pytorch torchcodec issues

pytorch / torchcodec

PyTorch video decoding

BSD 3-Clause "New" or "Revised" License

77 stars 9 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

Use test-infra's main branch now that https://github.com/pytorch/test-infra/pull/5789 is merged

#292 ahmadsharif1 closed 1 week ago
0
Add option for the user to pass in ffmpeg thread count

#291 ahmadsharif1 closed 1 week ago
0
No-op change: move benchmarking code to a library

#290 ahmadsharif1 closed 1 week ago
3
Prevent CUDA tests from running if there is a build problem

#289 scotts closed 2 weeks ago
0
Indexing a 4D FrameBatch: Should this return a Frame or a 3D FrameBatch?

#288 NicolasHug closed 1 week ago
0
Fix pts -> index conversion in `get_frame_at_timestamps[_in_range]`

#287 NicolasHug closed 2 weeks ago
2
Fix binary search of getFramesDisplayedByTimestamps

#286 NicolasHug closed 2 weeks ago
2
Added benchmark notebook to generate videos and run benchmarks

#285 ahmadsharif1 closed 1 week ago
2
Speed-up time-based samplers by 20X and index-based by 1.5X

#284 NicolasHug closed 2 weeks ago
1
Frame and FrameBatch improvements

#283 NicolasHug closed 2 weeks ago
0
Add sort and dedup logic in C++ with new `getFramesDisplayedByTimestamps` method / core API

#282 NicolasHug closed 2 weeks ago
0
Use tensor accessors in C++ to avoid going through the dispatcher

#281 NicolasHug opened 2 weeks ago
0
Add sort and dedup logic in C++ to `getFramesAtIndices`

#280 NicolasHug closed 2 weeks ago
0
Reorder inits to match declaration

#279 scotts closed 2 weeks ago
0
Let getFramesAtIndices rely on getFrameAtIndex

#278 NicolasHug closed 2 weeks ago
1
Rework HWC / CHW dimension order conversions

#277 NicolasHug closed 2 weeks ago
0
Features to gpu_benchmark.py

#276 ahmadsharif1 opened 2 weeks ago
0
Lint fix.

#275 scotts closed 2 weeks ago
1
Fix cuda builds for lower ffmpeg versions

#274 ahmadsharif1 closed 1 week ago
0
Fix cuda builds for lower than 6.1 ffmpeg versions

#273 ahmadsharif1 closed 3 weeks ago
1
Add ability to build cuda wheels

#272 ahmadsharif1 closed 2 weeks ago
0
Remove unnecessary optionality

#271 scotts closed 3 weeks ago
1
Add Nvidia licenses link to README.md

#270 ahmadsharif1 opened 3 weeks ago
0
Review of frame tensor creation / allocation

#269 NicolasHug opened 3 weeks ago
0
Tensor allocation of batched APIs are incorrect on GPU

#268 NicolasHug closed 3 weeks ago
1
[IGNORE]

#267 NicolasHug closed 3 weeks ago
0
Pass pre-allocate tensors in batch APIs to avoid copies

#266 NicolasHug closed 2 weeks ago
3
Allow to specify a path for ffmpeg binary

#265 gdippolito opened 3 weeks ago
2
Unbreak CI by using C++17 code

#264 ahmadsharif1 closed 3 weeks ago
1
Reuse existing cuda context if possible when creating decoders

#263 ahmadsharif1 closed 3 weeks ago
0
Refactor seeking to only store pts as int64 timestamp

#262 scotts closed 3 weeks ago
3
Fix typo in logger

#261 NicolasHug closed 3 weeks ago
0
Review of C++ decoding entry-points

#260 NicolasHug closed 3 weeks ago
2
Split samplers implementation across different files

#259 NicolasHug closed 3 weeks ago
0
Improve cuda decoding performance by ~2x using decoder caching

#258 ahmadsharif1 closed 3 weeks ago
6
Indexes -> Indices

#257 NicolasHug closed 4 weeks ago
0
Time-based samplers are significantly slower than index-based samplers

#256 NicolasHug closed 2 weeks ago
0
Add Random time-based sampler

#255 NicolasHug closed 3 weeks ago
1
Add stream_index as an option to VideoDecoder

#254 scotts closed 3 weeks ago
1
testing the CI.

#253 ahmadsharif1 closed 1 week ago
0
Just testing the CI

#252 ahmadsharif1 closed 3 weeks ago
0
Debug https://github.com/pytorch/torchcodec/pull/230

#251 huydhn closed 3 weeks ago
0
Add time-based regular sampler

#250 NicolasHug closed 4 weeks ago
0
Use pathlib instead of os.path

#249 NicolasHug closed 3 weeks ago
0
New samplers benchmark

#248 NicolasHug closed 1 month ago
0
Refactor Frame and FrameBatch into their own source file

#247 scotts closed 1 month ago
5
Add frame index to output with metadata.

#246 scotts opened 1 month ago
0
Speed-up samplers by avoiding backwards seeks

#245 NicolasHug closed 1 month ago
2
Make it clear that the device interface is only for non-CPU devices. This makes the code simpler to reason about

#244 ahmadsharif1 closed 1 month ago
2
Fix broken C++ code

#243 ahmadsharif1 closed 1 month ago
1

Previous Next