v1.0.1; add tools

UPDATE.md

@@ -1,3 +1,6 @@
+# 20260117
+## v1.0.1
+1. 加入日志数据读取脚本`logs_merge.py`，压缩日志`logs_compress.py`工具
 # 20260113
 ## v1.0.0
 1. 确定最大训练步数为120k

setup.py

@@ -2,7 +2,7 @@ from setuptools import find_packages
 from distutils.core import setup
 
 setup(name='go2_rl_gym',
-      version='1.0.0',
+      version='1.0.1',
       author='Wu Tianyang',
       license="MIT",
       packages=find_packages(),

tools/logs_compress.py

@@ -0,0 +1,79 @@
+# -*- coding: utf-8 -*-
+'''
+@File    : batch_compress.py
+@Time    : 2026/01/09 17:49:54
+@Author  : wty-yy, Gemini 3
+@Version : 1.0
+@Blog    : https://wty-yy.github.io/
+@Desc    : None
+'''
+import os
+import subprocess
+from pathlib import Path
+
+def smart_compress(logs_root):
+    logs_path = Path(logs_root).resolve()
+    if not logs_path.exists():
+        print(f"❌ 找不到目录: {logs_root}")
+        return
+
+    # 遍历 logs 下的第一层子目录 (cts_vanilla, go2_moe_cts 等)
+    projects = [d for d in logs_path.iterdir() if d.is_dir()]
+
+    for project in projects:
+        project_name = project.name
+        print(f"\n🚀 正在处理项目: {project_name}")
+
+        # 1. 搜寻需要包含的内容
+        include_items = []
+
+        # 检查是否存在 exported 文件夹
+        if (project / "exported").exists():
+            include_items.append("exported")
+
+        # 搜寻所有包含 tfevents 的文件夹 (如 Jan04_15-55-59_)
+        # 我们寻找 events 文件，然后取其父目录名（相对于项目根目录）
+        event_folders = set()
+        for event_file in project.rglob("events.out.tfevents*"):
+            # 计算相对于项目根目录的路径
+            relative_folder = event_file.parent.relative_to(project)
+            event_folders.add(str(relative_folder))
+        
+        include_items.extend(list(event_folders))
+
+        if not include_items:
+            print(f"⚠️  跳过 {project_name}: 未发现符合条件的训练数据或 exported 文件夹")
+            continue
+
+        # 2. 构造压缩命令
+        output_zst = logs_path / f"{project_name}.tar.zst"
+        
+        # 命令解释:
+        # -C: 切换到项目所在目录，这样压缩包内的路径不会带一堆无用的父级前缀
+        # --exclude='*.pt': 显式排除所有模型权重文件
+        # -T0: zstd 开启全核并行
+        tar_cmd = [
+            "tar",
+            "-I", "zstd -T0 -3",
+            "-C", str(project), 
+            "--exclude=*.pt",
+            "--exclude=*.pth", # 预防万一有 .pth
+            "-cf", str(output_zst)
+        ] + include_items
+
+        print(f"📦 正在打包 (已排除 .pt 文件)...")
+        
+        try:
+            # 执行压缩
+            subprocess.run(tar_cmd, check=True)
+            
+            # 统计结果
+            final_size = output_zst.stat().st_size / (1024 * 1024)
+            print(f"✅ 完成! 压缩包: {output_zst.name} ({final_size:.2f} MB)")
+        except subprocess.CalledProcessError as e:
+            print(f"❌ {project_name} 压缩失败: {e}")
+
+if __name__ == "__main__":
+    # 执行目录
+    TARGET_LOGS_DIR = "./logs"
+    smart_compress(TARGET_LOGS_DIR)

