add decode logging

This commit is contained in:
Pepijn
2025-08-30 16:00:55 +02:00
parent aed90c8042
commit 0be53ef3e1
4 changed files with 213 additions and 1 deletions
+74
View File
@@ -0,0 +1,74 @@
#!/usr/bin/env python
"""
Convert video dataset to image dataset for faster training.
This pre-extracts all frames from MP4 files to PNG images.
"""
import argparse
from pathlib import Path
import logging
import shutil
def convert_dataset_videos_to_images(repo_id: str, root: str | None = None):
"""Convert all videos in a LeRobot dataset to individual image files."""
from lerobot.datasets.lerobot_dataset import LeRobotDataset
from lerobot.datasets.video_utils import decode_video_frames
import torch
# Load dataset
dataset = LeRobotDataset(repo_id, root=root, download_videos=True)
total_frames_processed = 0
for ep_idx in range(dataset.meta.total_episodes):
logging.info(f"Processing episode {ep_idx}/{dataset.meta.total_episodes}")
for vid_key in dataset.meta.video_keys:
video_path = dataset.root / dataset.meta.get_video_file_path(ep_idx, vid_key)
if not video_path.exists():
logging.warning(f"Video not found: {video_path}")
continue
# Create image directory
img_dir = dataset.root / f"images/chunk-{dataset.meta.get_episode_chunk(ep_idx)}/{vid_key}"
img_dir.mkdir(parents=True, exist_ok=True)
# Decode all frames from video
# Get episode length to decode all frames
ep_length = dataset.meta.episodes[ep_idx]["length"]
timestamps = [i / dataset.fps for i in range(ep_length)]
try:
frames = decode_video_frames(video_path, timestamps, dataset.tolerance_s, dataset.video_backend)
# Save each frame as PNG
for i, frame in enumerate(frames.squeeze(0)):
img_path = img_dir / f"episode_{ep_idx:06d}_{i:06d}.png"
# Convert tensor to PIL and save
import torchvision.transforms as T
to_pil = T.ToPILImage()
pil_frame = to_pil(frame)
pil_frame.save(img_path)
total_frames_processed += len(frames.squeeze(0))
logging.info(f" Extracted {len(frames.squeeze(0))} frames to {img_dir}")
except Exception as e:
logging.error(f"Failed to process {video_path}: {e}")
continue
logging.info(f"Conversion complete! Processed {total_frames_processed} total frames")
logging.info(f"You can now use download_videos=False to use the extracted images")
if __name__ == "__main__":
parser = argparse.ArgumentParser(description="Convert LeRobot video dataset to images")
parser.add_argument("repo_id", help="Dataset repo ID (e.g., 'kenmacken/record-test-2')")
parser.add_argument("--root", help="Local root directory", default=None)
args = parser.parse_args()
logging.basicConfig(level=logging.INFO)
convert_dataset_videos_to_images(args.repo_id, args.root)
@@ -67,6 +67,7 @@ class RLearNConfig(PreTrainedConfig):
# Performance optimizations # Performance optimizations
use_amp: bool = True # Mixed precision training for speed boost use_amp: bool = True # Mixed precision training for speed boost
compile_model: bool = True # torch.compile for additional speedup compile_model: bool = True # torch.compile for additional speedup
video_backend: str = "pyav" # Use PyAV for faster video decoding (vs torchcodec)
# ReWiND-specific parameters # ReWiND-specific parameters
use_video_rewind: bool = True # Enable video rewinding augmentation use_video_rewind: bool = True # Enable video rewinding augmentation
+65 -1
View File
@@ -86,6 +86,58 @@ def _add_video_decoding_timing(dataset):
instrument_dataset(dataset) instrument_dataset(dataset)
else: else:
print(f"Warning: Unknown dataset type {type(dataset)}, skipping video timing instrumentation") print(f"Warning: Unknown dataset type {type(dataset)}, skipping video timing instrumentation")
def _add_video_frame_caching(dataset, cache_size=1000):
"""Add LRU caching to video decoding to avoid re-decoding the same frames."""
