MOSS-VL-Instruct-0408 / video_processing_moss_vl.py
CCCCyx's picture
Upload folder using huggingface_hub
b66ac48
# coding=utf-8
# Copyright 2025 The FNLP Vision Team and The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""video processor class for Moss-VL."""
import json
import logging as system_logging
import math
import os
import re
import subprocess
import traceback
from typing import Any, Dict, List, Optional, Union
import numpy as np
import torch
from joblib import Parallel, delayed
from torchcodec.decoders import VideoDecoder
from transformers.feature_extraction_utils import BatchFeature
from transformers.image_utils import ChannelDimension, PILImageResampling, SizeDict, get_image_size, validate_kwargs
from transformers.processing_utils import Unpack, VideosKwargs
from transformers.utils import TensorType, add_start_docstrings, logging
from transformers.video_processing_utils import BASE_VIDEO_PROCESSOR_DOCSTRING, BaseVideoProcessor
from transformers.video_utils import VideoMetadata, group_videos_by_shape, reorder_videos
logger = logging.get_logger(__name__)
# -----------------------------------------------------------------------------
# Torchcodec video frame extraction utilities
# -----------------------------------------------------------------------------
def check_video_for_extra_streams_and_errors(video_path: str) -> dict:
"""
Check if video file has abnormal streams or errors reported by ffprobe.
Args:
video_path: Path to the video file.
Returns:
A dictionary containing:
- 'has_extra_streams': bool, whether there are streams other than video and audio.
- 'unsupported_codec_errors': list, all "Unsupported codec" error messages.
- 'ffprobe_output_error': str, other errors/warnings from ffprobe stderr.
- 'ffprobe_successful': bool, whether ffprobe command executed successfully (return code 0).
- 'stream_details': list, codec_type and index for each stream.
- 'num_streams': int, total number of streams identified in the video file.
"""
result = {
'has_extra_streams': False,
'unsupported_codec_errors': [],
'ffprobe_output_error': '',
'ffprobe_successful': False,
'stream_details': [],
'num_streams': 0
}
command = [
"ffprobe",
"-v", "error",
"-show_streams",
"-show_format",
"-of", "json",
video_path
]
try:
process = subprocess.run(
command,
capture_output=True,
text=True,
check=False
)
result['ffprobe_successful'] = (process.returncode == 0)
if process.stderr:
result['ffprobe_output_error'] = process.stderr
unsupported_codec_pattern = re.compile(r"Unsupported codec with id \d+ for input stream \d+")
result['unsupported_codec_errors'] = unsupported_codec_pattern.findall(process.stderr)
if process.stdout:
ffprobe_data = json.loads(process.stdout)
if 'streams' in ffprobe_data:
result['num_streams'] = len(ffprobe_data['streams'])
for stream in ffprobe_data['streams']:
stream_type = stream.get('codec_type')
stream_index = stream.get('index')
result['stream_details'].append({'index': stream_index, 'codec_type': stream_type})
if stream_type not in ['video', 'audio']:
result['has_extra_streams'] = True
if 'format' in ffprobe_data and 'nb_streams' in ffprobe_data['format']:
if result['num_streams'] == 0:
result['num_streams'] = ffprobe_data['format']['nb_streams']
elif result['num_streams'] != ffprobe_data['format']['nb_streams']:
logger.warning(
f"Number of streams in 'streams' list ({result['num_streams']}) "
f"differs from 'nb_streams' in 'format' ({ffprobe_data['format']['nb_streams']})."
)
except FileNotFoundError:
result['ffprobe_output_error'] = "ffprobe command not found. Please ensure FFmpeg is installed and in your PATH."
result['ffprobe_successful'] = False
except json.JSONDecodeError:
result['ffprobe_output_error'] = "Failed to parse ffprobe JSON output. Check ffprobe installation or video file."
result['ffprobe_successful'] = False
except Exception as e:
result['ffprobe_output_error'] = f"An unexpected error occurred: {e}"
result['ffprobe_successful'] = False
return result
def remove_video_extra_stream_ffmpeg(input_video: str, output_video: str) -> bool:
"""
Remove extra streams from video using ffmpeg.
Args:
input_video: Path to input video.
output_video: Path to output video.
Returns:
bool: True if successful, False otherwise.
"""
command_list = [
"ffmpeg", "-y", "-i", input_video,
"-map", "0:v:0",
"-c", "copy",
"-an",
"-sn",
"-dn",
"-map_metadata", "-1",
"-map_chapters", "-1",
"-movflags", "faststart",
output_video,
]
try:
subprocess.run(command_list, shell=False, check=True, capture_output=True)
return True
except subprocess.CalledProcessError as e:
system_logging.error(f"Command execution failed with return code: {e.returncode}, video: {input_video}")
system_logging.error(f"Error output:\n{e.stderr}")
return False
except FileNotFoundError:
system_logging.error("Error: ffmpeg command not found. Please ensure ffmpeg is installed and in PATH.")
return False
except Exception as e:
system_logging.error(f"Unexpected error executing command: {e}, video: {input_video}", exc_info=True)
return False
def clean_video_streams(video_path: str) -> str:
"""
Clean video streams if extra streams are detected.
