File size: 10,211 Bytes
685d968 |
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 |
"""
Balanced Dataset Generation with Concurrent API Calls
Generates 10 items simultaneously per batch for faster generation.
"""
import json
import random
import time
import sys
import asyncio
import os
from typing import List, Dict, Any, Optional
from datetime import datetime
from concurrent.futures import ThreadPoolExecutor
import cohere
from dotenv import load_dotenv
load_dotenv()
# Target counts per category (balanced)
CATEGORY_TARGETS = {
"company.brand_core": 77,
"company.strategic_signatures": 77,
"company.knowledge_artifacts": 77,
"company.business_priorities": 77,
"company.tools_config": 77,
"company.performance_context": 77,
"user.communication_style": 77,
"user.strategic_approach": 77,
"user.role_context": 77,
"user.workflow_patterns": 77,
"user.session_history": 77,
"user.interaction_preferences": 77,
"none": 77,
}
CATEGORY_EXAMPLES = {
"company.brand_core": {
"signals": ["brand voice is warm", "primary color is #2563EB", "never use jargon", "tagline is..."],
},
"company.strategic_signatures": {
"signals": ["always prioritize retention", "80/20 rule", "never launch without testing"],
},
"company.knowledge_artifacts": {
"signals": ["style guide says", "playbook recommends", "SOP for launches", "template includes"],
},
"company.business_priorities": {
"signals": ["Q4 focus is", "this quarter's target", "holiday campaign", "prioritizing APAC"],
},
"company.tools_config": {
"signals": ["Slack webhook URL", "HubSpot sync", "API key is", "Zapier integration"],
},
"company.performance_context": {
"signals": ["24% open rate", "CTR improved by", "retrospective showed", "conversion dropped"],
},
"user.communication_style": {
"signals": ["prefer bullet points", "keep it under 200 words", "casual tone", "data-driven"],
},
"user.strategic_approach": {
"signals": ["prioritize speed over perfection", "test fast fail fast", "customer feedback"],
},
"user.role_context": {
"signals": ["As VP of Marketing", "report to CMO", "budget authority up to", "manage team of"],
},
"user.workflow_patterns": {
"signals": ["review drafts Monday", "don't send Friday", "async via Slack", "weekly sync Tuesday"],
},
"user.session_history": {
"signals": ["as we discussed yesterday", "continuing from last", "proposal we started"],
},
"user.interaction_preferences": {
"signals": ["push back on my ideas", "give me options", "be direct", "ask clarifying questions"],
},
"none": {
"signals": ["what time is meeting", "checking status", "confirming receipt", "quick question"],
},
}
class BalancedAsyncGenerator:
def __init__(self):
self.api_key = os.getenv("COHERE_API_KEY")
if not self.api_key:
raise ValueError("COHERE_API_KEY not found")
self.client = cohere.ClientV2(api_key=self.api_key)
self.model = "command-r-plus-08-2024"
self.executor = ThreadPoolExecutor(max_workers=10)
def _extract_text(self, response) -> Optional[str]:
if not response or not getattr(response, "message", None):
return None
blocks = getattr(response.message, "content", []) or []
for block in blocks:
text = getattr(block, "text", None)
if isinstance(text, str) and text.strip():
return text
return None
def _generate_sync(self, category: str) -> Optional[Dict]:
"""Synchronous generation for a single category."""
signals = CATEGORY_EXAMPLES.get(category, {}).get("signals", [])
signals_text = "\n".join(f"- {s}" for s in signals[:4])
if category == "none":
prompt = f"""Generate a marketing conversation with NO long-term memory value.
Transactional, vague, or temporary only. Examples: status check, scheduling, confirming.
4-6 turns, no greetings, start mid-conversation.
