File size: 16,362 Bytes
96b6673 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 |
import hashlib
import copy
import json
from citekit.prompt.prompt import Prompt, ALCEDocPrompt,DocPrompt,NewALCEVanillaPrompt
try:
example_results = json.load(open('res.json'))
except:
example_results = []
def weight_check(module, target):
if module.model_type == 'verifier':
test_module = copy.deepcopy(module)
test_module.last_message = True
true_target = test_module.send()
if str(target) == str(true_target) or ('output' in str(target).lower() and 'output' in str(true_target).lower()):
return 1
test_module.last_message = False
true_target = test_module.send()
if str(target) == str(true_target) or ('output' in str(target).lower() and 'output' in str(true_target).lower()):
return -1
return 0
def get_params(module):
import copy
componets = copy.deepcopy(module.prompt_maker.components if hasattr(module, 'prompt_maker') and hasattr(module.prompt_maker, 'components') else None)
if componets:
if isinstance(componets, tuple):
componets = componets[0]
componets.update(module.self_prompt)
destination = module.get_destinations()
if len(destination) == 1:
destination = str(destination[0])
elif not destination:
destination = 'N/A'
else:
destination = 'Not Available for multiple destinations'
if hasattr(module, 'iterative'):
mode = 'iterative' if module.iterative else ('parallel' if module.parallel else None)
else:
mode = None
if hasattr(module, 'if_add_output_to_head') and hasattr(module, 'head_key'):
global_prompt = 'N/A' if module.if_add_output_to_head == False else module.head_key
else:
global_prompt = 'N/A'
params = {
'type': module.model_type,
'Model': getattr(module, 'model_name', None),
'Mode': mode,
'Max Turn': getattr(module, 'max_turn', None),
'Destination': destination,
'Prompt': module.prompt_maker.make_prompt(componets) if hasattr(module,'prompt_maker') and module.prompt_maker else None,
'Global Prompt': global_prompt,
}
print(params)
non_empty_params = {}
for k, v in params.items():
if v:
non_empty_params[k] = v
return non_empty_params
class PipelineGraph:
def __init__(self, pipeline = None):
self.pipeline = pipeline
self.nodes = {'input': {}, 'output': {}}
self.edges = {} # 邻接表:{from_node: {to_node: weight}}
self.node_count = 0
if pipeline:
self.load_pipeline(pipeline)
print('FINDING ATTR:', [str(module) for module in pipeline.module])
def get_auto_node_name(self):
"""生成自动节点名。"""
self.node_count += 1
return f"Node{self.node_count}"
def update(self):
self.__init__(pipeline=self.pipeline)
def load_pipeline(self, pipeline):
"""从现有Pipeline对象加载图。"""
initial_module = pipeline.get_initial_module()
processed_nodes = []
if not initial_module:
initial_module = pipeline.llm
self.add_node(name = str(initial_module),
params = get_params(module=initial_module))
self.add_edge(from_node='input', to_node=str(initial_module), weight=0)
processed_nodes.append(initial_module)
subnodes = initial_module.get_destinations()
if not subnodes:
self.add_edge(from_node=str(initial_module), to_node='output', weight=0)
return
# print([str(subnode) for subnode in subnodes])
for subnode in subnodes:
weight = weight_check(initial_module, subnode)
self.load_node(subnode, processed_nodes=processed_nodes)
self.add_edge(from_node=str(initial_module), to_node=str(subnode), weight=0)
if hasattr(initial_module,'output_cond') and initial_module.output_cond:
weight = weight_check(initial_module, 'output')
self.add_edge(from_node=str(initial_module), to_node='output', weight=weight)
def load_node(self, module, processed_nodes = None):
if module in processed_nodes:
return
self.add_node(name = str(module),
params = get_params(module=module))
processed_nodes.append(module)
subnodes = module.get_destinations()
if not subnodes:
self.add_edge(from_node=str(module), to_node='output', weight=0)
return
for subnode in subnodes:
self.load_node(subnode, processed_nodes)
weight = weight_check(module, subnode)
self.add_edge(from_node=str(module), to_node=str(subnode), weight=weight)
if hasattr(module,'output_cond') and module.output_cond:
weight = weight_check(module, 'output')
self.add_edge(from_node=str(module), to_node='output', weight=0)
def export(self):
"""将图结构导出为字典格式,包含节点和边的信息。"""
nodes_export = {name: params.copy() for name, params in self.nodes.items()}
edges_export = []
for from_node, connections in self.edges.items():
for to_node, weight in connections.items():
edges_export.append({
'from': from_node,
'to': to_node,
'weight': weight
})
return {
'nodes': nodes_export,
'edges': edges_export
}
@classmethod
def import_from_dict(cls, data):
"""从字典数据创建新的PipelineGraph实例。"""
graph = cls() # 创建新实例,此时包含默认的input/output节点
# 清空默认节点(假设导入数据包含完整节点信息)
graph.nodes = {}
# 导入节点
for name, params in data['nodes'].items():
graph.add_node(name, params)
# 导入边
for edge in data['edges']:
graph.add_edge(edge['from'], edge['to'], edge['weight'])
return graph
def add_node(self, name, params=None):
"""添加或更新节点,参数可选。"""
if params is None:
params = {}
self.nodes[name] = params.copy()
def remove_node(self, name):
"""删除节点,自动移除相关边。不能删除input/output节点。"""
if name in ['input', 'output']:
raise ValueError("Cannot remove 'input' or 'output' nodes.")