Args:
video_path: Path to the video file.
Returns:
str: Path to cleaned video (or original if no cleaning needed).
"""
ffprobe_res = check_video_for_extra_streams_and_errors(video_path)
if ffprobe_res['has_extra_streams']:
base_name = os.path.basename(video_path)
output_folder = os.path.dirname(video_path)
file_name_without_ext, file_ext = os.path.splitext(base_name)
new_base_name = f"{file_name_without_ext}_fix{file_ext}"
video_path_output = os.path.join(output_folder, new_base_name)
process_flag = remove_video_extra_stream_ffmpeg(video_path, video_path_output)
if not process_flag:
logger.warning("Failed to remove extra streams with ffmpeg")
return video_path
return video_path_output
return video_path
def split_indices(indices: List[Union[int, float]], num_chunks: int) -> List[List[Union[int, float]]]:
"""
Split an index list into roughly equal chunks.
Args:
indices: List of indices to split.
num_chunks: Number of chunks to create.
Returns:
List of index chunks.
"""
chunk_size = len(indices) // num_chunks
chunks = []
for i in range(num_chunks - 1):
chunks.append(indices[i * chunk_size:(i + 1) * chunk_size])
chunks.append(indices[(num_chunks - 1) * chunk_size:])
return chunks
def decode_sequentially(indices: List[int], video_path: str, ffmpeg_threads: int = 0):
"""
Decode frames sequentially from a video.
Args:
indices: List of frame indices to decode.
video_path: Path to the video file.
ffmpeg_threads: Number of ffmpeg threads to use.
Returns:
FrameBatch from torchcodec.
"""
decoder = VideoDecoder(video_path, num_ffmpeg_threads=ffmpeg_threads)
try:
return decoder.get_frames_at(indices)
finally:
del decoder
def decode_with_multithreading(indices: List[int], num_threads: int, video_path: str) -> dict:
"""
Decode frames using multithreading with joblib.
Args:
indices: List of frame indices to decode.
num_threads: Number of threads to use.
video_path: Path to the video file.
Returns:
dict: Contains 'data', 'duration_seconds', 'pts_seconds' tensors.
"""
chunks = split_indices(indices, num_chunks=num_threads)
results = Parallel(n_jobs=num_threads, prefer="threads", verbose=0)(
delayed(decode_sequentially)(chunk, video_path) for chunk in chunks
)
return {
"data": torch.cat([frame_batch.data for frame_batch in results], dim=0),
"duration_seconds": torch.cat([frame_batch.duration_seconds for frame_batch in results], dim=0),
"pts_seconds": torch.cat([frame_batch.pts_seconds for frame_batch in results], dim=0)
}
def decode_sequentially_timestamp(timestamp_list: List[float], video_path: str, ffmpeg_threads: int = 0):
"""
Decode frames sequentially from a video based on timestamps.
Args:
timestamp_list: List of timestamps (in seconds) to decode.
video_path: Path to the video file.
ffmpeg_threads: Number of ffmpeg threads to use.
Returns:
FrameBatch from torchcodec.
"""
decoder = VideoDecoder(video_path, num_ffmpeg_threads=ffmpeg_threads)
try:
metadata = decoder.metadata
min_pts = metadata.begin_stream_seconds_from_content
if min_pts is None:
min_pts = 0.0
max_pts = None
if metadata.num_frames_from_content and metadata.average_fps:
max_pts = (metadata.num_frames_from_content - 1) / metadata.average_fps + min_pts
elif metadata.end_stream_seconds_from_content is not None:
max_pts = metadata.end_stream_seconds_from_content
else:
max_pts = metadata.duration_seconds
if max_pts is not None and max_pts > 0:
timestamp_list = [max(min_pts, min(t, max_pts)) for t in timestamp_list]
elif min_pts > 0:
timestamp_list = [max(min_pts, t) for t in timestamp_list]
return decoder.get_frames_played_at(timestamp_list)
finally:
del decoder
def timestamp_decode_with_multithreading(timestamp_list: List[float], num_threads: int, video_path: str) -> dict:
"""
Decode frames using multithreading based on timestamps.
Args:
timestamp_list: List of timestamps (in seconds) to decode.
num_threads: Number of threads to use.
video_path: Path to the video file.
Returns:
dict: Contains 'data', 'duration_seconds', 'pts_seconds' tensors.
