cyy/RoboGauge
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

v0.1.16

Browse Source
This commit is contained in:
wty-yy
2025-12-27 01:44:09 +08:00
parent dad65a4e27
commit 0a3985445a
5 changed files with 216 additions and 35 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
UPDATE.md
View File

@@ -1,8 +1,9 @@
# UPDATE
## 20251226
### v0.1.16
1. 基本完成StressPipeline, 但是绘制进度信息还有问题
1. 基本完成StressPipeline, 加入绘制进度条的线程, 其他进程通过Queue更新主进程的进度条
2. wave地形的穿模判定非常容易触发, 加入最多穿模判定重启次数为1次, 超过该次数后穿模就不再自动重启了
3. 当全部地形等级均失败时, 需要将全部metric都按0计算
FIX Bugs: (level_results会存储到MultiPipeline下, 因为MultiPipeline不是子进程启动的, 会覆盖LevelPipeline的Logger冲突) 通过创建新的Logger实现
## 20251225

7
robogauge/tasks/pipeline/level_pipeline.py
View File

@@ -11,31 +11,36 @@ import yaml
from robogauge.tasks.pipeline.multi_pipeline import MultiPipeline
from robogauge.utils.logger import Logger
from robogauge.utils.progress_monitor import report_progress, ProgressTypes, ProgressData
level_logger = Logger() # LevelPipeline logger
class LevelPipeline:
def __init__(self, args, console_output: bool = True):
def __init__(self, args, console_output=True, progress_data: ProgressData = None):
self.args = args
self.seeds = args.seeds
self.console_output = console_output
self.progress_data = progress_data
level_logger.create(args.experiment_name+'_level', args.run_name, console_output=console_output)
def run(self):
level_logger.info(f"🚀 Starting Level Searcher for '{self.args.experiment_name}'.")
level_logger.info(f"🔢 Seeds: {self.seeds}")
report_progress(self.progress_data, ProgressTypes.INIT, total=10, desc="🔍 Searching Max Level")
# binary search levels
l, r = 0, 10
all_level_results = {}
while l < r:
level = (l + r + 1) // 2
report_progress(self.progress_data, ProgressTypes.DESC, desc=f"🔍 Testing Level {level}")
all_success, results = self.test_level(level)
all_level_results[level] = results
if all_success:
l = level
else:
r = level - 1
report_progress(self.progress_data, ProgressTypes.UPDATE, value=1)
level = l
level_results = all_level_results.get(l, {
'model_path': results['model_path'],

13
robogauge/tasks/pipeline/multi_pipeline.py
View File

@@ -23,6 +23,7 @@ from robogauge.tasks.pipeline.base_pipeline import BasePipeline
from robogauge.utils.task_register import task_register
from robogauge.utils.logger import Logger
from robogauge.utils.process_utils import NoDaemonPool
from robogauge.utils.progress_monitor import report_progress, ProgressTypes, ProgressData
multi_logger = Logger() # MultiPipeline logger
@@ -63,11 +64,13 @@ def run_single_process(args, data):
return ret
class MultiPipeline:
def __init__(self, args, console_output: bool = True):
def __init__(self, args, console_output=True, progress_data: ProgressData = None):
self.args = args
self.seeds = args.seeds
self.frictions = args.frictions
self.base_masses = args.base_masses
self.console_output = console_output
self.progress_data = progress_data
self.num_processes = args.num_processes
self.static_info = {}
multi_logger.create(args.experiment_name+'_multi', args.run_name+'_multi', console_output=console_output)
@@ -83,16 +86,22 @@ class MultiPipeline:
multi_logger.info(f"🔢 Seeds: {self.seeds}, Frictions: {self.frictions}, Base masses: {self.base_masses}")
workers_data = list(product(self.seeds, self.base_masses, self.frictions))
report_progress(self.progress_data, ProgressTypes.INIT, total=len(workers_data), desc="🚀 MultiPipeline Executing")
ctx = multiprocessing.get_context('spawn')
worker_func = functools.partial(run_single_process, self.args)
results_list = []
with NoDaemonPool(processes=self.num_processes, context=ctx) as pool:
iterator = pool.imap_unordered(worker_func, workers_data)
for results in tqdm(iterator, total=len(workers_data), desc="Evaluation"):
bar = iterator
if self.console_output:
bar = tqdm(iterator, total=len(workers_data), desc="Evaluation")
for results in bar:
results_list.append(results)
self.add_static_info('model_path', results['model_path'])
self.add_static_info('terrain_name', results['results']['terrain_name'])
self.add_static_info('terrain_level', results['results']['terrain_level'])
report_progress(self.progress_data, ProgressTypes.UPDATE, value=1)
if results['status'] != 'success':
data = results['data']
multi_logger.error(f"❌ Process with seed={data[0]}, base_mass={data[1]}, friction={data[2]} failed with error: {results['error_msg']}")

