File size: 7,830 Bytes
1f3039b | 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 | #!/usr/bin/env python3
"""ShopRLVE Training Script -- run rollouts with synthetic data.
Usage:
python scripts/train.py --collection C1 --episodes 100 --seed 42
python scripts/train.py --collection C8 --episodes 1000 --model Qwen/Qwen2.5-1.5B-Instruct
Creates a ShopRLVEEnv with a synthetic catalog and runs batch rollouts
with DummyModelFn (or a real model if ``--model`` is specified). Prints
per-env stats and saves results to JSON.
"""
from __future__ import annotations
import argparse
import json
import logging
import sys
import time
from collections import defaultdict
from pathlib import Path
from typing import Any
import numpy as np
from rich.console import Console
from rich.table import Table
from shop_rlve.server.openenv import ShopRLVEEnv
from shop_rlve.training.collections import COLLECTIONS, get_collection
from shop_rlve.training.rollout import DummyModelFn, RolloutResult, run_rollout
logger = logging.getLogger(__name__)
console = Console()
def _build_dummy_model(
env_ids: list[str],
product_ids: list[str],
seed: int,
) -> DummyModelFn:
"""Build a DummyModelFn that cycles through env_ids.
Args:
env_ids: List of environment IDs.
product_ids: Product IDs to recommend from.
seed: Random seed.
Returns:
DummyModelFn instance.
"""
# The dummy model needs an env_id; for multi-env collections we default
# to the first env since DummyModelFn uses it for the answer schema.
return DummyModelFn(
env_id=env_ids[0],
product_ids=product_ids,
seed=seed,
)
def run_training(
collection: str,
n_episodes: int,
seed: int,
output_path: str | None,
) -> dict[str, Any]:
"""Run training rollouts and return aggregated results.
Args:
collection: Collection name (C1, C2, C4, C8).
n_episodes: Number of episodes to run.
seed: Master random seed.
output_path: Optional path to save JSON results.
Returns:
Dict with per-env stats and overall stats.
"""
console.print(
f"[bold cyan]Training[/bold cyan] collection=[bold]{collection}[/bold] "
f"episodes=[bold]{n_episodes}[/bold] seed=[bold]{seed}[/bold]"
)
t0 = time.monotonic()
env = ShopRLVEEnv(collection=collection, seed=seed)
env.dump_dir = "" # Disable trace dumping
env_ids = get_collection(collection)
product_ids = [p.id for p in env._products[:30]]
# Per-env accumulators
per_env: dict[str, list[RolloutResult]] = defaultdict(list)
for i in range(n_episodes):
ep_env_id = env_ids[i % len(env_ids)]
ep_seed = seed + i
# Create a fresh DummyModelFn per episode with the correct env_id
dummy = DummyModelFn(
env_id=ep_env_id,
product_ids=product_ids,
seed=ep_seed,
)
result = run_rollout(
env=env,
model_fn=dummy,
env_id=ep_env_id,
seed=ep_seed,
)
per_env[ep_env_id].append(result)
if (i + 1) % max(1, n_episodes // 10) == 0:
console.print(f" [{i + 1}/{n_episodes}] episodes completed")
elapsed = time.monotonic() - t0
# Aggregate stats
stats_table = Table(
title=f"Training Results: {collection} ({n_episodes} episodes, {elapsed:.1f}s)",
show_header=True,
header_style="bold cyan",
)
stats_table.add_column("Env", style="cyan", width=10)
stats_table.add_column("Episodes", justify="right", width=10)
stats_table.add_column("Mean Reward", justify="right", width=12)
stats_table.add_column("Std Reward", justify="right", width=12)
stats_table.add_column("Success Rate", justify="right", width=13)
stats_table.add_column("Mean Turns", justify="right", width=11)
all_results: dict[str, Any] = {
"collection": collection,
"n_episodes": n_episodes,
"seed": seed,
"elapsed_seconds": elapsed,
"per_env": {},
}
total_rewards: list[float] = []
total_correct = 0
for eid in env_ids:
results = per_env.get(eid, [])
if not results:
continue
rewards = [r.reward for r in results]
turns = [r.turns for r in results]
correct = sum(1 for r in results if r.is_correct)
rewards_arr = np.array(rewards, dtype=np.float64)
mean_r = float(np.mean(rewards_arr))
std_r = float(np.std(rewards_arr))
mean_t = float(np.mean(turns))
success = correct / len(results)
total_rewards.extend(rewards)
total_correct += correct
env_stats = {
"n_episodes": len(results),
"mean_reward": mean_r,
"std_reward": std_r,
"success_rate": success,
"mean_turns": mean_t,
}
all_results["per_env"][eid] = env_stats
stats_table.add_row(
eid,
str(len(results)),
f"{mean_r:.4f}",
f"{std_r:.4f}",
f"{success:.2%}",
f"{mean_t:.2f}",
)
# Overall row
if total_rewards:
overall_arr = np.array(total_rewards, dtype=np.float64)
stats_table.add_row(
"[bold]TOTAL[/bold]",
str(n_episodes),
f"[bold]{float(np.mean(overall_arr)):.4f}[/bold]",
f"{float(np.std(overall_arr)):.4f}",
f"[bold]{total_correct / n_episodes:.2%}[/bold]",
"",
)
console.print(stats_table)
# Difficulty progression
engine_state = env.adaptive_engine.get_all_states()
diff_table = Table(title="Difficulty Progression", show_header=True)
diff_table.add_column("Env", style="cyan", width=10)
diff_table.add_column("Low", justify="right", width=6)
diff_table.add_column("High", justify="right", width=6)
for eid in env_ids:
state = engine_state.get(eid)
if state:
diff_table.add_row(eid, str(state.low), str(state.high))
console.print(diff_table)
# Save results
if output_path:
p = Path(output_path)
p.parent.mkdir(parents=True, exist_ok=True)
with open(p, "w") as f:
json.dump(all_results, f, indent=2, default=str)
console.print(f"\n[dim]Results saved to {output_path}[/dim]")
env.close()
return all_results
def main() -> None:
"""Entry point."""
parser = argparse.ArgumentParser(
description="ShopRLVE Training Script",
formatter_class=argparse.RawDescriptionHelpFormatter,
epilog=__doc__,
)
parser.add_argument(
"--collection",
type=str,
default="C1",
choices=sorted(COLLECTIONS.keys()),
help="Environment collection (default: C1)",
)
parser.add_argument(
"--episodes", type=int, default=100, help="Number of episodes"
)
parser.add_argument("--seed", type=int, default=42, help="Random seed")
parser.add_argument(
"--model",
type=str,
default=None,
help="Model name (default: DummyModelFn)",
)
parser.add_argument(
"--config",
type=str,
default=None,
help="Path to config YAML",
)
parser.add_argument(
"--output",
type=str,
default="results/train_results.json",
help="Output JSON path",
)
args = parser.parse_args()
if args.model is not None:
console.print(
f"[yellow]Note: --model {args.model} specified but real model "
f"integration is not yet implemented. Using DummyModelFn.[/yellow]"
)
run_training(
collection=args.collection,
n_episodes=args.episodes,
seed=args.seed,
output_path=args.output,
)
if __name__ == "__main__":
main()
|