"""
chunks = split_indices(timestamp_list, num_chunks=num_threads)
results = Parallel(n_jobs=num_threads, prefer="threads", verbose=0)(
delayed(decode_sequentially_timestamp)(chunk, video_path) for chunk in chunks
)
# Concatenate results from all threads
data_list = [frame_batch.data for frame_batch in results]
duration_list = [frame_batch.duration_seconds for frame_batch in results]
pts_list = [frame_batch.pts_seconds for frame_batch in results]
if not data_list:
logger.warning("No frames were successfully decoded.")
return {"data": torch.empty(0), "duration_seconds": torch.empty(0), "pts_seconds": torch.empty(0)}
return {
"data": torch.cat(data_list, dim=0),
"duration_seconds": torch.cat(duration_list, dim=0),
"pts_seconds": torch.cat(pts_list, dim=0)
}
def extract_frames_with_torchcodec(
video_path: str,
sample_frames_count: int,
num_threads: int = 4,
) -> Optional[dict]:
"""
Extract frames from video using torchcodec with multithreading.
Args:
video_path: Path to the video file.
sample_frames_count: Number of frames to sample.
num_threads: Number of threads to use for extraction.
sampling_method: Sampling method, either "index" (uniform frame indices) or "timestamp" (uniform timestamps).
Returns:
dict: Contains 'data' (N, C, H, W), 'duration_seconds' (N,), 'pts_seconds' (N,) tensors.
Returns None if extraction fails.
"""
try:
video_path = clean_video_streams(video_path)
decoder = VideoDecoder(video_path, num_ffmpeg_threads=0)
metadata = decoder.metadata
total_frames_in_video = metadata.num_frames_from_content
effective_sample_count = min(sample_frames_count, total_frames_in_video)
if effective_sample_count == 0:
logger.error("Cannot extract frames: video has 0 frames or specified frame count is 0")
return None
# Generate uniform frame indices
frame_indices = np.linspace(0, total_frames_in_video - 1, effective_sample_count).astype(np.int32)
# Ensure indices are valid and remove duplicates
frame_indices = np.unique(np.clip(frame_indices, 0, total_frames_in_video - 1))
result = decode_with_multithreading(frame_indices.tolist(), num_threads=num_threads, video_path=video_path)
# Add frame_indices to the result for later use
result["frame_indices"] = frame_indices
return result
except Exception:
traceback.print_exc()
return None
def smart_resize(
num_frames: int,
height: int,
width: int,
temporal_factor: int = 1,
factor: int = 32,
min_pixels: int = 128 * 128,
max_pixels: int = 16 * 16 * 2 * 2 * 2 * 6144,
per_frame_min_pixels: int = None,
per_frame_max_pixels: int = None,
):
if num_frames < temporal_factor:
raise ValueError(f"t:{num_frames} must be larger than temporal_factor:{temporal_factor}")
if height < factor or width < factor:
raise ValueError(f"height:{height} or width:{width} must be larger than factor:{factor}")
elif max(height, width) / min(height, width) > 200:
raise ValueError(
f"absolute aspect ratio must be smaller than 200, got {max(height, width) / min(height, width)}"
)
h_bar = round(height / factor) * factor
w_bar = round(width / factor) * factor
t_bar = round(num_frames / temporal_factor) * temporal_factor
# Step 1: Apply per-frame upper limit constraint
if per_frame_max_pixels is not None and h_bar * w_bar > per_frame_max_pixels:
beta = math.sqrt((height * width) / per_frame_max_pixels)
h_bar = max(factor, math.floor(height / beta / factor) * factor)
w_bar = max(factor, math.floor(width / beta / factor) * factor)
# Step 2: Apply 3D volume constraints (frames * height * width)
if t_bar * h_bar * w_bar > max_pixels:
beta = math.sqrt((num_frames * height * width) / max_pixels)
h_bar = max(factor, math.floor(height / beta / factor) * factor)
w_bar = max(factor, math.floor(width / beta / factor) * factor)
elif t_bar * h_bar * w_bar < min_pixels:
beta = math.sqrt(min_pixels / (num_frames * height * width))
h_bar = math.ceil(height * beta / factor) * factor
w_bar = math.ceil(width * beta / factor) * factor
# Step 3: Ensure per-frame lower limit is respected (after volume constraint)
# This guarantees single frame stays within [per_frame_min_pixels, per_frame_max_pixels]
if per_frame_min_pixels is not None and h_bar * w_bar < per_frame_min_pixels:
beta = math.sqrt(per_frame_min_pixels / (height * width))
h_bar = math.ceil(height * beta / factor) * factor
w_bar = math.ceil(width * beta / factor) * factor
return h_bar, w_bar
class MossVLVideoProcessorInitKwargs(VideosKwargs):
patch_size: Optional[int]
temporal_patch_size: Optional[int]
merge_size: Optional[int]
min_frames: Optional[int]
max_frames: Optional[int]
video_fps: Optional[Union[int, float]]
num_extract_threads: Optional[int]
# Total 3D volume budget across all videos; distributed proportionally per video by T*H*W
video_max_pixels: Optional[int]
@add_start_docstrings(
"Constructs a fast Moss-VL video processor that dynamically resizes videos based on the original videos.",
BASE_VIDEO_PROCESSOR_DOCSTRING,
"""
patch_size (`int`, *optional*, defaults to 16):
The spacial patch size of the vision encoder.
temporal_patch_size (`int`, *optional*, defaults to 1):
The temporal patch size of the vision encoder.
merge_size (`int`, *optional*, defaults to 2):
The merge size of the vision encoder to llm encoder.
video_fps (`float`, *optional*, defaults to 1.0):
Target frames per second for video sampling.
min_frames (`int`, *optional*, defaults to 1):
Minimum number of frames to sample from a video.
max_frames (`int`, *optional*, defaults to 256):
Maximum number of frames to sample from a video.
num_extract_threads (`int`, *optional*, defaults to 4):
Number of threads to use for frame extraction.