63
robogauge/tasks/pipeline/stress_pipeline.py
View File

@@ -8,6 +8,7 @@
@Desc : Stress Pipeline for Robogauge
'''
import yaml
import traceback
import functools
import numpy as np
from tqdm import tqdm
@@ -18,45 +19,57 @@ from collections import defaultdict
from robogauge.utils.logger import Logger
from robogauge.utils.process_utils import NoDaemonPool
from robogauge.utils.progress_monitor import report_progress, ProgressTypes, start_progress_monitor_thread, ProgressData
from robogauge.tasks.pipeline import MultiPipeline, LevelPipeline
from robogauge.tasks.gauge.gauge_configs.terrain_levels_config import TerrainSearchLevelsConfig
stress_logger = Logger() # StressPipeline logger
def run_pipeline(args, data):
def run_pipeline(args, progress_queue, data):
args = deepcopy(args)
task_id = data['task_id']
search = data['search_max_level']
task_label = f"[{data['terrain_name']}]"
progress_data = ProgressData(
task_id=task_id,
msg_prefix=task_label + ' ',
progress_queue=progress_queue
)
if search is True:
task_label += f" M:{data['base_mass']} F:{data['friction']}"
progress_data.msg_prefix = task_label + ' '
args.friction = data['friction']
args.frictions = [data['friction']]
args.base_mass = data['base_mass']
args.base_masses = [data['base_mass']]
args.task_name = f"{data['task_robot_model']}.{data['terrain_name']}"
args.experiment_name = f"{args.experiment_name}_{data['terrain_name']}_baseMass{data['base_mass']}_friction{data['friction']}"
else:
args.task_name = f"{data['task_robot_model']}.{data['terrain_name']}"
args.experiment_name = f"{args.experiment_name}_{data['terrain_name']}"
level = None # flat terrain
if search:
level, results = LevelPipeline(args, console_output=False).run()
args.experiment_name = f"{args.experiment_name}_{data['terrain_name']}_M{data['base_mass']}_F{data['friction']}"
level, level_results = LevelPipeline(args, console_output=False, progress_data=progress_data).run()
if level == 0: # no valid level found
report_progress(progress_data, ProgressTypes.FINISH, desc=f"❌ Failed (Lv 0)")
results = {
'success': False,
'results': results,
'results': level_results,
'data': data,
'level': 0,
}
return results
report_progress(progress_data, ProgressTypes.RESET, total=0, desc=f"✅ Found Lv {level} -> Running")
else:
level = None # flat terrain
args.task_name = f"{data['task_robot_model']}.{data['terrain_name']}"
args.experiment_name = f"{args.experiment_name}_{data['terrain_name']}"
args.level = level
results = {
'success': True,
'results': MultiPipeline(args, console_output=False).run(),
'results': MultiPipeline(args, console_output=False, progress_data=progress_data).run(),
'data': data,
'level': level,
}
report_progress(progress_data, ProgressTypes.FINISH, desc=f"✅ Done (Lv {level})")
return results
class StressPipeline:
@@ -81,9 +94,6 @@ class StressPipeline:
terrain_names = self.args.stress_terrain_names
stress_logger.info(f"🌄 Stress Test Terrain Names: {terrain_names}")
ctx = multiprocessing.get_context('spawn')
worker_func = functools.partial(run_pipeline, self.args)
### Build worker data ###
workers_data = []
terrain_search_levels_config = TerrainSearchLevelsConfig()
@@ -109,13 +119,26 @@ class StressPipeline:
else:
workers_data.append(data)
### Start progress monitor ###
progress_queue, monitor_thread = start_progress_monitor_thread(len(workers_data))
for i, data in enumerate(workers_data):
data['task_id'] = i
### Run and collect results ###
ctx = multiprocessing.get_context('spawn')
worker_func = functools.partial(run_pipeline, self.args, progress_queue)
results_list = []
try:
with NoDaemonPool(processes=self.num_processes, context=ctx) as pool:
iterator = pool.imap_unordered(worker_func, workers_data)
for results in tqdm(iterator, total=len(workers_data), desc="Stress Benchmark"):
for results in iterator:
results_list.append(results)
self.add_static_info('model_path', results['results'].pop('model_path', None))
except Exception as e:
stress_logger.error(f"❌ Stress benchmark encountered an error: {e}, {traceback.format_exc()}")
finally:
progress_queue.put(None) # Stop the progress monitor thread
monitor_thread.join()
stress_logger.info("✅ Stress Benchmark Completed.")
stress_results = self.aggregate_results(results_list)
@@ -144,6 +167,7 @@ class StressPipeline:
summary = {**self.static_info, 'summary': {}}
value_collections = defaultdict(lambda: defaultdict(list))
zero_terrain_count = 0
for result in all_results:
terrain_name = result['data']['terrain_name']
terrain_level = result['level'] # None, 0, 1, ..., 10
@@ -151,7 +175,11 @@ class StressPipeline:
if terrain_level is not None:
key += f'_{terrain_level}'
key += f'_baseMass{result["data"]["base_mass"]}_friction{result["data"]["friction"]}'
summary[key] = result['results'] if terrain_level != 0 else None
if terrain_level == 0:
summary[key] = None
zero_terrain_count += 1
continue
summary[key] = result['results']
for metric, means in result['results']['summary'].items():
for mean_name, mean_value in means.items():
@@ -160,6 +188,7 @@ class StressPipeline:
for metric, means in value_collections.items():
summary['summary'][metric] = {}
for mean_name, values in means.items():
values.extend([0.0] * zero_terrain_count) # include zero terrains
summary['summary'][metric][mean_name] = f"{float(np.mean(values)):.4f} ± {float(np.std(values)):.4f}"
save_path = stress_logger.log_dir / "stress_benchmark_results.yaml"

