Closed HappyPiepie closed 1 year ago
For the smpl model refer here. For AMASS, you can download the corresponding datasets in SMPLH format from the AMASS website. (SMPLH-G)
It is by default in a pickle format. If you follow the instructions in the data section. You don't have to do anything special just SMPLH-G and then run my scripts. For SMPLH model follow the instruction in the link in my previous comment.
Thank you very much, i will try .
There is another problem, I get "(teach-env) (base) beibeigh@beibeigh-System-Product-Name:~/Research/Projects/teach$ python scripts/process_amass.py --input-path /path/to/data --output-path path/of/choice/defaultis/babel/babel-smplh-30fps-male --model-type smplh --use-betas --gender male usage: process_amass.py [-h] --input-path INPUT_PATH --output-path OUTPUT_PATH [--use-betas] --gender {male,female,neutral,amass} process_amass.py: error: unrecognized arguments: --model-type smplh " and i don't konw how to address?
Sorry, yes I have corrected the instruction the model-type argument should be removed. Use
python scripts/process_amass.py --input-path /path/to/amass/data --output-path path/of/choice/default/is/data/babel/babel-smplh-30fps-male --use-betas --gender male
.
i have another problem when i run "python train.py experiment=baseline logger=none, (teach-env) (base) beibeigh@beibeigh-System-Product-Name:~/Research/Projects/teach$ HYDRA_FULL_ERROR=1 python train.py experiment=baseline logger=none [12/10/22 17:35:27][main][INFO] - Training script. The outputs will be stored in: [12/10/22 17:35:27][main][INFO] - The working directory is:/home/beibeigh/Research/Projects/teach/teach/babel-amass/baseline/19qj42vz [12/10/22 17:35:27][main][INFO] - Loading libraries [12/10/22 17:35:28][main][INFO] - Libraries loaded [12/10/22 17:35:28][main][INFO] - Set the seed to 42 [12/10/22 17:35:28][pytorch_lightning.utilities.seed][INFO] - Global seed set to 42 [12/10/22 17:35:28][main][INFO] - Loading data module [12/10/22 17:35:28][main][INFO] - Data module 'babel-amass' loaded [12/10/22 17:35:28][main][INFO] - Loading model [12/10/22 17:35:28][torch.distributed.nn.jit.instantiator][INFO] - Created a temporary directory at /tmp/tmpt0eyfzcj [12/10/22 17:35:28][torch.distributed.nn.jit.instantiator][INFO] - Writing /tmp/tmpt0eyfzcj/_remote_module_non_sriptable.py [12/10/22 17:35:29][main][INFO] - Model 'teach' loaded [12/10/22 17:35:29][main][INFO] - Loading logger [12/10/22 17:35:29][main][INFO] - Logger 'none' ready [12/10/22 17:35:29][main][INFO] - Loading callbacks [12/10/22 17:35:29][OpenGL.acceleratesupport][INFO] - No OpenGL_accelerate module loaded: No module named 'OpenGL_accelerate' [12/10/22 17:35:29][main][INFO] - Callbacks initialized [12/10/22 17:35:29][main][INFO] - Loading trainer [12/10/22 17:35:29][py.warnings][WARNING] - /home/beibeigh/Research/Projects/teach/teach-env/lib/python3.9/site-packages/torch/cuda/init.py:145: UserWarning: NVIDIA GeForce RTX 3090 with CUDA capability sm_86 is not compatible with the current PyTorch installation. The current PyTorch install supports CUDA capabilities sm_37 sm_50 sm_60 sm_70. If you want to use the NVIDIA GeForce RTX 3090 GPU with PyTorch, please check the instructions at https://pytorch.org/get-started/locally/
warnings.warn(incompatible_device_warn.format(device_name, capability, " ".join(arch_list), device_name))
[12/10/22 17:35:29][pytorch_lightning.utilities.distributed][INFO] - GPU available: True, used: True
[12/10/22 17:35:29][pytorch_lightning.utilities.distributed][INFO] - TPU available: False, using: 0 TPU cores
[12/10/22 17:35:29][pytorch_lightning.utilities.distributed][INFO] - IPU available: False, using: 0 IPUs
[12/10/22 17:35:29][main][INFO] - Trainer initialized
[12/10/22 17:35:29][main][INFO] - Fitting the model..
