
The default interactive shell is now zsh.
To update your account to use zsh, please run `chsh -s /bin/zsh`.
For more details, please visit https://support.apple.com/kb/HT208050.
[?1034hlocal$ clear
[3J[H[2Jlocal$ scp -F ssh-config -r upload muse-cloud:/workspace/d11-input
Welcome to vast.ai. If authentication fails, try again after a few seconds, and double check your ssh key.
Have fun!
AI agents: READ /etc/vast-agents-guide.md (also linked as ./AGENTS.md) before acting on or describing this instance: it is the operating guide. ('vast-capabilities' shows live state only, not the guide.)
ssh -tt -F ssh-config muse-cloud "env PS1='gpu$ ' bash --noprofile --norc -i"

input.mp4                                                                           0%    0     0.0KB/s   --:-- ETA
input.mp4                                                                         100%   44KB  57.1KB/s   00:00    

sample.yaml                                                                         0%    0     0.0KB/s   --:-- ETA
sample.yaml                                                                       100%  115     0.3KB/s   00:00    

speech.wav                                                                          0%    0     0.0KB/s   --:-- ETA
speech.wav                                                                          0%    0     0.0KB/s   --:-- ETA
speech.wav                                                                        100%  114KB 103.7KB/s   00:01    
local$ ssh -tt -F ssh-config muse-cloud "env PS1='gpu$ ' bash --noprofile --norc -i"
Welcome to vast.ai. If authentication fails, try again after a few seconds, and double check your ssh key.
Have fun!
AI agents: READ /etc/vast-agents-guide.md (also linked as ./AGENTS.md) before acting on or describing this instance: it is the operating guide. ('vast-capabilities' shows live state only, not the guide.)
[?2004hgpu$ clear
[?2004l
[H[2J[3J[?2004hgpu$ python3 /workspace/d11-tools/record-cloud-settings.py
[?2004l