""",
)
class MossVLVideoProcessor(BaseVideoProcessor):
resample = PILImageResampling.BICUBIC
size = {"shortest_edge": 128 * 32 * 32, "longest_edge": 32 * 32 * 768}
image_mean = [0.5, 0.5, 0.5]
image_std = [0.5, 0.5, 0.5]
do_resize = True
do_rescale = True
do_normalize = True
do_convert_rgb = True
patch_size = 16
temporal_patch_size = 1
merge_size = 2
video_fps = 1.0
min_frames = 1
max_frames = 256
num_extract_threads = 4
do_sample_frames = True
# Total 3D volume budget across all videos; distributed proportionally per video by T*H*W
video_max_pixels = None # read from config
valid_kwargs = MossVLVideoProcessorInitKwargs
model_input_names = ["pixel_values_videos", "video_grid_thw"]
def __init__(self, **kwargs: Unpack[MossVLVideoProcessorInitKwargs]):
super().__init__(**kwargs)
if self.size is not None and (
self.size.get("shortest_edge", None) is None or self.size.get("longest_edge", None) is None
):
raise ValueError("size must contain 'shortest_edge' and 'longest_edge' keys.")
def _further_process_kwargs(
self,
size: Optional[SizeDict] = None,
**kwargs,
) -> dict:
"""
Update kwargs that need further processing before being validated
Can be overridden by subclasses to customize the processing of kwargs.
"""
if size is not None and ("shortest_edge" not in size or "longest_edge" not in size):
raise ValueError("size must contain 'shortest_edge' and 'longest_edge' keys.")
return super()._further_process_kwargs(size=size, **kwargs)
def _get_video_path_from_input(self, video_input: Union[str, Dict[str, Any]]) -> str:
"""Normalize a video input into a video path."""
if isinstance(video_input, dict):
return video_input["video_path"]
return video_input
def _get_video_duration_seconds(self, video_input: Union[str, Dict[str, Any]]) -> float:
"""Get video duration in seconds for weighted frame-budget allocation."""
video_path = clean_video_streams(self._get_video_path_from_input(video_input))
decoder = VideoDecoder(video_path, num_ffmpeg_threads=0)
try:
metadata = decoder.metadata
duration = None
if (
metadata.end_stream_seconds_from_content is not None
and metadata.begin_stream_seconds_from_content is not None
):
duration = metadata.end_stream_seconds_from_content - metadata.begin_stream_seconds_from_content
if duration is None or duration <= 0:
duration = metadata.duration_seconds
return max(0.0, float(duration or 0.0))
finally:
del decoder
def _allocate_max_frames_for_multiple_videos(
self,
video_inputs: List[Union[str, Dict[str, Any]]],
total_max_frames: Optional[int],
) -> List[Optional[int]]:
"""
Treat max_frames as a total budget for multi-video input and allocate it by duration.
The returned values are per-video max_frames. Segment dict inputs still keep their
existing per-segment weighting logic after receiving the video-level allocation.
"""
if not video_inputs:
return []
if total_max_frames is None or len(video_inputs) == 1:
return [total_max_frames] * len(video_inputs)
total_max_frames = int(total_max_frames)
num_videos = len(video_inputs)
if total_max_frames < num_videos:
logger.warning(
"Received max_frames=%s for %s videos. At least one frame per video is required, "
"so falling back to 1 frame per video.",
total_max_frames,
num_videos,
)
return [1] * num_videos
video_durations = [self._get_video_duration_seconds(video_input) for video_input in video_inputs]
total_duration = sum(video_durations)
# Reserve one frame per video first, then distribute the remaining budget by duration.
allocations = [1] * num_videos
remaining_budget = total_max_frames - num_videos
if remaining_budget == 0:
return allocations
if total_duration <= 0:
raw_extra_allocations = [remaining_budget / num_videos] * num_videos
else:
raw_extra_allocations = [
remaining_budget * (duration / total_duration) for duration in video_durations
]
base_extra_allocations = [int(math.floor(value)) for value in raw_extra_allocations]
allocations = [base + extra for base, extra in zip(allocations, base_extra_allocations)]
remainder = remaining_budget - sum(base_extra_allocations)
if remainder > 0:
fractional_parts = [
(raw_value - base_value, index)
for index, (raw_value, base_value) in enumerate(zip(raw_extra_allocations, base_extra_allocations))
]
fractional_parts.sort(key=lambda item: (-item[0], item[1]))
for _, index in fractional_parts[:remainder]:
allocations[index] += 1
return allocations
def calculate_num_frames(
self,
metadata: VideoMetadata,
num_frames: Optional[int] = None,
fps: Optional[Union[int, float]] = None,
min_frames: Optional[int] = None,
max_frames: Optional[int] = None,
**kwargs,
) -> int:
"""
Calculate the number of frames to sample using fps-based logic with min/max constraints.
