admin/any4lerobot
1
0
Fork 0
mirror of https://github.com/Tavish9/any4lerobot.git synced 2026-05-22 17:39:39 +00:00
Code Issues Packages Projects Releases Wiki Activity

make merging_utils support state_max_dim and action_max_dim (#30)

Browse Source 
* Update merge_lerobot_dataset.py

Split the original single max_dim parameter into two separate parameters: state_max_dim and action_max_dim

* Update README.md

* Update README.md
This commit is contained in:
zhipeng tang
2025-05-21 16:29:33 +08:00
committed by GitHub
parent 242f00a876
commit cf778e6e41
2 changed files with 182 additions and 130 deletions
Download Patch File Download Diff File Expand all files Collapse all files
utils/dataset_merging/README.md
+4 -3
View File
@@ -33,13 +33,14 @@ python dataset_merger.py --sources /path/to/dataset1 /path/to/dataset2 /path/to/
### （二）命令行参数 ### （二）命令行参数
- **--sources**：源数据集文件夹路径列表，至少需要指定一个源数据集路径。 - **--sources**：源数据集文件夹路径列表，至少需要指定一个源数据集路径。
- **--output**：输出数据集文件夹路径，用于指定合并后数据集的存储位置。 - **--output**：输出数据集文件夹路径，用于指定合并后数据集的存储位置。
- **--max_dim**：向量的最大维度，默认值为32。 - **--state_max_dim**：状态向量的最大维度，默认值为32。
- **--action_max_dim**：动作向量的最大维度，默认值为32。
- **--fps**：数据集的帧率，默认值为20。 - **--fps**：数据集的帧率，默认值为20。
- **--copy_images**: 是否将图像从源文件夹复制且合并到输出文件夹。(default: `False`) - **--copy_images**: 是否将图像从源文件夹复制且合并到输出文件夹。(default: `False`)
### （三）示例 ### （三）示例
```bash ```bash
python dataset_merger.py --sources ./robot_dataset_1 ./robot_dataset_2 --output ./merged_dataset --max_dim 18 --fps 30 python dataset_merger.py --sources ./robot_dataset_1 ./robot_dataset_2 --output ./merged_dataset --state_max_dim 32 --action_max_dim 18 --fps 30
``` ```
## 五、数据集格式 ## 五、数据集格式
@@ -85,4 +86,4 @@ dataset/
2. **Q: 如何处理不同FPS的数据集？** 2. **Q: 如何处理不同FPS的数据集？**
**A**: 暂时只支持相同FPS的数据集合并。 **A**: 暂时只支持相同FPS的数据集合并。
3. **Q: 能否只合并某些特定episode？** 3. **Q: 能否只合并某些特定episode？**
**A**: 当前版本会合并所有数据。如需更精细的控制，您可以先筛选数据集，然后再进行合并。 **A**: 当前版本会合并所有数据。如需更精细的控制，您可以先筛选数据集，然后再进行合并。也可以全部合并，然后使用lerobot加载特定的episode。
utils/dataset_merging/merge_lerobot_dataset.py
+178 -127
View File
@@ -532,7 +532,8 @@ def copy_data_files(
source_folders, source_folders,
output_folder, output_folder,
episode_mapping, episode_mapping,
max_dim=18, state_max_dim=32, # 默认状态向量维度为32
action_max_dim=32, # 默认动作向量维度为32
fps=None, fps=None,
episode_to_frame_index=None, episode_to_frame_index=None,
folder_task_mapping=None, folder_task_mapping=None,
@@ -540,25 +541,22 @@ def copy_data_files(
default_fps=20, default_fps=20,
): ):
""" """
复制并处理parquet数据文件，包括维度填充和索引更新 从源文件夹复制数据文件到输出文件夹，同时处理索引映射和维度填充
(Copy and process parquet data files, including dimension padding and index updates) (Copy data files from source folders to output folder, handling index mapping and dimension padding)
Args: Args:
source_folders (list): 源数据集文件夹路径列表 (List of source dataset folder paths) source_folders (list): 源数据集文件夹路径列表 (List of source dataset folder paths)
output_folder (str): 输出文件夹路径 (Output folder path) output_folder (str): 输出文件夹路径 (Output folder path)
episode_mapping (list): 包含(旧文件夹,旧索引,新索引)元组的列表 episode_mapping (list): 包含(旧文件夹,旧索引,新索引)元组的列表
(List of tuples containing (old_folder, old_index, new_index)) (List of tuples containing (old_folder, old_index, new_index))
max_dim (int): 向量的最大维度 (Maximum dimension for vectors) state_max_dim (int): 状态向量的最大维度 (Maximum dimension for state vectors)