>>> The output above is this instance's LIVE STATE, not the operating guide.
>>> Before acting on — or answering questions about — this instance, READ
>>> /etc/vast-agents-guide.md (also linked as ./AGENTS.md).
PROVIDER-REPORTED INSTANCE SETTINGS
GPU: RTX 4090 | count: 1 | VRAM: 24564 MiB
Template: ComfyUI | allocated disk: 130.0 GB
Effective CPU cores: 10.0 | reported RAM: 47870 MiB
Driver: 580.173.02 | workspace is persistent volume: False
Recorded UTC: 2026-10-07T05:22:58.404619+00:00
Compute rate: $0.36000/h | storage rate: $0.04815/h | total listed: $0.40815/h
Rates captured from this instance, not a final billed session total.
[?2004hgpu$ /workspace/d10-runtime/muse-env/bin/python /workspace/d11-tools/record-environment.py
[?2004l
VERIFIED MUSE TALK ENVIRONMENT
Python: 3.10.18 | executable: /workspace/d10-runtime/muse-env/bin/python
PyTorch: 2.0.1+cu118 | bundled CUDA: 11.8
GPU: NVIDIA GeForce RTX 4090 | CUDA available: True | FP16 GPU calculation: PASS
MMCV: 2.0.1 | MMPose: 1.1.0 | NumPy: 1.23.5
MuseTalk commit: 0a89dec45a0192b824e3cf4daf96c239440c5ed8
ffmpeg version 6.1.1-3ubuntu5 Copyright (c) 2000-2023 the FFmpeg developers
Environment, dependency and model records saved in /workspace/d11-records
[?2004hgpu$ clear
[?2004l
[H[2J[3J[?2004hgpu$ cd /workspace/d10-runtime/MuseTalk
[?2004l
[?2004hgpu$ mkdir -p cloud-sample
[?2004l
[?2004hgpu$ cp /workspace/d11-input/input.mp4 /workspace/d11-input/speech.wav /workspace/d11-input/sample.yaml cloud-sample/
[?2004l
[?2004hgpu$ cat cloud-sample/sample.yaml
[?2004l
task_0:
  video_path: cloud-sample/input.mp4
  audio_path: cloud-sample/speech.wav
  result_name: cloud-result.mp4
[?2004hgpu$ sha256sum cloud-sample/input.mp4 cloud-sample/speech.wav
[?2004l
895c94e8efff634b054abd17951e0c6e9cb8c1e864a0a9e4035b832f4a676a26  cloud-sample/input.mp4
89a902c22a5d0d75e8c37b1359ed5f71221d16f89fc9ca8e84b6ee1b336bec62  cloud-sample/speech.wav
[?2004hgpu$ ffprobe -v error -show_entries stream=codec_name,width,height,avg_frame_rate -show_entries format=duration -of default=noprint_wrappers=1 cloud-sample/input.mp4
[?2004l
codec_name=h264
width=512
height=512
avg_frame_rate=25/1
duration=3.640000
[?2004hgpu$ clear
[?2004l
[H[2J[3J[?2004hgpu$ set -o pipefail
[?2004l
[?2004hgpu$ test ! -e results/d11 && /usr/bin/time -f "Inference elapsed: %e s" /workspace/d10-runtime/muse-env/bin/python -m scripts.inference \
[?2004l
[?2004h>   --inference_config cloud-sample/sample.yaml --result_dir results/d11 \
[?2004l
[?2004h>   --unet_model_path models/musetalkV15/unet.pth \
[?2004l
[?2004h>   --unet_config models/musetalkV15/musetalk.json --whisper_dir models/whisper \
[?2004l
[?2004h>   --version v15 --batch_size 2 --use_float16 --ffmpeg_path /usr/bin \
[?2004l
[?2004h>   2>&1 | tee /workspace/d11-records/generation.log
[?2004l
bash: /usr/bin/time: No such file or directory
[?2004hgpu$ clear
[?2004l
[H[2J[3J[?2004hgpu$ TIMEFORMAT="Inference elapsed: %3R s"
[?2004l
[?2004hgpu$ time /workspace/d10-runtime/muse-env/bin/python -m scripts.inference \
[?2004l
[?2004h>   --inference_config cloud-sample/sample.yaml --result_dir results/d11 \
[?2004l
[?2004h>   --unet_model_path models/musetalkV15/unet.pth \
[?2004l
[?2004h>   --unet_config models/musetalkV15/musetalk.json --whisper_dir models/whisper \
[?2004l
[?2004h>   --version v15 --batch_size 2 --use_float16 --ffmpeg_path /usr/bin \
[?2004l
[?2004h>   2>&1 | tee /workspace/d11-records/generation.log
[?2004l
Loads checkpoint by local backend from path: ./models/dwpose/dw-ll_ucoco_384.pth
cuda start
An error occurred while trying to fetch models/sd-vae: Error no file named diffusion_pytorch_model.safetensors found in directory models/sd-vae.
Defaulting to unsafe serialization. Pass `allow_pickle=False` to raise an error instead.
/workspace/d10-runtime/muse-env/lib/python3.10/site-packages/torch/_utils.py:776: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly.  To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage()
  return self.fget.__get__(instance, owner)()
load unet model from models/musetalkV15/unet.pth
Loaded inference config: {'task_0': {'video_path': 'cloud-sample/input.mp4', 'audio_path': 'cloud-sample/speech.wav', 'result_name': 'cloud-result.mp4'}}
/workspace/d10-runtime/muse-env/lib/python3.10/site-packages/torch/nn/modules/conv.py:309: UserWarning: Applied workaround for CuDNN issue, install nvrtc.so (Triggered internally at ../aten/src/ATen/native/cudnn/Conv_v8.cpp:80.)
  return F.conv1d(input, weight, bias, self.stride,
Extracting landmarks... time-consuming operation
reading images...