Logic:
1. Calculate target_frames based on fps and video duration
2. Apply min_frames and max_frames constraints
3. Apply max_allowed_frames protection (rough cap from total video_max_pixels budget)
4. Return the number of frames to sample
Args:
metadata (`VideoMetadata`):
Metadata of the video containing information about total duration, fps and total number of frames.
num_frames (`int`, *optional*):
Maximum number of frames to sample. If provided, overrides fps-based calculation.
fps (`int` or `float`, *optional*):
Target frames to sample per second. Defaults to `self.video_fps`.
min_frames (`int`, *optional*):
Minimum number of frames to sample. If None, uses self.min_frames.
max_frames (`int`, *optional*):
Maximum number of frames to sample. If None, uses self.max_frames.
Returns:
int:
Number of frames to sample.
"""
if fps is not None and num_frames is not None:
raise ValueError("`num_frames` and `fps` are mutually exclusive arguments, please use only one!")
total_num_frames = metadata.total_num_frames
# Use provided min/max or fall back to defaults
effective_min_frames = min_frames if min_frames is not None else self.min_frames
effective_max_frames = max_frames if max_frames is not None else self.max_frames
# Rough per-video frame cap derived from the multi-video total budget
# (exact allocation happens later in _preprocess via weighted distribution)
per_frame_min_pixels = self.size.get("shortest_edge", None) if self.size else None
video_max_pixels = getattr(self, "video_max_pixels", None)
if per_frame_min_pixels is not None and video_max_pixels is not None and per_frame_min_pixels > 0:
max_allowed_frames = video_max_pixels // per_frame_min_pixels
effective_max_frames = min(effective_max_frames, max_allowed_frames)
# Get video duration
if hasattr(metadata, 'duration') and metadata.duration is not None:
duration = metadata.duration
else:
video_fps = metadata.fps
if video_fps is not None and video_fps > 0:
duration = total_num_frames / video_fps
else:
# Fallback: assume 24 fps
video_fps = 24.0
duration = total_num_frames / video_fps
logger.warning_once(
"Could not determine video fps from metadata, defaulting to 24 fps for duration calculation."
)
# Use provided fps or default
target_fps = fps if fps is not None else self.video_fps
# Calculate target frames based on fps and duration
if num_frames is None:
# Calculate how many frames we should sample based on target fps
target_total_frames = int(math.ceil(duration * target_fps - 1e-6))
# Apply min/max constraints
sample_frames = max(target_total_frames, effective_min_frames)
sample_frames = min(sample_frames, effective_max_frames, total_num_frames)
else:
# If num_frames is explicitly provided, use it directly with constraints
sample_frames = min(max(num_frames, effective_min_frames), effective_max_frames, total_num_frames)
return sample_frames
def _fetch_video_segment(
self,
video_path: str,
segment: List[float],
min_frames: Optional[int] = None,
max_frames: Optional[int] = None,
video_fps: Optional[float] = None,
):
"""
Fetch video frames for a specific segment.
Args:
video_path: Path to the video file
segment: [start, end] for a segment (left-closed, right-open) or [time] for a single frame
min_frames: Minimum frames for this segment (weighted). Defaults to self.min_frames. Must be >= 1.
max_frames: Maximum frames for this segment (weighted). Defaults to self.max_frames. Must be >= 1.
video_fps: Target frames per second for video sampling. If None, uses self.video_fps.