if name not in self.nodes:
return
del self.nodes[name]
# 移除该节点作为起点的边
if name in self.edges:
del self.edges[name]
# 移除其他节点到该节点的边
for from_node in list(self.edges.keys()):
if name in self.edges[from_node]:
del self.edges[from_node][name]
# 若该节点无其他出边,删除空字典以保持整洁
if not self.edges[from_node]:
del self.edges[from_node]
def add_edge(self, from_node, to_node, weight):
"""添加或更新边,权重必须是0、-1或+1。"""
if weight not in {0, -1, 1}:
raise ValueError("Weight must be 0, -1, or 1.")
if from_node not in self.nodes:
raise KeyError(f"Node '{from_node}' does not exist.")
if to_node not in self.nodes:
raise KeyError(f"Node '{to_node}' does not exist.")
if from_node not in self.edges:
self.edges[from_node] = {}
self.edges[from_node][to_node] = weight
def remove_edge(self, from_node, to_node):
"""删除指定边。"""
if from_node in self.edges and to_node in self.edges[from_node]:
del self.edges[from_node][to_node]
# 清理空字典
if not self.edges[from_node]:
del self.edges[from_node]
def get_nodes(self):
"""返回所有节点名。"""
return list(self.nodes.keys())
def get_edges(self):
"""返回所有边的列表,形式为(from, to, weight)。"""
edges = []
for from_node in self.edges:
for to_node, weight in self.edges[from_node].items():
edges.append((from_node, to_node, weight))
return edges
def update_node_params(self, name, params):
"""更新节点的参数表。"""
if name not in self.nodes:
raise KeyError(f"Node '{name}' does not exist.")
self.nodes[name].update(params.copy())
def visualize(self, use_graphviz=True):
"""
可视化图结构:
- 默认尝试用graphviz生成矢量图(需安装graphviz)
- 若未安装graphviz,则用文本模式输出
"""
# 尝试用graphviz生成专业图表
if use_graphviz:
try:
import graphviz
except ImportError:
print("Graphviz未安装,切换到文本模式。安装命令:pip install graphviz")
return self._visualize_text()
# 创建有向图,设置排版方向为从左到右
dot = graphviz.Digraph(comment='Pipeline Graph', graph_attr={'rankdir': 'LR'})
# 添加所有节点(包含参数信息)
for node_name in self.nodes:
params = self.nodes[node_name]
# 将参数字典转换为易读字符串(例如 "size=2, color=red")
params_str = ", ".join([f"{k}={v}" for k, v in params.items()])
# 特殊样式标记input/output节点
if node_name in ('input', 'output'):
dot.node(
node_name,
label=f"{node_name}\n({params_str})" if params_str else node_name,
shape='doublecircle', # 双圆表示端点
style='filled', # 填充颜色
fillcolor='#e6f3ff' # 浅蓝色背景
)
else:
dot.node(
node_name,
label=f"{node_name}\n({params_str})" if params_str else node_name,
shape='box', # 方框表示普通节点
style='rounded,filled',# 圆角填充
fillcolor='#f0f0f0' # 浅灰色背景
)
# 添加所有带权重的边
for from_node, connections in self.edges.items():
for to_node, weight in connections.items():
dot.edge(
from_node,
to_node,
label=str(weight), # 显示权重
color='#666666', # 边颜色
fontcolor='#ff5555' # 权重文字颜色
)
# 生成并自动打开可视化文件(Mac默认用Preview打开PDF)
dot.render('pipeline_graph', view=True, cleanup=True)
return
# 文本模式回退方案
self._visualize_text()
def _visualize_text(self):
"""纯文本模式的可视化(备用方案)"""
print("="*40 + "\nPipeline Graph 文本视图\n" + "="*40)
# 列出所有节点及其参数
print("\n[节点列表]")
for node, params in self.nodes.items():
param_desc = " ".join([f"{k}={v}" for k, v in params.items()])
print(f"· {node:10} {param_desc}")
# 列出所有边及其权重
print("\n[边列表]")
if not self.edges:
print("(暂无边连接)")
else:
for from_node, connections in self.edges.items():
