Hi, thanks for dropping this wonderful tool for video upscale! I'm currently trying to upscale a course video, which is around 40 minutes. But even with streaming mode the RAM usage still keeps increasing and end up with consuming all the 250G RAM I got.
My inference command is:
python inference_cli.py /vid/01.mp4 \
--resolution 720 \
--cuda_device 0,1,2,3,4,5,6,7 \
--batch_size 81 \
--uniform_batch_size \
--temporal_overlap 3 \
--prepend_frames 4 \
--chunk_size 81 \
--dit_model "seedvr2_ema_3b-Q8_0.gguf" \
--vae_offload_device "cpu" \
--cache_vae \
--dit_offload_device "cpu" \
--video_backend ffmpeg \
--cache_ditAnd here is my machine's information:
_,met$$$$$gg. tim@localhost
,g$$$$$$$$$$$$$$$P. -------------
,g$$P"" """Y$$.". OS: Debian GNU/Linux 13 (trixie) x86_64
,$$P' `$$$. Host: IIMS (00001)
',$$P ,ggs. `$$b: Kernel: Linux 6.12.57+deb13-amd64
`d$$' ,$P"' . $$$ Uptime: 17 hours, 52 mins
$$P d$' , $$P Packages: 403 (dpkg)
$$: $$. - ,d$$' Shell: bash 5.2.37
$$; Y$b._ _,d$P' Display (VGA-1): 1024x768 [External]
Y$$. `.`"Y$$$$P"' Terminal: /dev/pts/7
`$$b "-.__ CPU: 2 x Intel(R) Xeon(R) Gold 6130 (64) @ 3.70 GHz
`Y$$b GPU 1: NVIDIA Tesla V100 PCIe 32GB [Discrete]
`Y$$. GPU 2: ASPEED Technology, Inc. ASPEED Graphics Family
`$$b. GPU 3: NVIDIA Tesla V100 PCIe 32GB [Discrete]
`Y$$b. GPU 4: NVIDIA Tesla V100 PCIe 32GB [Discrete]
`"Y$b._ GPU 5: NVIDIA Tesla V100 PCIe 32GB [Discrete]
`"""" GPU 6: NVIDIA Tesla V100 PCIe 32GB [Discrete]
GPU 7: NVIDIA Tesla V100 PCIe 32GB [Discrete]
GPU 8: NVIDIA Tesla V100 PCIe 32GB [Discrete]
GPU 9: NVIDIA Tesla V100 PCIe 32GB [Discrete]
Memory: 4.64 GiB / 250.58 GiB (2%)
Swap: 0 B / 11.53 GiB (0%)Is this expected behavior or there is some way to fix? Thanks for your help!
Hi, thanks for dropping this wonderful tool for video upscale! I'm currently trying to upscale a course video, which is around 40 minutes. But even with streaming mode the RAM usage still keeps increasing and end up with consuming all the 250G RAM I got.
My inference command is:
python inference_cli.py /vid/01.mp4 \ --resolution 720 \ --cuda_device 0,1,2,3,4,5,6,7 \ --batch_size 81 \ --uniform_batch_size \ --temporal_overlap 3 \ --prepend_frames 4 \ --chunk_size 81 \ --dit_model "seedvr2_ema_3b-Q8_0.gguf" \ --vae_offload_device "cpu" \ --cache_vae \ --dit_offload_device "cpu" \ --video_backend ffmpeg \ --cache_ditAnd here is my machine's information:
Is this expected behavior or there is some way to fix? Thanks for your help!