Loading BABEL train: 100%|██████████████████████████████████████████████████████████████████████████████████████████| 113/113 [00:00<00:00, 563.06it/s]
[12/10/22 17:35:30][teach.data.babel][INFO] - Processed 113 sequences and found 57 invalid cases based on the datatype.
[12/10/22 17:35:30][teach.data.babel][INFO] - 122 sequences -- datatype:separate_pairs.
[12/10/22 17:35:30][teach.data.babel][INFO] - 23.27% of the sequences which are rejected by the sampler in total.
[12/10/22 17:35:30][teach.data.babel][INFO] - 0.0% of the sequence which are rejected by the sampler, because of the excluded actions.
[12/10/22 17:35:30][teach.data.babel][INFO] - 23.27% of the sequence which are rejected by the sampler, because they are too short(<0.5 secs) or too long(>25.0 secs).
[12/10/22 17:35:30][teach.data.babel][INFO] - Discard from BML: 0
[12/10/22 17:35:30][teach.data.babel][INFO] - Discard not KIT: 0
Error executing job with overrides: ['experiment=baseline', 'logger=none']
Traceback (most recent call last):
File "/home/beibeigh/Research/Projects/teach/train.py", line 140, in
Besides, i don't konw how to get "train.json" and "val.json",and “what's the mean of babel_v2.1“(If it can create in run "python scripts/amass_splits_babel.py", however, i just get "train.pth.tar","val.pth.tar","train_tiny.pth.tar" and "val_tiny.pth.tar"). All in all , there are somethins you should do i suggest : Firstly , delete "model_type = args.model_type" in scripts/process_amass.py delete "model_type" and "model_type=body_model_type" in read_data function. Secondly,add "model_type" in get_body_model function, and "eidx-sidx" is batch_size ? and how many?
The last, i suggest you can try your project and sure it can work with your indruction.
First: The error is here and clear:
NVIDIA GeForce RTX 3090 with CUDA capability sm_86 is not compatible with the current PyTorch installation.
The current PyTorch install supports CUDA capabilities sm_37 sm_50 sm_60 sm_70.
If you want to use the NVIDIA GeForce RTX 3090 GPU with PyTorch, please check the instructions at https://pytorch.org/get-started/locally/
You should try my code with another version of pytorch that supports your GPU or another GPU. Second: Since you have run the amass preprocessing. The instructions are clear:
Download the data from TEACH website, after signing in. (...) Finally, download the male SMPLH body model from the SMPLX website in pickle format. The run this script and change your paths accordingly inside it extract the different babel splits from amass:
python scripts/amass_splits_babel.py
changing the paths would be straightforward and is based to where your data are.
I have another problem,when i run"python sample_seq.py folder=/path/to/experiment align=full slerp_ws=8" and i can't find
"test_set_seqs_nowalk, test_set_seqs_walk" in "teach.utils.inference".
(teach-env) (base) beibeigh@beibeigh-System-Product-Name:~/Research/Projects/teach$ python sample_seq.py folder=/path/to/experiment align=full slerp_ws=8
Traceback (most recent call last):
File "/home/beibeigh/Research/Projects/teach/sample_seq.py", line 29, in
Sorry I was going to release instruction about running sampling and evaluation. But I made a commit that fixes this you can just delete those imports.
Thanks for your reply.
But,there is another problem when i run "python sample_seq.py folder=/path/to/experiment align=full slerp_ws=8
"
(teach-env) (base) beibeigh@beibeigh-System-Product-Name:~/Research/Projects/teach$ python sample_seq.py folder=/path/to/experiment align=full slerp_ws=8
Traceback (most recent call last):
File "/home/beibeigh/Research/Projects/teach/sample_seq.py", line 30, in
Hi, I don't konw which kind(SMPL+HG, SMPL+XG,or Render) dataset on AMASS should be down, and how to "in a pickle format"?