fps (float, optional): 帧率，如果未提供则从第一个数据集获取 (Frame rate, if not provided will be obtained from the first dataset) action_max_dim (int): 动作向量的最大维度 (Maximum dimension for action vectors)
episode_to_frame_index (dict, optional): 每个新episode索引对应的起始帧索引映射 fps (float): 帧率 (frames per second)
(Mapping of each new episode index to its starting frame index) episode_to_frame_index (dict): 每个episode对应的起始帧索引
folder_task_mapping (dict, optional): 每个文件夹中task_index的映射关系 (Start frame index for each episode)
(Mapping of task_index for each folder) folder_task_mapping (dict): 文件夹任务映射 (Folder task mapping)
chunks_size (int): 每个chunk包含的episode数量 (Number of episodes per chunk) chunks_size (int): 数据块大小 (Chunk size)
default_fps (float): 默认帧率，当无法从数据集获取时使用 (Default frame rate when unable to obtain from dataset) default_fps (float): 默认帧率 (Default frame rate)
Returns:
bool: 是否成功复制了至少一个文件 (Whether at least one file was successfully copied)
""" """
# 获取第一个数据集的FPS（如果未提供）(Get FPS from first dataset if not provided) # 获取第一个数据集的FPS（如果未提供）(Get FPS from first dataset if not provided)
if fps is None: if fps is None:
@@ -572,7 +570,7 @@ def copy_data_files(
else: else:
fps = default_fps # 使用变量替代硬编码的20 (Use variable instead of hardcoded 20) fps = default_fps # 使用变量替代硬编码的20 (Use variable instead of hardcoded 20)
print(f"使用FPS={fps} (Using FPS={fps})") print(f"使用FPS={fps}")
# 为每个episode复制和处理数据文件 (Copy and process data files for each episode) # 为每个episode复制和处理数据文件 (Copy and process data files for each episode)
total_copied = 0 total_copied = 0
@@ -601,26 +599,48 @@ def copy_data_files(
# 读取parquet文件 (Read parquet file) # 读取parquet文件 (Read parquet file)
df = pd.read_parquet(source_path) df = pd.read_parquet(source_path)
# 检查是否需要填充维度 (Check if dimensions need padding) # 检查是否需要填充维度 - 为不同特征类型使用不同的最大维度
for feature in ["observation.state", "action"]: # 为状态向量填充
if feature in df.columns: if "observation.state" in df.columns:
# 检查第一个非空值 (Check first non-null value) # 检查第一个非空值 (Check first non-null value)
for _idx, value in enumerate(df[feature]): for _idx, value in enumerate(df["observation.state"]):
if value is not None and isinstance(value, (list, np.ndarray)): if value is not None and isinstance(value, (list, np.ndarray)):
current_dim = len(value) current_dim = len(value)
if current_dim < max_dim: if current_dim < state_max_dim:
print( print(
f"填充 {feature} 从 {current_dim} 维到 {max_dim} 维 (Padding {feature} from {current_dim} to {max_dim} dimensions)" f"填充状态向量从 {current_dim} 维到 {state_max_dim} 维"
) f" (Padding state vector from {current_dim} to {state_max_dim} dimensions)"
# 使用零填充到目标维度 (Pad with zeros to target dimension) )
df[feature] = df[feature].apply( # 使用零填充到目标维度 (Pad with zeros to target dimension)
lambda x: np.pad(x, (0, max_dim - len(x)), "constant").tolist() df["observation.state"] = df["observation.state"].apply(
if x is not None lambda x: np.pad(x, (0, state_max_dim - len(x)), "constant").tolist()
and isinstance(x, (list, np.ndarray)) if x is not None
and len(x) < max_dim and isinstance(x, (list, np.ndarray))
else x and len(x) < state_max_dim
) else x
break )
break
# 为动作向量填充
if "action" in df.columns:
# 检查第一个非空值 (Check first non-null value)
for _idx, value in enumerate(df["action"]):
if value is not None and isinstance(value, (list, np.ndarray)):
current_dim = len(value)