for to_node, weight in connections.items():
arrow = {
1: "──(+)──>", # +1权重用(+)表示
-1: "──(-)──>", # -1权重用(-)表示
0: "───────>" # 0权重无特殊标记
}[weight]
print(f"{from_node:10} {arrow} {to_node}")
print("="*40 + "\n")
def generate_html(self, results = example_results):
assert isinstance(results, str)
import json
result_path = results
results = json.load(open(results))
"""生成交互式可视化HTML,使用D3.js力导向图布局"""
nodes_data = []
for name in self.nodes:
params = self.nodes[name]
nodes_data.append({
"id": name,
"type": params.get("type", name),
"params": params
})
edges_data = []
for from_node, to_nodes in self.edges.items():
for to_node, weight in to_nodes.items():
edges_data.append({
"source": from_node,
"target": to_node,
"weight": weight
})
nodes_js = json.dumps(nodes_data)
edges_js = json.dumps(edges_data)
with open('htmls/html_templates/pipeline_template.txt', 'r', encoding='utf-8') as f:
template = f.read()
template = template.replace('<NODE_JS>', nodes_js).replace('<EDGE_JS>', edges_js).replace('<FILE_PATH>', result_path).replace('<OPTIONS>', "".join(f'<option value="{i}">Result {i+1}</option>' for i in range(len(results))))
return template
def get_nodes(self):
nodes_data = []
for name in self.nodes:
params = self.nodes[name]
nodes_data.append({
"id": name,
"type": params.get("type", name),
"params": params
})
return json.dumps(nodes_data)
def get_edges(self):
edges_data = []
for from_node, to_nodes in self.edges.items():
for to_node, weight in to_nodes.items():
edges_data.append({
"source": from_node,
"target": to_node,
"weight": weight
})
return json.dumps(edges_data)
def get_json(self):
return {
'nodes': self.get_nodes(),
'edges': self.get_edges()
}
def generate_html_embed(self, results = example_results):
assert isinstance(results, str) or isinstance(results, list)
import json
result_path = results
if isinstance(results, str):
results = json.load(open(results))
"""生成交互式可视化HTML,使用D3.js力导向图布局"""
nodes_data = []
for name in self.nodes:
params = self.nodes[name]
nodes_data.append({
"id": name,
"type": params.get("type", name),
"params": params
})
edges_data = []
for from_node, to_nodes in self.edges.items():
for to_node, weight in to_nodes.items():
edges_data.append({
"source": from_node,
"target": to_node,
"weight": weight
})
nodes_js = json.dumps(nodes_data)
edges_js = json.dumps(edges_data)
with open('htmls/html_templates/pipeline_template_emb.txt', 'r', encoding='utf-8') as f:
template = f.read()
template = template.replace('<NODE_JS>', nodes_js).replace('<EDGE_JS>', edges_js).replace('<RESULTS>', json.dumps(results)).replace('<OPTIONS>', "".join(f'<option value="{i}">Result {i+1}</option>' for i in range(len(results))))
return template
if __name__ == '__main__':
# 初始化图
graph = PipelineGraph()
# 添加节点
graph.add_node('processor', {'type': 'transform', 'rate': 0.8})
graph.add_node('validator', {'threshold': 0.5})
# 添加边
graph.add_edge('input', 'processor', 1)
graph.add_edge('processor', 'validator', -1)
graph.add_edge('validator', 'output', 0)
# 更新参数
graph.update_node_params('processor', {'rate': 1.0})
graph.visualize() |