PR #21654 opened by Steven Xiao (younengxiao)
URL: https://code.ffmpeg.org/FFmpeg/FFmpeg/pulls/21654
Patch URL: https://code.ffmpeg.org/FFmpeg/FFmpeg/pulls/21654.patch
When using `dnn_processing` filter with torch backend, FFmpeg hangs
indefinitely because no inference is actually performed.
Resolve this problem by add "else" branch for synchronous execution path.
Usage:
ffmpeg -i input.mp4 -vf
scale=224:224,format=rgb24,dnn_processing=dnn_backend=torch:model=sr_model_torch.pt:device=cpu
output.mp4
>From 5b5324d6242eee72cba10365ac9c36ad69481d78 Mon Sep 17 00:00:00 2001
From: stevxiao <[email protected]>
Date: Thu, 5 Feb 2026 11:34:07 -0500
Subject: [PATCH] avfilter/dnn: add synchronous execution path in torch backend
When using `dnn_processing` filter with torch backend, FFmpeg hangs
indefinitely because no inference is actually performed.
Resolve this problem by add "else" branch for synchronous execution path.
Usage:
ffmpeg -i input.mp4 -vf
scale=224:224,format=rgb24,dnn_processing=dnn_backend=torch:model=sr_model_torch.pt:device=cpu
output.mp4
---
libavfilter/dnn/dnn_backend_torch.cpp | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/libavfilter/dnn/dnn_backend_torch.cpp
b/libavfilter/dnn/dnn_backend_torch.cpp
index 33809bf983..56fbe372d1 100644
--- a/libavfilter/dnn/dnn_backend_torch.cpp
+++ b/libavfilter/dnn/dnn_backend_torch.cpp
@@ -424,6 +424,14 @@ static int execute_model_th(THRequestItem *request, Queue
*lltask_queue)
}
th_model->cond->notify_one();
return 0;
+ } else {
+ // Synchronous execution path
+ ret = th_start_inference((void *)(request));
+ if (ret != 0) {
+ goto err;
+ }
+ infer_completion_callback(request);
+ return (task->inference_done == task->inference_todo) ? 0 :
DNN_GENERIC_ERROR;
}
err:
--
2.52.0
_______________________________________________
ffmpeg-devel mailing list -- [email protected]
To unsubscribe send an email to [email protected]