if current_dim < action_max_dim:
print(
f"填充动作向量从 {current_dim} 维到 {action_max_dim} 维"
f" (Padding action vector from {current_dim} to {action_max_dim} dimensions)"
)
# 使用零填充到目标维度 (Pad with zeros to target dimension)
df["action"] = df["action"].apply(
lambda x: np.pad(x, (0, action_max_dim - len(x)), "constant").tolist()
if x is not None
and isinstance(x, (list, np.ndarray))
and len(x) < action_max_dim
else x
)
break
# 更新episode_index列 (Update episode_index column) # 更新episode_index列 (Update episode_index column)
if "episode_index" in df.columns: if "episode_index" in df.columns:
@@ -699,28 +719,48 @@ def copy_data_files(
# 读取parquet文件 (Read parquet file) # 读取parquet文件 (Read parquet file)
df = pd.read_parquet(source_path) df = pd.read_parquet(source_path)
# 检查是否需要填充维度 (Check if dimensions need padding) # 检查是否需要填充维度 - 为不同特征类型使用不同的最大维度
for feature in ["observation.state", "action"]: # 为状态向量填充
if feature in df.columns: if "observation.state" in df.columns:
# 检查第一个非空值 (Check first non-null value) # 检查第一个非空值 (Check first non-null value)
for _idx, value in enumerate(df[feature]): for _idx, value in enumerate(df["observation.state"]):
if value is not None and isinstance(value, (list, np.ndarray)): if value is not None and isinstance(value, (list, np.ndarray)):
current_dim = len(value) current_dim = len(value)
if current_dim < max_dim: if current_dim < state_max_dim:
print( print(
f"填充 {feature} 从 {current_dim} 维到 {max_dim} 维 (Padding {feature} from {current_dim} to {max_dim} dimensions)" f"填充状态向量从 {current_dim} 维到 {state_max_dim} 维"
) f" (Padding state vector from {current_dim} to {state_max_dim} dimensions)"
# 使用零填充到目标维度 (Pad with zeros to target dimension) )
df[feature] = df[feature].apply( # 使用零填充到目标维度 (Pad with zeros to target dimension)
lambda x: np.pad( df["observation.state"] = df["observation.state"].apply(
x, (0, max_dim - len(x)), "constant" lambda x: np.pad(x, (0, state_max_dim - len(x)), "constant").tolist()
).tolist() if x is not None
if x is not None and isinstance(x, (list, np.ndarray))
and isinstance(x, (list, np.ndarray)) and len(x) < state_max_dim
and len(x) < max_dim else x
else x )
) break
break
# 为动作向量填充
if "action" in df.columns:
# 检查第一个非空值 (Check first non-null value)
for _idx, value in enumerate(df["action"]):
if value is not None and isinstance(value, (list, np.ndarray)):
current_dim = len(value)
if current_dim < action_max_dim:
print(
f"填充动作向量从 {current_dim} 维到 {action_max_dim} 维"
f" (Padding action vector from {current_dim} to {action_max_dim} dimensions)"
)
# 使用零填充到目标维度 (Pad with zeros to target dimension)
df["action"] = df["action"].apply(
lambda x: np.pad(x, (0, action_max_dim - len(x)), "constant").tolist()
if x is not None
and isinstance(x, (list, np.ndarray))
and len(x) < action_max_dim
else x
)
break
# 更新episode_index列 (Update episode_index column) # 更新episode_index列 (Update episode_index column)
if "episode_index" in df.columns: if "episode_index" in df.columns:
@@ -1205,7 +1245,8 @@ def copy_images(source_folders, output_folder, episode_mapping, default_fps=20,
def merge_datasets( def merge_datasets(
source_folders, output_folder, validate_ts=False, tolerance_s=1e-4, max_dim=18, default_fps=20 source_folders, output_folder, validate_ts=False, tolerance_s=1e-4,
state_max_dim=32, action_max_dim=32, default_fps=20
): ):
""" """
