Spaces:
Running
Running
File size: 5,580 Bytes
30fabb4 |
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 |
import re
from dataclasses import dataclass
from typing import List, Optional, Dict
from collections import Counter
from cot_reasoning import CoTStep, CoTResponse, VisualizationConfig, wrap_text
@dataclass
class SCRPath:
"""Data class representing a single self-consistency reasoning path"""
path_id: int
steps: List[CoTStep]
answer: Optional[str] = None
@dataclass
class SCRResponse:
"""Data class representing a complete self-consistency response"""
question: str
paths: List[SCRPath]
final_answer: Optional[str] = None
vote_counts: Optional[Dict[str, int]] = None
def parse_scr_response(response_text: str, question: str, num_paths: int = 5) -> SCRResponse:
"""
Parse self-consistency response text to extract multiple reasoning paths and answers.
Args:
response_text: The raw response from the API containing multiple paths
question: The original question
num_paths: Expected number of reasoning paths
Returns:
SCRResponse object containing all paths and aggregated answer
"""
# Split the response into individual paths
path_pattern = r'Path\s+(\d+):(.*?)(?=Path\s+\d+:|$)'
path_matches = re.finditer(path_pattern, response_text, re.DOTALL)
paths = []
answers = []
for match in path_matches:
path_id = int(match.group(1))
path_content = match.group(2).strip()
# Extract steps for this path
step_pattern = r'<step number="(\d+)">\s*(.*?)\s*</step>'
steps = []
for step_match in re.finditer(step_pattern, path_content, re.DOTALL):
number = int(step_match.group(1))
content = step_match.group(2).strip()
steps.append(CoTStep(number=number, content=content))
# Extract answer for this path
answer_pattern = r'<answer>\s*(.*?)\s*</answer>'
answer_match = re.search(answer_pattern, path_content, re.DOTALL)
answer = answer_match.group(1).strip() if answer_match else None
if answer:
answers.append(answer)
# Sort steps by number
steps.sort(key=lambda x: x.number)
paths.append(SCRPath(path_id=path_id, steps=steps, answer=answer))
# Determine final answer through voting
vote_counts = Counter(answers)
final_answer = vote_counts.most_common(1)[0][0] if vote_counts else None
return SCRResponse(
question=question,
paths=paths,
final_answer=final_answer,
vote_counts=dict(vote_counts)
)
def create_mermaid_diagram(scr_response: SCRResponse, config: VisualizationConfig) -> str:
"""
Convert self-consistency paths to Mermaid diagram.
Args:
scr_response: SCRResponse object containing multiple reasoning paths
config: VisualizationConfig for text formatting
Returns:
Mermaid diagram markup as a string
"""
diagram = ['<div class="mermaid">', 'graph TD']
# Add question node
question_content = wrap_text(scr_response.question, config)
diagram.append(f' Q["{question_content}"]')
# Process each path
for path in scr_response.paths:
path_id = f'P{path.path_id}'
# Add path label
diagram.append(f' {path_id}["Path {path.path_id}"]')
diagram.append(f' Q --> {path_id}')
# Add steps for this path
prev_node = path_id
for step in path.steps:
content = wrap_text(step.content, config)
node_id = f'P{path.path_id}S{step.number}'
diagram.append(f' {node_id}["{content}"]')
diagram.append(f' {prev_node} --> {node_id}')
prev_node = node_id
# Add path answer
if path.answer:
answer_content = wrap_text(path.answer, config)
answer_id = f'A{path.path_id}'
diagram.append(f' {answer_id}["{answer_content}"]')
diagram.append(f' {prev_node} --> {answer_id}')
# Add final answer with vote counts
if scr_response.final_answer and scr_response.vote_counts:
vote_info = [f"{ans}: {count} votes" for ans, count in scr_response.vote_counts.items()]
final_content = wrap_text(
f"Final Answer (by voting):\\n{scr_response.final_answer}\\n\\n" +
"Vote Distribution:\\n" + "\\n".join(vote_info),
config
)
diagram.append(f' F["{final_content}"]')
# Connect all path answers to final answer
for path in scr_response.paths:
if path.answer:
diagram.append(f' A{path.path_id} --> F')
# Add styles
diagram.extend([
' classDef default fill:#f9f9f9,stroke:#333,stroke-width:2px;',
' classDef question fill:#e3f2fd,stroke:#1976d2,stroke-width:2px;',
' classDef path fill:#fff3e0,stroke:#f57c00,stroke-width:2px;',
' classDef answer fill:#d4edda,stroke:#28a745,stroke-width:2px;',
' classDef final fill:#d4edda,stroke:#28a745,stroke-width:2px;',
' class Q question;',
' class F final;'
])
# Apply path style to all path nodes
for path in scr_response.paths:
diagram.append(f' class P{path.path_id} path;')
# Apply answer style to all answer nodes
for path in scr_response.paths:
if path.answer:
diagram.append(f' class A{path.path_id} answer;')
diagram.append('</div>')
return '\n'.join(diagram) |