Returns:
Tuple of (video_tensor, video_metadata)
"""
# Use provided min/max or fall back to defaults, ensure >= 1
min_frames = max(1, min_frames if min_frames is not None else self.min_frames)
max_frames = max(1, max_frames if max_frames is not None else self.max_frames)
# Use provided video_fps or fall back to self.video_fps
target_video_fps = video_fps if video_fps is not None else self.video_fps
video_path = clean_video_streams(video_path)
decoder = VideoDecoder(video_path, num_ffmpeg_threads=0)
try:
torchcodec_metadata = decoder.metadata
video_fps = torchcodec_metadata.average_fps
# Calculate duration
duration = None
if torchcodec_metadata.end_stream_seconds_from_content is not None and torchcodec_metadata.begin_stream_seconds_from_content is not None:
duration = torchcodec_metadata.end_stream_seconds_from_content - torchcodec_metadata.begin_stream_seconds_from_content
if duration is None or duration <= 0:
duration = torchcodec_metadata.duration_seconds
if len(segment) == 1:
# Single frame at specified time
timestamp = segment[0]
frame_batch = decoder.get_frames_played_at([timestamp])
video_tensor = frame_batch.data
actual_timestamps = [timestamp]
sample_count = 1
else:
# Segment [start, end) - left-closed, right-open interval
start_time, end_time = segment
segment_duration = end_time - start_time
# Calculate number of frames to sample for this segment
target_frames = int(math.ceil(segment_duration * target_video_fps))
target_frames = max(target_frames, min_frames)
target_frames = min(target_frames, max_frames)
# Generate timestamps for uniform sampling within segment
if target_frames == 1:
actual_timestamps = [start_time] # Use start_time for single frame
else:
# Sample uniformly within [start, end), endpoint=False for left-closed right-open
actual_timestamps = np.linspace(start_time, end_time, target_frames, endpoint=False).tolist()
# Use multithreading for extraction
result = timestamp_decode_with_multithreading(actual_timestamps, self.num_extract_threads, video_path)
video_tensor = result["data"]
sample_count = len(actual_timestamps)
# Create VideoMetadata
video_metadata = VideoMetadata(
total_num_frames=sample_count,
fps=video_fps,
duration=duration,
video_backend="torchcodec",
height=torchcodec_metadata.height,
width=torchcodec_metadata.width,
frames_indices=None
)
# Store actual timestamps as a custom attribute for _calculate_timestamps to use
video_metadata.actual_timestamps = actual_timestamps
return video_tensor, video_metadata
finally:
del decoder
def fetch_videos(
self,
video_url_or_urls: Union[str, Dict[str, Any], List[Union[str, Dict[str, Any]]]],
sample_indices_fn=None,
video_fps: Optional[float] = None,
min_frames: Optional[int] = None,
max_frames: Optional[int] = None,
):
"""
Override fetch_videos to use torchcodec for frame extraction.
This method uses torchcodec with multithreading for efficient frame extraction.
Frame count is calculated by the calculate_num_frames method
(fps-based with min/max constraints).
Args:
video_url_or_urls: Can be one of:
- str: Single video path
- Dict: Video with segments {"video_path": str, "segments": List[List[float]]}
- List[Union[str, Dict]]: List of video paths or segment dicts
sample_indices_fn: (Not used) Kept for compatibility with base class signature.
video_fps: Target frames per second for video sampling. If None, uses self.video_fps.
min_frames: Minimum number of frames to sample. If None, uses self.min_frames.
max_frames: Maximum number of frames to sample. If None, uses self.max_frames.
Returns:
Tuple of (videos, metadata) where videos are torch.Tensors and metadata are VideoMetadata objects.
"""
# Use provided values or fall back to self defaults
effective_video_fps = video_fps if video_fps is not None else self.video_fps
effective_min_frames = min_frames if min_frames is not None else self.min_frames
effective_max_frames = max_frames if max_frames is not None else self.max_frames
# Handle recursive calls for lists
if isinstance(video_url_or_urls, list):
all_videos = []
all_metadata = []
if len(video_url_or_urls) == 1:
per_video_max_frames = [effective_max_frames]
else:
per_video_max_frames = self._allocate_max_frames_for_multiple_videos(
video_url_or_urls,
effective_max_frames,
)
for x, allocated_max_frames in zip(video_url_or_urls, per_video_max_frames):
result = self.fetch_videos(
x,
video_fps=effective_video_fps,
min_frames=effective_min_frames,
max_frames=allocated_max_frames,
)
# Check if result is from segment expansion (returns lists) or single item
if isinstance(result[0], list):
all_videos.extend(result[0])
all_metadata.extend(result[1])
else:
all_videos.append(result[0])
all_metadata.append(result[1])
return all_videos, all_metadata
# Handle dict with segments - returns lists (one per segment)
if isinstance(video_url_or_urls, dict):
video_path = video_url_or_urls["video_path"]
segments = video_url_or_urls["segments"]
# Calculate total duration of all time-range segments (len == 2) for weighted min/max frames
# Single-frame segments (len == 1) are excluded from weighting
segment_durations = []
for seg in segments:
if len(seg) == 2:
segment_durations.append(seg[1] - seg[0])