OUTPUT (JSON only):
{{"scenario_id": "none_{random.randint(100,999)}", "conversation": [{{"role": "user", "content": "..."}}, {{"role": "assistant", "content": "..."}}], "labels": {{"categories": ["none"], "persistence_horizon": "short", "memory_scope": "none", "rationale": "..."}}, "metadata": {{"primary_category": "none", "turn_count": 4}}}}"""
else:
prompt = f"""Generate a marketing conversation demonstrating: {category}
SIGNALS FOR THIS CATEGORY:
{signals_text}
REQUIREMENTS:
1. MUST contain clear signals for {category}
2. 4-6 turns, no greetings, start mid-conversation
3. Include specific details (names, numbers, dates)
CRITICAL: categories array MUST include "{category}"
OUTPUT (JSON only):
{{"scenario_id": "{category.replace('.', '_')}_{random.randint(100,999)}", "conversation": [{{"role": "user", "content": "..."}}, {{"role": "assistant", "content": "..."}}], "labels": {{"categories": ["{category}"], "persistence_horizon": "long", "memory_scope": "company", "rationale": "..."}}, "metadata": {{"primary_category": "{category}", "turn_count": 4}}}}"""
try:
response = self.client.chat(
messages=[{"role": "user", "content": prompt}],
temperature=0.7,
model=self.model,
response_format={"type": "json_object"}
)
content = self._extract_text(response)
if not content:
return None
if content.startswith("```json"):
content = content[7:]
if content.endswith("```"):
content = content[:-3]
data = json.loads(content.strip())
# Validate target category is present
categories = data.get("labels", {}).get("categories", [])
if category.lower() not in [c.lower() for c in categories]:
return None
# Clean: Remove "none" if other categories exist
if len(categories) > 1 and "none" in [c.lower() for c in categories]:
data["labels"]["categories"] = [c for c in categories if c.lower() != "none"]
return data
except Exception as e:
return None
async def generate_batch(self, categories: List[str]) -> List[Dict]:
"""Generate a batch of items concurrently."""
loop = asyncio.get_event_loop()
tasks = [
loop.run_in_executor(self.executor, self._generate_sync, cat)
for cat in categories
]
results = await asyncio.gather(*tasks, return_exceptions=True)
return [r for r in results if isinstance(r, dict)]
async def run_balanced_generation_async():
"""Run balanced generation with concurrent batches."""
generator = BalancedAsyncGenerator()
timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
output_file = f"synthetic_data/balanced_dataset_{timestamp}.jsonl"
# Track progress per category
category_counts = {cat: 0 for cat in CATEGORY_TARGETS}
all_data = []
print("=" * 70, flush=True)
print("BALANCED CONCURRENT DATASET GENERATION", flush=True)
print("=" * 70, flush=True)
print(f"Target per category: 77", flush=True)
print(f"Total categories: {len(CATEGORY_TARGETS)}", flush=True)
print(f"Expected total: {77 * len(CATEGORY_TARGETS)}", flush=True)
print(f"Batch size: 10 concurrent requests", flush=True)
print(flush=True)
batch_num = 0
while True:
# Find categories that still need examples
needed = []
for cat, target in CATEGORY_TARGETS.items():
remaining = target - category_counts[cat]
needed.extend([cat] * min(remaining, 2)) # Up to 2 per category per batch
if not needed:
break
# Take up to 10 for this batch
batch_categories = needed[:10]
batch_num += 1
print(f"\n[Batch {batch_num}] Generating {len(batch_categories)} items...", flush=True)
results = await generator.generate_batch(batch_categories)
# Process results
for result in results:
if result:
primary = result.get("metadata", {}).get("primary_category") or \
result.get("labels", {}).get("categories", ["unknown"])[0]
if primary in category_counts:
category_counts[primary] += 1
all_data.append(result)
# Save incrementally
with open(output_file, "a") as f:
f.write(json.dumps(result) + "\n")
# Progress report
total_done = sum(category_counts.values())
total_target = sum(CATEGORY_TARGETS.values())
print(f" Success: {len(results)}/{len(batch_categories)} | Total: {total_done}/{total_target}", flush=True)
# Show category progress every 10 batches
if batch_num % 10 == 0:
print("\n Category Progress:", flush=True)
for cat, count in sorted(category_counts.items()):
target = CATEGORY_TARGETS[cat]
bar = "█" * (count * 20 // target) + "░" * (20 - count * 20 // target)
print(f" {cat:<35} [{bar}] {count}/{target}", flush=True)
# Rate limit: wait 3 seconds between batches
await asyncio.sleep(3)
# Final summary
print("\n" + "=" * 70, flush=True)
print("GENERATION COMPLETE", flush=True)
print("=" * 70, flush=True)
print(f"\nFinal Distribution:", flush=True)
for cat, count in sorted(category_counts.items(), key=lambda x: -x[1]):
pct = count / len(all_data) * 100 if all_data else 0
print(f" {cat:<40} {count:>4} ({pct:.1f}%)", flush=True)
print(f"\nTotal examples: {len(all_data)}", flush=True)
print(f"Output file: {output_file}", flush=True)
return output_file
if __name__ == "__main__":
asyncio.run(run_balanced_generation_async())
|