vanint / datasetexpansion Goto Github PK

Hello seniors, can this data augmentation method be used in the field of speech.

Hello, I would like to ask if the environment for this code has requirements for torch and python.Thanks

Hi,How should I train my dataset?

Hello, I would like to confirm that the "Dataset Expansion" step is for training. It does not directly provide us with the checkpoint after training. Is that so?

Hi!

I am wondering how you save MedMNIST dataset on your local directory.
To my understanding, we can download .npz file for each dataset from the official website(https://zenodo.org/records/10519652).
Based on your implementation, it seems you saved them as .png or .jpg files in the different folders for each label.
Do you have the code to share?

Is it possible to perform data augmentation with a homemade railroad scene dataset, and if so, how should it be done? Thanks.

if we need to fine-tune the stable diffusion for datasets in special scenarios .
Thank you

Hello.

I am currently trying your work on medical domain based on Dreambooth.

But there are a lot of missing details on finetuning with SD in this paper (Only number of training image mentioned).

Can you tell about more details based on the repository above? (e.g. class word, prompt, training time, etc.)

Hi, thank you for your work. Initialize GIF_ Encountered some issues while modeling. It cannot your code run directly.
when I run the following:

model = instantiate_from_config(config.model)

I got the following error message:
/data/zhuhaowei/anaconda/anaconda3/envs/data-expand/lib/python3.10/site-packages/torchvision/transforms/transforms.py:329: UserWarning: Argument 'interpolation' of type int is deprecated since 0.13 and will be removed in 0.15. Please use InterpolationMode enum. warnings.warn( /data/zhuhaowei/anaconda/anaconda3/envs/data-expand/lib/python3.10/site-packages/pytorch_lightning/utilities/distributed.py:258: LightningDeprecationWarning: pytorch_lightning.utilities.distributed.rank_zero_onlyhas been deprecated in v1.8.1 and will be removed in v2.0.0. You can import it frompytorch_lightning.utilitiesinstead. rank_zero_deprecation( ==> Preparing dataset fgvc_aircraft Load fgvc_aircraft data finished. ==> creating model 'CLIP-VIT-B32' Model CLIP loaded. Global Step: 470000 LatentDiffusion: Running in eps-prediction mode DiffusionWrapper has 859.52 M params. Working with z of shape (1, 4, 32, 32) = 4096 dimensions. Traceback (most recent call last): File "/data/zhuhaowei/code/DatasetExpansion-main/GIF_SD/custom/dataset_expansion_stable_diffusion_CLIP_batch_optimization_final.py", line 732, in <module> main() File "/data/zhuhaowei/code/DatasetExpansion-main/GIF_SD/custom/dataset_expansion_stable_diffusion_CLIP_batch_optimization_final.py", line 555, in main GIF_model = load_model_from_config(config, f"{args.ckpt}") File "/data/zhuhaowei/code/DatasetExpansion-main/GIF_SD/custom/dataset_expansion_stable_diffusion_CLIP_batch_optimization_final.py", line 445, in load_model_from_config model = instantiate_from_config(config.model) File "/data/zhuhaowei/code/DatasetExpansion-main/GIF_SD/custom/ldm/util.py", line 85, in instantiate_from_config return get_obj_from_str(config["target"])(**config.get("params", dict())) File "/data/zhuhaowei/code/DatasetExpansion-main/GIF_SD/custom/ldm/models/diffusion/ddpm.py", line 461, in __init__ self.instantiate_cond_stage(cond_stage_config) File "/data/zhuhaowei/code/DatasetExpansion-main/GIF_SD/custom/ldm/models/diffusion/ddpm.py", line 519, in instantiate_cond_stage model = instantiate_from_config(config) File "/data/zhuhaowei/code/DatasetExpansion-main/GIF_SD/custom/ldm/util.py", line 85, in instantiate_from_config return get_obj_from_str(config["target"])(**config.get("params", dict())) File "/data/zhuhaowei/code/DatasetExpansion-main/GIF_SD/custom/ldm/modules/encoders/modules.py", line 141, in __init__ self.tokenizer = CLIPTokenizer.from_pretrained(version) File "/data/zhuhaowei/anaconda/anaconda3/envs/data-expand/lib/python3.10/site-packages/transformers/tokenization_utils_base.py", line 1784, in from_pretrained return cls._from_pretrained( File "/data/zhuhaowei/anaconda/anaconda3/envs/data-expand/lib/python3.10/site-packages/transformers/tokenization_utils_base.py", line 1825, in _from_pretrained init_kwargs = json.load(tokenizer_config_handle) File "/data/zhuhaowei/anaconda/anaconda3/envs/data-expand/lib/python3.10/json/__init__.py", line 293, in load return loads(fp.read(), File "/data/zhuhaowei/anaconda/anaconda3/envs/data-expand/lib/python3.10/json/__init__.py", line 346, in loads return _default_decoder.decode(s) File "/data/zhuhaowei/anaconda/anaconda3/envs/data-expand/lib/python3.10/json/decoder.py", line 337, in decode obj, end = self.raw_decode(s, idx=_w(s, 0).end()) File "/data/zhuhaowei/anaconda/anaconda3/envs/data-expand/lib/python3.10/json/decoder.py", line 355, in raw_decode raise JSONDecodeError("Expecting value", s, err.value) from None json.decoder.JSONDecodeError: Expecting value: line 1 column 1 (char 0)

I think there are some issues with my environment configuration. Now I am using transformers=4.19.2, torch=1.13.1, torch vision=0.14.1, and pytorch_ Lighting=1.9.0.

Can you provide a more detailed version of the environment? Thanks!

I trained on the cifar dataset, but encountered some problems. In dataset_expansion_stable_diffusion_CLIP_batch_optimization_final, bugs appear at this location in the program.

The model has problems in calculating the gradient, which hopefully can be solved later.

Hi, do I need to fine tune CLIP on professional datasets?