from functools import lru_cache
from lerobot.datasets.lerobot_dataset import LeRobotDataset, MultiLeRobotDataset
def instrument_dataset_caching(ds):
if not hasattr(ds, '_query_videos'):
return
# Store original method
original_query_videos = ds._query_videos
# Create cache key from timestamps and episode
def make_cache_key(query_timestamps, ep_idx):
# Convert to hashable tuple
key_parts = [ep_idx]
for vid_key in sorted(query_timestamps.keys()):
ts_tuple = tuple(round(ts, 6) for ts in query_timestamps[vid_key]) # Round to microsecond precision
key_parts.append((vid_key, ts_tuple))
return tuple(key_parts)
# Create LRU cached version
@lru_cache(maxsize=cache_size)
def cached_decode_frames(cache_key, ep_idx):
# Reconstruct query_timestamps from cache_key
query_timestamps = {}
for item in cache_key[1:]: # Skip ep_idx
vid_key, ts_tuple = item
query_timestamps[vid_key] = list(ts_tuple)
return original_query_videos(query_timestamps, ep_idx)
def cached_query_videos(self, query_timestamps, ep_idx):
cache_key = make_cache_key(query_timestamps, ep_idx)
return cached_decode_frames(cache_key, ep_idx)
# Bind the cached method to the instance
import types
ds._query_videos = types.MethodType(cached_query_videos, ds)
ds._cached_decode_frames = cached_decode_frames # Keep reference for cache info
print(f"Added video frame caching with size {cache_size}")
# Handle both single and multi datasets
if isinstance(dataset, MultiLeRobotDataset):
for ds in dataset._datasets:
instrument_dataset_caching(ds)
elif isinstance(dataset, LeRobotDataset):
instrument_dataset_caching(dataset)
else:
print(f"Warning: Unknown dataset type {type(dataset)}, skipping video caching")
from termcolor import colored from termcolor import colored
from torch.amp import GradScaler from torch.amp import GradScaler
from torch.optim import Optimizer from torch.optim import Optimizer
@@ -243,7 +295,13 @@ def train(cfg: TrainPipelineConfig):
torch.backends.cuda.matmul.allow_tf32 = True torch.backends.cuda.matmul.allow_tf32 = True
logging.info("Creating dataset") logging.info("Creating dataset")
dataset = make_dataset(cfg) # Pass video backend to dataset for RLearN optimization
dataset_kwargs = {}
if getattr(cfg.policy, "type", None) == "rlearn" and hasattr(cfg.policy, "video_backend"):
dataset_kwargs["video_backend"] = cfg.policy.video_backend
logging.info(f"Using video backend: {cfg.policy.video_backend}")
dataset = make_dataset(cfg, **dataset_kwargs)
# Add video decoding timing for RLearN debugging # Add video decoding timing for RLearN debugging
if getattr(cfg.policy, "type", None) == "rlearn": if getattr(cfg.policy, "type", None) == "rlearn":
@@ -432,6 +490,12 @@ def train(cfg: TrainPipelineConfig):
if recent_decodes: if recent_decodes:
avg_video_decode = sum(recent_decodes) / len(recent_decodes) avg_video_decode = sum(recent_decodes) / len(recent_decodes)
print(f" └─ Video decoding: ~{avg_video_decode:.2f} ms/call (included in data loading)") print(f" └─ Video decoding: ~{avg_video_decode:.2f} ms/call (included in data loading)")
# Show cache hit rate if available
if hasattr(ds, '_cached_decode_frames'):
cache_info = ds._cached_decode_frames.cache_info()
hit_rate = cache_info.hits / max(cache_info.hits + cache_info.misses, 1) * 100
print(f" └─ Cache hit rate: {hit_rate:.1f}% ({cache_info.hits}H/{cache_info.misses}M, size={cache_info.currsize})")
except Exception: except Exception:
pass pass
+73
View File
@@ -0,0 +1,73 @@
#!/usr/bin/env python
"""
Quick benchmark to test video decoding speed across different backends.
"""
import time
from pathlib import Path
import torch
def test_video_backend(video_path, backend_name, num_frames=10):
"""Test video decoding speed for a specific backend."""
try:
from lerobot.datasets.video_utils import decode_video_frames
# Create timestamps for first N frames
fps = 30 # Assume 30fps, adjust if needed
timestamps = [i / fps for i in range(num_frames)]
# Time the decoding
start_time = time.perf_counter()
frames = decode_video_frames(video_path, timestamps, tolerance_s=1e-4, backend=backend_name)
decode_time = time.perf_counter() - start_time
frames_decoded = frames.shape[1] if frames.dim() > 1 else frames.shape[0]
ms_per_frame = (decode_time * 1000) / max(frames_decoded, 1)
print(f"{backend_name:12} | {decode_time*1000:6.1f}ms total | {ms_per_frame:6.1f}ms/frame | {frames_decoded} frames")
return decode_time, frames_decoded
except Exception as e:
print(f"{backend_name:12} | ERROR: {str(e)[:50]}...")
return float('inf'), 0
def main():
# Find your video files
video_dir = Path.home() / ".cache/huggingface/lerobot/kenmacken/record-test-2/videos"
video_files = list(video_dir.rglob("*.mp4"))
if not video_files:
print("❌ No video files found! Check the path.")
return
test_video = video_files[0]
print(f"Testing video: {test_video.name}")
print(f"File size: {test_video.stat().st_size / 1024 / 1024:.1f} MB")
print("-" * 60)
backends = ["torchcodec", "pyav", "video_reader"]
results = {}
for backend in backends:
decode_time, frames = test_video_backend(test_video, backend)
results[backend] = (decode_time, frames)
print("-" * 60)
print("RECOMMENDATION:")
# Find fastest backend
valid_results = {k: v for k, v in results.items() if v[0] != float('inf')}
if valid_results:
fastest = min(valid_results.items(), key=lambda x: x[1][0])
print(f"🚀 Use '{fastest[0]}' - fastest backend!")
print(f" Add to your config: video_backend: \"{fastest[0]}\"")
slowest_time = max(valid_results.values())[0]
speedup = slowest_time / fastest[1][0]
print(f" Speedup vs slowest: {speedup:.1f}x faster")
else:
print("❌ No backends worked!")
if __name__ == "__main__":
main()