Hi, thank you for releasing this great project.
I am trying to run the official inference pipeline with:
```bash
cd All_LargeDanceAR
DATA_ROOT=../InfiniteDanceData \
CHECKPOINT_PATH=./output/exp_m2d_infinitedance/best_model_stage2.pt \
PROCESSES_PER_GPU=1 \
bash infer.sh
The model checkpoint, VQ-VAE checkpoint, MuQ test features, retrieval data, style metadata, and SMPL files have all been downloaded and placed according to the README expected layout.
However, inference fails in the retrieval branch with the following error:
ModuleNotFoundError: No module named 'RetrievalNet.models'
The error comes from All_LargeDanceAR/RetrievalNet/interclip_rp.py, which imports:
from RetrievalNet.models.rp_clip.clip_encoder import CLIPTextEncoder
from RetrievalNet.models.rp_clip.trans_ae import TransEncoder
I checked both the GitHub repository and the Hugging Face release, but I could not find:
All_LargeDanceAR/RetrievalNet/models/rp_clip/clip_encoder.py
All_LargeDanceAR/RetrievalNet/models/rp_clip/trans_ae.py
I also searched locally with:
find . -path "*rp_clip*" -print
find . -name "clip_encoder.py" -o -name "trans_ae.py"
grep -R "TransEncoder" -n .
The only occurrences of TransEncoder are its usage in RetrievalNet/interclip_rp.py and the config entries such as TransEncoder: / CondEncoder:. I could not find the implementation of TransEncoder.
Could you please clarify whether RetrievalNet/models/rp_clip/ is missing from the current release? If so, would it be possible to release the missing files or provide instructions on how to run bash infer.sh with retrieval enabled?
Thank you very much!
The model checkpoint, VQ-VAE checkpoint, MuQ test features, retrieval data, style metadata, and SMPL files have all been downloaded and placed according to the README expected layout.
However, inference fails in the retrieval branch with the following error:
The error comes from
All_LargeDanceAR/RetrievalNet/interclip_rp.py, which imports:I checked both the GitHub repository and the Hugging Face release, but I could not find:
I also searched locally with:
The only occurrences of
TransEncoderare its usage inRetrievalNet/interclip_rp.pyand the config entries such asTransEncoder:/CondEncoder:. I could not find the implementation ofTransEncoder.Could you please clarify whether
RetrievalNet/models/rp_clip/is missing from the current release? If so, would it be possible to release the missing files or provide instructions on how to runbash infer.shwith retrieval enabled?Thank you very much!