else:
segment_durations.append(None) # Single frame, no weighting
total_segment_duration = sum(d for d in segment_durations if d is not None)
videos = []
metadata = []
for i, segment in enumerate(segments):
if len(segment) == 1:
# Single frame - no weighted min/max, just extract directly
video, meta = self._fetch_video_segment(video_path, segment, video_fps=effective_video_fps)
else:
# Time-range segment - apply weighted min/max frames
if total_segment_duration > 0:
weight = segment_durations[i] / total_segment_duration
else:
# Fallback: equal weight among time-range segments
num_range_segments = sum(1 for d in segment_durations if d is not None)
weight = 1.0 / num_range_segments if num_range_segments > 0 else 1.0
# Calculate weighted min/max frames (ensure >= 1)
weighted_min_frames = max(1, int(round(effective_min_frames * weight)))
weighted_max_frames = max(1, int(round(effective_max_frames * weight)))
video, meta = self._fetch_video_segment(
video_path, segment,
min_frames=weighted_min_frames,
max_frames=weighted_max_frames,
video_fps=effective_video_fps,
)
videos.append(video)
metadata.append(meta)
return videos, metadata
# Single video path
video_path = video_url_or_urls
# Clean video streams first (remove extra streams if needed)
video_path = clean_video_streams(video_path)
decoder = None
try:
# Create VideoDecoder only once for both metadata and frame extraction
decoder = VideoDecoder(video_path, num_ffmpeg_threads=0)
torchcodec_metadata = decoder.metadata
duration = None
if torchcodec_metadata.end_stream_seconds_from_content is not None and torchcodec_metadata.begin_stream_seconds_from_content is not None:
duration = torchcodec_metadata.end_stream_seconds_from_content - torchcodec_metadata.begin_stream_seconds_from_content
if duration is None or duration <= 0:
duration = torchcodec_metadata.duration_seconds
# Use num_frames_from_content for accurate frame count (consistent with extraction)
total_frames_in_video = torchcodec_metadata.num_frames_from_content
# Create VideoMetadata object for sample_frames method
temp_metadata = VideoMetadata(
total_num_frames=total_frames_in_video,
fps=torchcodec_metadata.average_fps,
duration=duration,
video_backend="torchcodec",
height=torchcodec_metadata.height,
width=torchcodec_metadata.width,
frames_indices=None
)
# Use calculate_num_frames method to get the number of frames to sample
sample_frames_count = self.calculate_num_frames(
temp_metadata,
fps=effective_video_fps,
min_frames=effective_min_frames,
max_frames=effective_max_frames,
)
# Ensure sample count is valid
effective_sample_count = min(sample_frames_count, total_frames_in_video)
if effective_sample_count == 0:
raise ValueError(f"Cannot extract frames: video has 0 frames or specified frame count is 0")
# Generate uniform frame indices
frame_indices = np.linspace(0, total_frames_in_video - 1, effective_sample_count).astype(np.int32)
# Ensure indices are valid and remove duplicates
frame_indices = np.unique(np.clip(frame_indices, 0, total_frames_in_video - 1))
# Extract frames using multithreading (decoder is created inside each thread for thread safety)
result = decode_with_multithreading(frame_indices.tolist(), num_threads=self.num_extract_threads, video_path=video_path)
# Extract frame tensor (N, C, H, W)
frames_tensor = result["data"]
# Create final VideoMetadata object
video_metadata = VideoMetadata(
total_num_frames=len(frame_indices),
fps=torchcodec_metadata.average_fps,
duration=duration,
video_backend="torchcodec",
height=torchcodec_metadata.height,
width=torchcodec_metadata.width,
frames_indices=frame_indices
)
# Ensure frames are in (T, C, H, W) format
if frames_tensor.dim() == 4: # (N, C, H, W)
video_tensor = frames_tensor
else:
raise ValueError(f"Unexpected frame tensor shape: {frames_tensor.shape}")
return video_tensor, video_metadata
except Exception as e:
logger.error(f"Error loading video {video_path}: {e}")
traceback.print_exc()
raise ValueError(f"Failed to load video {video_path}: {e}")
finally:
if decoder is not None:
del decoder
def _preprocess(
self,
videos: list[torch.Tensor],
do_convert_rgb: bool = True,
do_resize: bool = True,
size: Optional[SizeDict] = None,
interpolation: PILImageResampling = PILImageResampling.BICUBIC,
do_rescale: bool = True,
rescale_factor: float = 1 / 255.0,
do_normalize: bool = True,
image_mean: Optional[Union[float, list[float]]] = None,
image_std: Optional[Union[float, list[float]]] = None,
patch_size: Optional[int] = None,
temporal_patch_size: Optional[int] = None,
merge_size: Optional[int] = None,
return_tensors: Optional[Union[str, TensorType]] = None,
**kwargs,
):
grouped_videos, grouped_videos_index = group_videos_by_shape(videos)
resized_videos_grouped = {}
video_max_pixels = getattr(self, "video_max_pixels", None)
if video_max_pixels is not None:
total_volume = sum(
sv.shape[0] * sv.shape[1] * sv.shape[3] * sv.shape[4]
for sv in grouped_videos.values()
)
else:
total_volume = 0
for shape, stacked_videos in grouped_videos.items():
B, T, C, H, W = stacked_videos.shape
num_frames, height, width = T, H, W
# Convert to RGB if needed (reuse from base class)
if do_convert_rgb:
stacked_videos = self.convert_to_rgb(stacked_videos)
if do_resize:
if video_max_pixels is not None and total_volume > 0:
allocated_max_pixels = int(video_max_pixels * (T * H * W) / total_volume)
else:
allocated_max_pixels = size.longest_edge
resized_height, resized_width = smart_resize(
num_frames=num_frames,
height=height,
width=width,
temporal_factor=temporal_patch_size,
factor=patch_size * merge_size,
min_pixels=size.shortest_edge,
max_pixels=allocated_max_pixels,
per_frame_min_pixels=size.shortest_edge,
per_frame_max_pixels=size.longest_edge,
)
stacked_videos = stacked_videos.view(B * T, C, H, W)
stacked_videos = self.resize(
stacked_videos,
size=SizeDict(height=resized_height, width=resized_width),
interpolation=interpolation,
)
stacked_videos = stacked_videos.view(B, T, C, resized_height, resized_width)
resized_videos_grouped[shape] = stacked_videos
resized_videos = reorder_videos(resized_videos_grouped, grouped_videos_index)
# Group videos by size for further processing
# Needed in case do_resize is False, or resize returns videos with different sizes
grouped_videos, grouped_videos_index = group_videos_by_shape(resized_videos)
processed_videos_grouped = {}
processed_grids = {}
for shape, stacked_videos in grouped_videos.items():
resized_height, resized_width = get_image_size(stacked_videos[0], channel_dim=ChannelDimension.FIRST)
# Fused rescale and normalize
stacked_videos = self.rescale_and_normalize(
stacked_videos, do_rescale, rescale_factor, do_normalize, image_mean, image_std
)
patches = stacked_videos
# Check that videos have `num_frames` divisible by `temporal_patch_size`
if patches.shape[1] % temporal_patch_size != 0:
repeats = patches[:, -1:].repeat(1, temporal_patch_size - 1, 1, 1, 1)
patches = torch.cat([patches, repeats], dim=1)
batch_size, grid_t, channel = patches.shape[:3]
grid_t = grid_t // temporal_patch_size
grid_h, grid_w = resized_height // patch_size, resized_width // patch_size
patches = patches.view(
batch_size,
grid_t,
temporal_patch_size,
channel,
grid_h // merge_size,
merge_size,
patch_size,
grid_w // merge_size,
merge_size,
patch_size,
)
patches = patches.permute(0, 1, 4, 7, 5, 8, 3, 2, 6, 9)
flatten_patches = patches.reshape(
batch_size,
grid_t * grid_h * grid_w,
channel * temporal_patch_size * patch_size * patch_size,
)
processed_videos_grouped[shape] = flatten_patches
processed_grids[shape] = [[grid_t, grid_h, grid_w]] * batch_size
processed_videos = reorder_videos(processed_videos_grouped, grouped_videos_index)
processed_grids = reorder_videos(processed_grids, grouped_videos_index)
pixel_values_videos = torch.cat(processed_videos, dim=0)
video_grid_thw = torch.tensor(processed_grids)
data = {
"pixel_values_videos": pixel_values_videos,
"video_grid_thw": video_grid_thw,
}
return BatchFeature(data=data, tensor_type=return_tensors)
def preprocess(
self,
videos: Union[str, Dict[str, Any], List[Union[str, Dict[str, Any]]]],
**kwargs,
) -> BatchFeature:
"""
Preprocess videos for the model.
This method overrides the base class to handle two video input formats:
1. String path: "path/to/video.mp4"
2. Dict with segments: {"video_path": "...", "segment": [[start, end], [time], ...]}
Args:
videos: Video input(s) in one of the supported formats.
**kwargs: Additional arguments passed to _preprocess.
Returns:
BatchFeature with pixel_values_videos, video_grid_thw, and optionally video_metadata.
"""
# Validate kwargs
validate_kwargs(
captured_kwargs=kwargs.keys(),
valid_processor_keys=list(self.valid_kwargs.__annotations__.keys()) + ["return_tensors"],
)
# Set default kwargs from self
for kwarg_name in self.valid_kwargs.__annotations__:
kwargs.setdefault(kwarg_name, getattr(self, kwarg_name, None))
# Pop kwargs that are handled separately
return_tensors = kwargs.pop("return_tensors", None)
return_metadata = kwargs.pop("return_metadata", False)
input_data_format = kwargs.pop("input_data_format", None)
device = kwargs.pop("device", None)
kwargs.pop("video_metadata", None) # We generate our own metadata
kwargs.pop("do_sample_frames", None) # We handle sampling ourselves
kwargs.pop("data_format", None) # Not used
# Normalize input to list format
if not isinstance(videos, list):
videos = [videos]
# Get video processing params from kwargs (may be passed explicitly for per-batch configuration)
video_fps = kwargs.pop("video_fps", None)
min_frames = kwargs.pop("min_frames", None)
max_frames = kwargs.pop("max_frames", None)
# Use fetch_videos to handle both string and dict formats
video_tensors, video_metadata = self.fetch_videos(
videos,
video_fps=video_fps,
min_frames=min_frames,
max_frames=max_frames,
)
# Prepare video tensors using _prepare_input_videos
prepared_videos = self._prepare_input_videos(
videos=video_tensors,
input_data_format=input_data_format,
device=device,
)
# Process kwargs for _preprocess
kwargs = self._further_process_kwargs(**kwargs)
self._validate_preprocess_kwargs(**kwargs)
# Call _preprocess with prepared videos
result = self._preprocess(videos=prepared_videos, return_tensors=return_tensors, **kwargs)
# Add metadata if requested
if return_metadata:
result["video_metadata"] = video_metadata
return result
__all__ = ["MossVLVideoProcessor"]