137
robogauge/utils/progress_monitor.py Normal file
View File

@@ -0,0 +1,137 @@
# -*- coding: utf-8 -*-
'''
@File : progress_monitor.py
@Time : 2025/12/27 00:10:07
@Author : wty-yy
@Version : 1.0
@Blog : https://wty-yy.github.io/
@Desc : Centralized progress monitoring for multiprocessing tasks using tqdm
'''
from tqdm import tqdm
import multiprocessing
from typing import Dict
from threading import Thread
from dataclasses import dataclass
class ProgressTypes:
INIT = 'init' # Init progress (set total, desc)
UPDATE = 'update' # Update progress value (set value)
DESC = 'desc' # Update description text only (set desc)
RESET = 'reset' # Reset progress bar (set total, desc)
FINISH = 'finish' # Mark completion (set desc)
ERROR = 'error' # Mark error (set desc)
@dataclass
class ProgressData:
progress_queue: multiprocessing.Queue
task_id: int = 0
msg_prefix: str = ''
def report_progress(progress_data: ProgressData, msg_type, value=None, desc=None, total=None):
"""
Assistant function to report progress to the main process.
Args:
queue: multiprocessing.Queue
task_id: int, line number of the task
msg_type: ProgressTypes
value: int, value to update (for UPDATE type)
desc: str, description text (for DESC, INIT, FINISH types)
total: int, total value (for INIT, RESET types)
"""
if progress_data is None:
return
queue = progress_data.progress_queue
task_id = progress_data.task_id
msg_prefix = progress_data.msg_prefix
if desc is not None:
desc = msg_prefix + desc
queue.put((task_id, msg_type, {'value': value, 'desc': desc, 'total': total}))
class ProgressMonitor:
def __init__(self, total_rows):
self.total_rows = total_rows
self.bars: Dict[int, tqdm] = {}
def listener_loop(self, queue):
"""
Run in a separate thread in the main process to consume the Queue and update tqdm.
"""
# print(f"Monitor started for {self.total_rows} tasks...")
for i in range(self.total_rows):
self.bars[i] = tqdm(
total=100,
position=i,
desc=f"Task {i} Pending...",
bar_format="{l_bar}{bar}| {n_fmt}/{total_fmt} [{elapsed}]",
leave=True
)
active_tasks = self.total_rows
while active_tasks > 0:
record = queue.get()
if record is None: # Poison pill signal
break
task_id, msg_type, data = record
if task_id not in self.bars:
continue
bar = self.bars[task_id]
if msg_type == ProgressTypes.INIT:
total = data['total']
desc = data['desc']
bar.reset(total=total)
bar.set_description(desc)
bar.refresh()
elif msg_type == ProgressTypes.UPDATE:
val = data['value']
bar.update(val)
elif msg_type == ProgressTypes.DESC:
desc = data['desc']
if desc:
bar.set_description(desc)
elif msg_type == ProgressTypes.RESET:
# Scenario: Level search finished, starting MultiPipeline, reset progress bar
total = data['total']
desc = data['desc']
bar.reset(total=total)
bar.set_description(desc)
bar.refresh()
elif msg_type == ProgressTypes.FINISH:
desc = data['desc']
if desc:
bar.set_description(desc)
bar.refresh()
# Note: We do not close the bar here to keep it displayed until all tasks are finished and closed together.
active_tasks -= 1
elif msg_type == ProgressTypes.ERROR:
desc = data['desc']
bar.set_description(desc)
bar.refresh()
active_tasks -= 1
# After all tasks are finished, close all bars
for bar in self.bars.values():
bar.close()
def start_progress_monitor_thread(total_rows):
"""
Create and start a ProgressMonitor thread. Queue is returned for reporting progress.
Args:
total_rows: int, number of tasks to monitor
Returns:
queue: multiprocessing.Queue
monitor_thread: threading.Thread
"""
queue = multiprocessing.Manager().Queue()
monitor = ProgressMonitor(total_rows)
monitor_thread = Thread(target=monitor.listener_loop, args=(queue,))
monitor_thread.start()
return queue, monitor_thread