tools/logs_merge.py

@@ -0,0 +1,121 @@
+import time
+import yaml
+import argparse
+import pandas as pd
+from tqdm import tqdm
+from pathlib import Path
+from collections import defaultdict
+
+from tensorboard.compat.proto import event_pb2
+from tensorboard.backend.event_processing import event_file_loader
+
+PATH_PARENT = Path(__file__).parent.resolve()
+BASE_COLUMNS = [
+    'it', 'benchmark',
+    'lin_vel_err',
+    'ang_vel_err',
+    'dof_limits',
+    'dof_power',
+    'orientation_stability',
+    'torque_smoothness',
+    'flat', 'wave', 'obstacle',
+    'slope_fd', 'slope_bd',
+    'stairs_fd', 'stairs_bd',
+    'terrain_level'
+]
+
+def fast_read(event_file_path, tag_names):
+    loader = event_file_loader.RawEventFileLoader(event_file_path)
+    steps = []
+    values = []
+
+    for raw_event in loader.Load():
+        event = event_pb2.Event.FromString(raw_event)
+        
+        if event.HasField('summary'):
+            for value in event.summary.value:
+                if value.tag in tag_names:
+                    steps.append(event.step)
+                    values.append(value.simple_value)
+                    
+    return pd.DataFrame({'step': steps, 'value': values})
+
+class Collector:
+    def __init__(self, log_dirs):
+        self.log_dirs = Path(log_dirs)
+        assert self.log_dirs.exists(), f"Log directory {log_dirs} does not exist."
+        assert self.log_dirs.is_dir(), f"{log_dirs} is not a directory."
+        alg_name = self.log_dirs.parent.name
+        date_str = self.log_dirs.name
+        self.output_dir = PATH_PARENT / f"{alg_name}_{date_str}"
+        if self.output_dir.exists():
+            s = input(f"[Warning] Output directory {self.output_dir} already exists, press Enter to continue and overwrite or type 'q' to quit...")
+            if s.lower() == 'q':
+                exit(0)
+        self.output_dir.mkdir(parents=True, exist_ok=True)
+        self.output_csv = self.output_dir / f"{alg_name}_{date_str}.csv"
+        self.datas = defaultdict(list)
+
+
+        self.output_tb = self.output_dir / "tb.csv"
+        if self.output_tb.exists():
+            print(f"Loading existing tensorboard data from {self.output_tb}")
+            self.tb_df = pd.read_csv(self.output_tb)
+        else:
+            start_time = time.time()
+            print(f"Start reading tensorboard events at {time.ctime(start_time)}")
+            self.tb_df = fast_read(str(self.log_dirs.glob("events.out.tfevents.*").__next__()), ['Terrain/terrain_level_all', 'Episode/terrain_level_all'])
+            print(f"Finished reading tensorboard events in {time.time() - start_time:.2f} seconds.")
+            self.tb_df.to_csv(self.output_tb, index=False)
+            print(f"Saved tensorboard data to {self.output_tb}")
+    
+    def collect(self):
+        robogauge_results_path = self.log_dirs / "robogauge_results"
+        results = list(robogauge_results_path.glob("*.yaml"))
+        results = sorted(results, key=lambda x: int(x.stem.split("_")[-1]))
+        for result in tqdm(results):
+            it = int(result.stem.split("_")[-1])
+            with open(result, 'r', encoding='utf-8') as f:
+                data = yaml.safe_load(f)
+            self.datas['it'].append(it)
+            self.datas['benchmark'].append(float(data['benchmark_score']))
+            for metric_name in [
+                'lin_vel_err',
+                'ang_vel_err',
+                'dof_limits',
+                'dof_power',
+                'orientation_stability',
+                'torque_smoothness'
+            ]:
+                self.datas[f'{metric_name}_mean'].append(float(data['summary'][metric_name]['mean'].split(' ')[0]))
+                self.datas[f'{metric_name}_mean@25'].append(float(data['summary'][metric_name]['mean@25'].split(' ')[0]))
+                self.datas[f'{metric_name}_mean@50'].append(float(data['summary'][metric_name]['mean@50'].split(' ')[0]))
+            for terrain_name in [
+                'flat',
+                'wave',
+                'obstacle',
+                'slope_fd',
+                'slope_bd',
+                'stairs_fd',
+                'stairs_bd',
+            ]:
+                if data['robust_score'][terrain_name] is None:
+                    self.datas[f'{terrain_name}_mean'].append(0.0)
+                    self.datas[f'{terrain_name}_mean@25'].append(0.0)
+                    self.datas[f'{terrain_name}_mean@50'].append(0.0)
+                    continue
+                self.datas[f'{terrain_name}_mean'].append(float(data['robust_score'][terrain_name]['mean']))
+                self.datas[f'{terrain_name}_mean@25'].append(float(data['robust_score'][terrain_name]['mean@25']))
+                self.datas[f'{terrain_name}_mean@50'].append(float(data['robust_score'][terrain_name]['mean@50']))
+            
+            self.datas['terrain_level'].append(float(self.tb_df[self.tb_df['step'] == it]['value'].values[0]))
+        df = pd.DataFrame(self.datas)
+        df.to_csv(self.output_csv, index=False)
+        print(f"Saved merged results to {self.output_csv}")
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--log-dirs")
+    args = parser.parse_args()
+    collector = Collector(args.log_dirs)
+    collector.collect()