[VLM] Replace decord with torchcodec for video decoding (#20055)

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: BakerBunker <17872844+BakerBunker@users.noreply.github.com>
This commit is contained in:
Xinyuan Tong
2026-03-09 11:23:49 +00:00
committed by GitHub
parent 2d6eb7dff0
commit 4a757990a1
17 changed files with 234 additions and 144 deletions

View File

@@ -23,7 +23,7 @@ dependencies = [
"build",
"compressed-tensors",
"cuda-python==12.9",
"decord2",
"decord2 ; sys_platform == 'linux' and (platform_machine == 'aarch64' or platform_machine == 'arm64' or platform_machine == 'armv7l')",
"datasets",
"einops",
"fastapi",
@@ -68,7 +68,8 @@ dependencies = [
"torch==2.9.1",
"torchao==0.9.0",
"torchaudio==2.9.1",
"torchcodec==0.8.0 ; sys_platform != 'linux' or (sys_platform == 'linux' and platform_machine != 'aarch64' and platform_machine != 'arm64' and platform_machine != 'armv7l')", # torchcodec does not exist in those systems. If not provided, transformer will use torchvision instead by default.
"torchcodec==0.9.1 ; sys_platform != 'linux' or (sys_platform == 'linux' and platform_machine != 'aarch64' and platform_machine != 'arm64' and platform_machine != 'armv7l')", # torchcodec 0.9.1 for torch 2.9.x. Not available on Linux ARM.
"av ; sys_platform == 'linux' and (platform_machine == 'aarch64' or platform_machine == 'arm64' or platform_machine == 'armv7l')",
"torchvision",
"tqdm",
"transformers==4.57.1",