  0%|          | 0/91 [00:00<?, ?it/s]
 38%|███▊      | 35/91 [00:00<00:00, 340.93it/s]
 77%|███████▋  | 70/91 [00:00<00:00, 341.87it/s]
100%|██████████| 91/91 [00:00<00:00, 341.92it/s]
get key_landmark and face bounding boxes with the default value

  0%|          | 0/91 [00:00<?, ?it/s]
  1%|          | 1/91 [00:00<00:51,  1.73it/s]
  4%|▍         | 4/91 [00:00<00:12,  7.15it/s]
  9%|▉         | 8/91 [00:00<00:06, 13.59it/s]
 13%|█▎        | 12/91 [00:00<00:04, 18.65it/s]
 18%|█▊        | 16/91 [00:01<00:03, 22.46it/s]
 22%|██▏       | 20/91 [00:01<00:02, 25.47it/s]
 26%|██▋       | 24/91 [00:01<00:02, 27.70it/s]
 31%|███       | 28/91 [00:01<00:02, 29.30it/s]
 35%|███▌      | 32/91 [00:01<00:01, 30.49it/s]
 40%|███▉      | 36/91 [00:01<00:01, 31.36it/s]
 44%|████▍     | 40/91 [00:01<00:01, 31.96it/s]
 48%|████▊     | 44/91 [00:01<00:01, 32.40it/s]
 53%|█████▎    | 48/91 [00:02<00:01, 32.73it/s]
 57%|█████▋    | 52/91 [00:02<00:01, 32.94it/s]
 62%|██████▏   | 56/91 [00:02<00:01, 33.09it/s]
 66%|██████▌   | 60/91 [00:02<00:00, 33.12it/s]
 70%|███████   | 64/91 [00:02<00:00, 33.12it/s]
 75%|███████▍  | 68/91 [00:02<00:00, 33.17it/s]
 79%|███████▉  | 72/91 [00:02<00:00, 33.20it/s]
 84%|████████▎ | 76/91 [00:02<00:00, 33.22it/s]
 88%|████████▊ | 80/91 [00:02<00:00, 33.12it/s]
 92%|█████████▏| 84/91 [00:03<00:00, 33.17it/s]
 97%|█████████▋| 88/91 [00:03<00:00, 33.18it/s]
100%|██████████| 91/91 [00:03<00:00, 27.48it/s]
********************************************bbox_shift parameter adjustment**********************************************************
Total frame:「91」 Manually adjust range : [ -18~18 ] , the current value: 0
*************************************************************************************************************************************
Number of frames: 91
Starting inference

  0%|          | 0/45 [00:00<?, ?it/s]
  2%|▏         | 1/45 [00:00<00:28,  1.56it/s]
  7%|▋         | 3/45 [00:00<00:08,  4.91it/s]
 13%|█▎        | 6/45 [00:00<00:04,  9.20it/s]
 20%|██        | 9/45 [00:01<00:02, 12.40it/s]
 27%|██▋       | 12/45 [00:01<00:02, 14.75it/s]
 31%|███       | 14/45 [00:01<00:01, 15.85it/s]
 38%|███▊      | 17/45 [00:01<00:01, 17.23it/s]
 44%|████▍     | 20/45 [00:01<00:01, 18.27it/s]
 51%|█████     | 23/45 [00:01<00:01, 18.94it/s]
 58%|█████▊    | 26/45 [00:01<00:00, 19.41it/s]
 64%|██████▍   | 29/45 [00:02<00:00, 19.78it/s]
 71%|███████   | 32/45 [00:02<00:00, 20.02it/s]
 78%|███████▊  | 35/45 [00:02<00:00, 20.22it/s]
 84%|████████▍ | 38/45 [00:02<00:00, 20.36it/s]
 91%|█████████ | 41/45 [00:02<00:00, 20.42it/s]
 98%|█████████▊| 44/45 [00:02<00:00, 20.47it/s]
100%|██████████| 45/45 [00:02<00:00, 16.12it/s]
Padding generated images to original video size