将多个数据集文件夹合并为一个，处理索引、维度和元数据 将多个数据集文件夹合并为一个，处理索引、维度和元数据
@@ -1216,38 +1257,14 @@ def merge_datasets(
output_folder (str): 输出文件夹路径 (Output folder path) output_folder (str): 输出文件夹路径 (Output folder path)
validate_ts (bool): 是否验证时间戳 (Whether to validate timestamps) validate_ts (bool): 是否验证时间戳 (Whether to validate timestamps)
tolerance_s (float): 时间戳不连续性的容差值，以秒为单位 (Tolerance for timestamp discontinuities in seconds) tolerance_s (float): 时间戳不连续性的容差值，以秒为单位 (Tolerance for timestamp discontinuities in seconds)
max_dim (int): 向量的最大维度 (Maximum dimension for vectors) state_max_dim (int): 状态向量的最大维度 (Maximum dimension for state vectors)
action_max_dim (int): 动作向量的最大维度 (Maximum dimension for action vectors)
default_fps (float): 默认帧率 (Default frame rate) default_fps (float): 默认帧率 (Default frame rate)
这个函数执行以下操作:
(This function performs the following operations:)
1. 合并所有的episodes、tasks和stats (Merges all episodes, tasks and stats)
2. 重新编号所有的索引以保持连续性 (Renumbers all indices to maintain continuity)
3. 填充向量维度使其一致 (Pads vector dimensions for consistency)
4. 更新元数据文件 (Updates metadata files)
5. 复制并处理数据和视频文件 (Copies and processes data and video files)
6. 复制并验证图像文件 (Copies and validates image files)
""" """
# Create output folder if it doesn't exist # Create output folder if it doesn't exist
os.makedirs(output_folder, exist_ok=True) os.makedirs(output_folder, exist_ok=True)
os.makedirs(os.path.join(output_folder, "meta"), exist_ok=True) os.makedirs(os.path.join(output_folder, "meta"), exist_ok=True)
# 注释掉时间戳验证
# if validate_ts:
# issues, fps_values = validate_timestamps(source_folders, tolerance_s)
# if issues:
# print("时间戳验证发现以下问题:")
# for issue in issues:
# print(f" - {issue}")
#
# # 获取共同的FPS值
# if fps_values:
# fps = max(set(fps_values), key=fps_values.count)
# print(f"使用共同FPS值: {fps}")
# else:
# fps = default_fps
# print(f"未找到FPS值，使用默认值: {default_fps}")
# else:
fps = default_fps fps = default_fps
print(f"使用默认FPS值: {fps}") print(f"使用默认FPS值: {fps}")
@@ -1266,7 +1283,8 @@ def merge_datasets(
all_stats_data = [] all_stats_data = []
# Track dimensions for each folder # Track dimensions for each folder
folder_dimensions = {} folder_state_dimensions = {} # 存储每个文件夹的状态向量维度
folder_action_dimensions = {} # 存储每个文件夹的动作向量维度
# 添加一个变量来跟踪累积的帧数 # 添加一个变量来跟踪累积的帧数
cumulative_frame_count = 0 cumulative_frame_count = 0
@@ -1310,8 +1328,9 @@ def merge_datasets(
f"从{folder}的info.json中读取到视频数量: {folder_videos} (Read video count from {folder}'s info.json: {folder_videos})" f"从{folder}的info.json中读取到视频数量: {folder_videos} (Read video count from {folder}'s info.json: {folder_videos})"
) )
# Check dimensions of state vectors in this folder # 分别检查状态和动作向量的维度
folder_dim = max_dim # 使用变量替代硬编码的18 folder_state_dim = state_max_dim # 默认使用传入的状态最大维度
folder_action_dim = action_max_dim # 默认使用传入的动作最大维度
# Try to find a parquet file to determine dimensions # Try to find a parquet file to determine dimensions
for root, _dirs, files in os.walk(folder): for root, _dirs, files in os.walk(folder):
@@ -1319,19 +1338,34 @@ def merge_datasets(
if file.endswith(".parquet"): if file.endswith(".parquet"):
try: try:
df = pd.read_parquet(os.path.join(root, file)) df = pd.read_parquet(os.path.join(root, file))
# 检查状态向量维度
if "observation.state" in df.columns: if "observation.state" in df.columns:
first_state = df["observation.state"].iloc[0] for state_val in df["observation.state"]:
if isinstance(first_state, (list, np.ndarray)): if state_val is not None and isinstance(state_val, (list, np.ndarray)):
folder_dim = len(first_state) folder_state_dim = len(state_val)
print(f"Detected {folder_dim} dimensions in {folder}") print(f"Detected {folder_state_dim} dimensions for state in {folder}")
break break
# 检查动作向量维度
if "action" in df.columns:
for action_val in df["action"]:
if action_val is not None and isinstance(action_val, (list, np.ndarray)):
folder_action_dim = len(action_val)
print(f"Detected {folder_action_dim} dimensions for action in {folder}")
break
# 如果两个维度都已检测到，可以停止搜索
if folder_state_dim != state_max_dim and folder_action_dim != action_max_dim:
break
except Exception as e: except Exception as e:
print(f"Error checking dimensions in {folder}: {e}") print(f"Error checking dimensions in {folder}: {e}")
break break
if folder_dim != max_dim: # 使用变量替代硬编码的18 # 如果两个维度都已检测到，可以停止搜索
if folder_state_dim != state_max_dim and folder_action_dim != action_max_dim:
break break
folder_dimensions[folder] = folder_dim folder_state_dimensions[folder] = folder_state_dim
folder_action_dimensions[folder] = folder_action_dim
# Load episodes # Load episodes
episodes_path = os.path.join(folder, "meta", "episodes.jsonl") episodes_path = os.path.join(folder, "meta", "episodes.jsonl")
@@ -1391,25 +1425,25 @@ def merge_datasets(
stats = stats_map[old_index] stats = stats_map[old_index]
stats["episode_index"] = new_index stats["episode_index"] = new_index
# Pad stats data if needed # 填充统计数据
if "stats" in stats and folder_dimensions[folder] < max_dim: # 使用变量替代硬编码的18 if "stats" in stats:
# Pad observation.state and action stats # 分别填充observation.state的统计数据
for feature in ["observation.state", "action"]: if "observation.state" in stats["stats"] and folder_state_dimensions[folder] < state_max_dim:
if feature in stats["stats"]: for stat_type in ["mean", "std", "max", "min"]:
for stat_type in ["mean", "std", "max", "min"]: if stat_type in stats["stats"]["observation.state"]:
if stat_type in stats["stats"][feature]: values = stats["stats"]["observation.state"][stat_type]
# Get current values if isinstance(values, list) and len(values) < state_max_dim:
values = stats["stats"][feature][stat_type] padded = values + [0.0] * (state_max_dim - len(values))
stats["stats"]["observation.state"][stat_type] = padded
# Check if it's a list/array that needs padding
if ( # 分别处理action的统计数据
isinstance(values, list) and len(values) < max_dim if "action" in stats["stats"] and folder_action_dimensions[folder] < action_max_dim:
): # 使用变量替代硬编码的18 for stat_type in ["mean", "std", "max", "min"]:
# Pad with zeros if stat_type in stats["stats"]["action"]:
padded = values + [0.0] * ( values = stats["stats"]["action"][stat_type]
max_dim - len(values) if isinstance(values, list) and len(values) < action_max_dim:
) # 使用变量替代硬编码的18 padded = values + [0.0] * (action_max_dim - len(values))
stats["stats"][feature][stat_type] = padded stats["stats"]["action"][stat_type] = padded
all_episodes_stats.append(stats) all_episodes_stats.append(stats)
@@ -1577,16 +1611,25 @@ def merge_datasets(
# Update feature dimensions to the maximum dimension # Update feature dimensions to the maximum dimension
if "features" in info: if "features" in info:
# Find the maximum dimension across all folders # 使用检测到的最大状态和动作维度
actual_max_dim = max_dim # 使用变量替代硬编码的18 actual_state_max_dim = state_max_dim
for _folder, dim in folder_dimensions.items(): actual_action_max_dim = action_max_dim
actual_max_dim = max(actual_max_dim, dim)
for _folder, dim in folder_state_dimensions.items():
# Update observation.state and action dimensions actual_state_max_dim = max(actual_state_max_dim, dim)
for feature_name in ["observation.state", "action"]:
if feature_name in info["features"] and "shape" in info["features"][feature_name]: for _folder, dim in folder_action_dimensions.items():
info["features"][feature_name]["shape"] = [actual_max_dim] actual_action_max_dim = max(actual_action_max_dim, dim)
print(f"Updated {feature_name} shape to {actual_max_dim}")
# 更新状态向量维度
if "observation.state" in info["features"] and "shape" in info["features"]["observation.state"]:
info["features"]["observation.state"]["shape"] = [actual_state_max_dim]
print(f"Updated observation.state shape to {actual_state_max_dim}")
# 更新动作向量维度
if "action" in info["features"] and "shape" in info["features"]["action"]:
info["features"]["action"]["shape"] = [actual_action_max_dim]
print(f"Updated action shape to {actual_action_max_dim}")
# 更新视频总数 (Update total videos) # 更新视频总数 (Update total videos)
info["total_videos"] = total_videos info["total_videos"] = total_videos
@@ -1608,7 +1651,8 @@ def merge_datasets(
source_folders, source_folders,
output_folder, output_folder,
episode_mapping, episode_mapping,
max_dim=max_dim, state_max_dim=state_max_dim,
action_max_dim=action_max_dim,
fps=fps, fps=fps,
episode_to_frame_index=episode_to_frame_index, episode_to_frame_index=episode_to_frame_index,
folder_task_mapping=folder_task_mapping, folder_task_mapping=folder_task_mapping,
@@ -1631,12 +1675,19 @@ if __name__ == "__main__":
# Add arguments # Add arguments
parser.add_argument("--sources", nargs="+", required=True, help="List of source folder paths") parser.add_argument("--sources", nargs="+", required=True, help="List of source folder paths")
parser.add_argument("--output", required=True, help="Output folder path") parser.add_argument("--output", required=True, help="Output folder path")
parser.add_argument("--max_dim", type=int, default=32, help="Maximum dimension (default: 32)") parser.add_argument("--state_max_dim", type=int, default=32, help="Maximum state vector dimension (default: 32)")
parser.add_argument("--action_max_dim", type=int, default=32, help="Maximum action vector dimension (default: 32)")
parser.add_argument("--fps", type=int, default=20, help="Your datasets FPS (default: 20)") parser.add_argument("--fps", type=int, default=20, help="Your datasets FPS (default: 20)")
parser.add_argument("--copy_images", action="store_true", help="Whether copy images from source folders to output folder with validation. (default: False)",) parser.add_argument("--copy_images", action="store_true", help="Whether to copy images (default: False)")
# Parse arguments # Parse arguments
args = parser.parse_args() args = parser.parse_args()
# Use parsed arguments # Use parsed arguments
merge_datasets(args.sources, args.output, max_dim=args.max_dim, default_fps=args.fps) merge_datasets(
args.sources,
args.output,
state_max_dim=args.state_max_dim,
action_max_dim=args.action_max_dim,
default_fps=args.fps
)