logger.error(f"构建任务{task_name}失败: {e!s}")
|
||||
results[task_index] = self._get_default_result_for_task(task_name) # type: ignore
|
||||
results[task_index] = self._get_default_result_for_task(task_name)
|
||||
elif isinstance(result, Exception):
|
||||
# 处理其他异常
|
||||
logger.error(f"构建任务{task_name}失败: {result!s}")
|
||||
results[task_index] = self._get_default_result_for_task(task_name)
|
||||
else:
|
||||
# 成功完成
|
||||
results[task_index] = result
|
||||
logger.debug(f"构建任务{task_name}完成 ({task_timeout}s)")
|
||||
|
||||
# --- 步骤 3: 合并所有结果 ---
|
||||
context_data = {}
|
||||
|
||||
@@ -27,6 +27,7 @@ class PromptParameters:
|
||||
enable_relation: bool = True
|
||||
enable_cross_context: bool = True
|
||||
enable_knowledge: bool = True
|
||||
enable_attention_optimization: bool = True # 注意力优化开关
|
||||
|
||||
# 性能控制
|
||||
max_context_messages: int = 50
|
||||
|
||||
@@ -977,9 +977,6 @@ def filter_system_format_content(content: str | None) -> str:
|
||||
# [图片(描述生成失败)] 等错误格式
|
||||
cleaned_content = re.sub(r"\[图片\([^)]*\)\]", "", cleaned_content)
|
||||
|
||||
# 清理多余空格
|
||||
cleaned_content = re.sub(r"\s+", " ", cleaned_content).strip()
|
||||
|
||||
# 记录过滤操作
|
||||
if cleaned_content != original_content.strip():
|
||||
logger.info(
|
||||
|
||||
@@ -13,6 +13,7 @@ from src.common.logger import get_logger
|
||||
from src.config.config_base import ValidatedConfigBase
|
||||
from src.config.official_configs import (
|
||||
AffinityFlowConfig,
|
||||
AttentionOptimizationConfig,
|
||||
BotConfig,
|
||||
ChatConfig,
|
||||
ChineseTypoConfig,
|
||||
@@ -391,6 +392,9 @@ class Config(ValidatedConfigBase):
|
||||
tool: ToolConfig = Field(..., description="工具配置")
|
||||
debug: DebugConfig = Field(..., description="调试配置")
|
||||
custom_prompt: CustomPromptConfig = Field(..., description="自定义提示配置")
|
||||
attention_optimization: AttentionOptimizationConfig = Field(
|
||||
default_factory=lambda: AttentionOptimizationConfig(), description="注意力优化配置"
|
||||
)
|
||||
voice: VoiceConfig = Field(..., description="语音配置")
|
||||
permission: PermissionConfig = Field(..., description="权限配置")
|
||||
command: CommandConfig = Field(..., description="命令系统配置")
|
||||
|
||||
@@ -531,6 +531,16 @@ class CustomPromptConfig(ValidatedConfigBase):
|
||||
planner_custom_prompt_content: str = Field(default="", description="规划器自定义提示词内容")
|
||||
|
||||
|
||||
class AttentionOptimizationConfig(ValidatedConfigBase):
|
||||
"""注意力优化配置类 - 防止提示词过度相似导致LLM注意力退化"""
|
||||
|
||||
enable_noise: bool = Field(default=True, description="启用轻量级噪声注入(空白字符调整)")
|
||||
enable_semantic_variants: bool = Field(default=False, description="启用语义变体替换(实验性功能)")
|
||||
noise_strength: Literal["light", "medium", "heavy"] = Field(
|
||||
default="light", description="噪声强度: light(轻量) | medium(中等) | heavy(强力)"
|
||||
)
|
||||
|
||||
|
||||
class ResponsePostProcessConfig(ValidatedConfigBase):
|
||||
"""回复后处理配置类"""
|
||||
|
||||
|
||||
@@ -1,5 +1,5 @@
|
||||
[inner]
|
||||
version = "7.6.8"
|
||||
version = "7.6.9"
|
||||
|
||||
#----以下是给开发人员阅读的,如果你只是部署了MoFox-Bot,不需要阅读----
|
||||
#如果你想要修改配置文件,请递增version的值
|
||||
@@ -348,6 +348,12 @@ reaction = "请按照以下模板造句:[n]是这样的,xx只要xx就可以
|
||||
image_prompt = "请用中文描述这张图片的内容。如果有文字,请把文字描述概括出来,请留意其主题,直观感受,输出为一段平文本,最多30字,请注意不要分点,就输出一段文本"
|
||||
planner_custom_prompt_content = "" # 决策器自定义提示词内容,如果这里没有内容则不生效
|
||||
|
||||
# 注意力优化配置 - 防止提示词过度相似导致LLM注意力退化
|
||||
[attention_optimization]
|
||||
enable_noise = true # 启用轻量级噪声注入(空白字符调整)
|
||||
enable_semantic_variants = false # 启用语义变体替换(实验性功能)
|
||||
noise_strength = "light" # 噪声强度: "light"(轻量) | "medium"(中等) | "heavy"(强力),推荐使用light
|
||||
|
||||
[response_post_process]
|
||||
enable_response_post_process = true # 是否启用回复后处理,包括错别字生成器,回复分割器
|
||||
|
||||
|
||||
Reference in New Issue
Block a user