Open pkhk-1 opened 1 year ago
Thank you for your work. When I tried to use my own model for benchmark experiments with reference to open_clip.py, an error occurred during model.eval(). I also encountered the same error when running the ja_clip example provided in the readme.
python3 clip_benchmark/cli.py eval \ --model_type "ja_clip" \ --pretrained "rinna/japanese-cloob-vit-b-16" \ --language "jp" \ --task "zeroshot_classification" \ --dataset "imagenet1k" \ Error:
python3 clip_benchmark/cli.py eval \ --model_type "ja_clip" \ --pretrained "rinna/japanese-cloob-vit-b-16" \ --language "jp" \ --task "zeroshot_classification" \ --dataset "imagenet1k" \
I used weights that are consistent with open_clip, but only replaced the .bin and config files. Can you advise me on how to properly load the weights?
It seems like JaCLIPForBenchmark is not an torch.nn.Module, can you maybe share your code ?
torch.nn.Module
Thank you for your work. When I tried to use my own model for benchmark experiments with reference to open_clip.py, an error occurred during model.eval(). I also encountered the same error when running the ja_clip example provided in the readme.
python3 clip_benchmark/cli.py eval \ --model_type "ja_clip" \ --pretrained "rinna/japanese-cloob-vit-b-16" \ --language "jp" \ --task "zeroshot_classification" \ --dataset "imagenet1k" \
Error:I used weights that are consistent with open_clip, but only replaced the .bin and config files. Can you advise me on how to properly load the weights?