  0%|          | 0/90 [00:00<?, ?it/s]
  1%|          | 1/90 [00:00<00:19,  4.52it/s]
  4%|▍         | 4/90 [00:00<00:06, 12.36it/s]
  7%|▋         | 6/90 [00:00<00:05, 14.33it/s]
  9%|▉         | 8/90 [00:00<00:05, 15.40it/s]
 11%|█         | 10/90 [00:00<00:04, 16.12it/s]
 13%|█▎        | 12/90 [00:00<00:04, 16.46it/s]
 16%|█▌        | 14/90 [00:00<00:04, 17.05it/s]
 18%|█▊        | 16/90 [00:01<00:04, 17.39it/s]
 21%|██        | 19/90 [00:01<00:03, 19.05it/s]
 24%|██▍       | 22/90 [00:01<00:03, 20.10it/s]
 28%|██▊       | 25/90 [00:01<00:03, 20.64it/s]
 31%|███       | 28/90 [00:01<00:02, 21.13it/s]
 34%|███▍      | 31/90 [00:01<00:02, 21.45it/s]
 38%|███▊      | 34/90 [00:01<00:02, 21.70it/s]
 41%|████      | 37/90 [00:01<00:02, 21.67it/s]
 44%|████▍     | 40/90 [00:02<00:02, 21.81it/s]
 48%|████▊     | 43/90 [00:02<00:02, 21.95it/s]
 51%|█████     | 46/90 [00:02<00:02, 21.93it/s]
 54%|█████▍    | 49/90 [00:02<00:01, 21.98it/s]
 58%|█████▊    | 52/90 [00:02<00:01, 22.05it/s]
 61%|██████    | 55/90 [00:02<00:01, 22.08it/s]
 64%|██████▍   | 58/90 [00:02<00:01, 22.05it/s]
 68%|██████▊   | 61/90 [00:03<00:01, 22.01it/s]
 71%|███████   | 64/90 [00:03<00:01, 22.08it/s]
 74%|███████▍  | 67/90 [00:03<00:01, 22.12it/s]
 78%|███████▊  | 70/90 [00:03<00:00, 22.00it/s]
 81%|████████  | 73/90 [00:03<00:00, 22.07it/s]
 84%|████████▍ | 76/90 [00:03<00:00, 22.15it/s]
 88%|████████▊ | 79/90 [00:03<00:00, 22.12it/s]
 91%|█████████ | 82/90 [00:04<00:00, 21.92it/s]
 94%|█████████▍| 85/90 [00:04<00:00, 22.01it/s]
 98%|█████████▊| 88/90 [00:04<00:00, 22.10it/s]
100%|██████████| 90/90 [00:04<00:00, 20.50it/s]
Video generation command: ffmpeg -y -v warning -r 25.0 -f image2 -i results/d11/v15/input_speech/%08d.png -vcodec libx264 -vf format=yuv420p -crf 18 results/d11/v15/temp_input_speech.mp4
Audio combination command: ffmpeg -y -v warning -i cloud-sample/speech.wav -i results/d11/v15/temp_input_speech.mp4 results/d11/v15/cloud-result.mp4
[aist#0:0/pcm_s16le @ 0x5ab6c81c5c00] Guessed Channel Layout: mono
Results saved to results/d11/v15/cloud-result.mp4
Inference elapsed: 27.034 s
[?2004hgpu$ clear
[?2004l
[H[2J[3J[?2004hgpu$ ffprobe -v error -show_entries stream=codec_name,codec_type,width,height,avg_frame_rate -show_entries format=duration -of default=noprint_wrappers=1 results/d11/v15/cloud-result.mp4
[?2004l
codec_name=h264
codec_type=video
width=512
height=512
avg_frame_rate=25/1
codec_name=aac
codec_type=audio
avg_frame_rate=0/0
duration=3.632000
[?2004hgpu$ ffmpeg -v error -i results/d11/v15/cloud-result.mp4 -f null - && echo "Full decode: PASS"
[?2004l
Full decode: PASS
[?2004hgpu$ ha256sum results/d11/v15/cloud-result.mp4
[?2004l
bash: ha256sum: command not found
[?2004hgpu$ exit
[?2004l
exit
Connection to [connection host redacted] closed.

local$ clear
[3J[H[2Jlocal$ scp -F ssh-config muse-cloud:/workspace/d10-runtime/MuseTalk/results/d11/v15/cloud-result.mp4 ./cloud-result. 
mp4
Welcome to vast.ai. If authentication fails, try again after a few seconds, and double check your ssh key.
Have fun!
AI agents: READ /etc/vast-agents-guide.md (also linked as ./AGENTS.md) before acting on or describing this instance: it is the operating guide. ('vast-capabilities' shows live state only, not the guide.)

cloud-result.mp4                                                                    0%    0     0.0KB/s   --:-- ETA
cloud-result.mp4                                                                    0%    0     0.0KB/s   --:-- ETA
cloud-result.mp4                                                                  100%   88KB  50.3KB/s   00:01    
local$ shasum -a 256 cloud-result.mp4
9a603827b0b47a8ac748ac69ae5923b20f593b409e933ba27378fd1b9f1b76f4  cloud-result.mp4
local$ ssh -F ssh-config muse-cloud "sha256sum /workspace/d10-runtime/MuseTalk/results/d11/v15/cloud-result.mp4"
Welcome to vast.ai. If authentication fails, try again after a few seconds, and double check your ssh key.
Have fun!
AI agents: READ /etc/vast-agents-guide.md (also linked as ./AGENTS.md) before acting on or describing this instance: it is the operating guide. ('vast-capabilities' shows live state only, not the guide.)
9a603827b0b47a8ac748ac69ae5923b20f593b409e933ba27378fd1b9f1b76f4  /workspace/d10-runtime/MuseTalk/results/d11/v15/cloud-result.mp4
local$ ssh -F ssh-config muse-cloud "cd /workspace/d10-runtime/MuseTalk && ffprobe -v error -show_entries stream=cod 
ec_name,codec_type,width,height -show_entries format=duration -of default=noprint_wrappers=1 results/d11/v15/cloud-r 
esult.mp4 && ffmpeg -nostdin -v error -i results/d11/v15/cloud-result.mp4 -f null - && echo Full-decode-PASS && sha2 
56sum results/d11/v15/cloud-result.mp4"
Welcome to vast.ai. If authentication fails, try again after a few seconds, and double check your ssh key.
Have fun!
AI agents: READ /etc/vast-agents-guide.md (also linked as ./AGENTS.md) before acting on or describing this instance: it is the operating guide. ('vast-capabilities' shows live state only, not the guide.)
codec_name=h264
codec_type=video
width=512
height=512
codec_name=aac
codec_type=audio
duration=3.632000
Full-decode-PASS
9a603827b0b47a8ac748ac69ae5923b20f593b409e933ba27378fd1b9f1b76f4  results/d11/v15/cloud-result.mp4
local$ clear
[3J[H[2Jlocal$ ssh -F ssh-config muse-cloud "cd /workspace/d10-runtime/MuseTalk && cat cloud-sample/sample.yaml && sha256sum 
 cloud-sample/input.mp4 cloud-sample/speech.wav"
Welcome to vast.ai. If authentication fails, try again after a few seconds, and double check your ssh key.
Have fun!
AI agents: READ /etc/vast-agents-guide.md (also linked as ./AGENTS.md) before acting on or describing this instance: it is the operating guide. ('vast-capabilities' shows live state only, not the guide.)
task_0:
  video_path: cloud-sample/input.mp4
  audio_path: cloud-sample/speech.wav
  result_name: cloud-result.mp4
895c94e8efff634b054abd17951e0c6e9cb8c1e864a0a9e4035b832f4a676a26  cloud-sample/input.mp4
89a902c22a5d0d75e8c37b1359ed5f71221d16f89fc9ca8e84b6ee1b336bec62  cloud-sample/speech.wav
local$ clear
[3J[H[2Jlocal$ ssh -F ssh-config muse-cloud "cd /workspace/d10-runtime/MuseTalk && ffprobe -v error -show_entries stream=cod 
ec_name,codec_type,width,height -show_entries format=duration -of default=noprint_wrappers=1 results/d11/v15/cloud-r 
esult.mp4 && ffmpeg -nostdin -v error -i results/d11/v15/cloud-result.mp4 -f null - && echo Full-decode-PASS && sha2 
56sum results/d11/v15/cloud-result.mp4"
Welcome to vast.ai. If authentication fails, try again after a few seconds, and double check your ssh key.
Have fun!
AI agents: READ /etc/vast-agents-guide.md (also linked as ./AGENTS.md) before acting on or describing this instance: it is the operating guide. ('vast-capabilities' shows live state only, not the guide.)
codec_name=h264
codec_type=video
width=512
height=512
codec_name=aac
codec_type=audio
duration=3.632000
Full-decode-PASS
9a603827b0b47a8ac748ac69ae5923b20f593b409e933ba27378fd1b9f1b76f4  results/d11/v15/cloud-result.mp4
local$ 