fix

README.md

@@ -8,7 +8,7 @@ Git clone our repository, creating a python environment and activate it via the
 
 ```bash
 git clone https://github.com/DoubleClass/GMM
-cd MiniGPT-4
+cd GMM
 conda env create -n GMM python=3.9
 conda install --yes --file requirements.txt
 conda activate GMM
@@ -22,17 +22,17 @@ You can get the LLM Vicuna in [huggingface](https://huggingface.co/meta-llama/Ll
 Then set the downloaded vicuna folder path [here](minigpt4/configs/models/minigpt4_vicuna0.yaml) and the initial checkpoint [here](train_configs/minigpt4_stage2_finetune.yaml#L9)
 
 ### EVA_VIT_G
-The code will automatically downloading the eva_vit_g.pth, we alse put it [here](https://pan.baidu.com/s/1kyc6gp7f2CXkocljhERKVg?pwd=2mux), you can manually download it and put it in 'root/.cache/torch/hub/checkpoints/eva_vit_g.pth'
+The code will automatically downloading the eva_vit_g.pth, we alse put it [here](https://pan.baidu.com/s/1kyc6gp7f2CXkocljhERKVg?pwd=2mux), you can manually download it and put it in the cache dir
 
 ### bert-base-uncased
-The code will automatically downloading this, but in case you don't have access to huggingface, we also put it [here](https://pan.baidu.com/s/1XzAidcFinjsNxdz58M465w?pwd=b98f), you can manually download it and put it in '~/.cache/huggingface/hub/models--bert-base-uncased'
+The code will automatically downloading this, but in case you don't have access to huggingface, we also put it [here](https://pan.baidu.com/s/1XzAidcFinjsNxdz58M465w?pwd=b98f), you can manually download it and alse put it in cache dir
 ### datasets
 #### ImageNet-R
 You can download it [here](https://people.eecs.berkeley.edu/~hendrycks/imagenet-r.tar)
 
-Then set the dataset folder path [here](clip_base/datasets.py#L281)
+Then set the dataset folder path [here](clip_base/datasets.py#L134)
 
-Besides, you need to customize the dataset for the GPT fine-tuning process. We prepare a example here you can follow. [baidu](https://pan.baidu.com/s/1xMkqOiSylWyKY74Oef4h4g?pwd=yyea)
+Besides, you need to customize the dataset for the GPT fine-tuning process. We prepare a example here you can follow: [download](https://pan.baidu.com/s/1xMkqOiSylWyKY74Oef4h4g?pwd=yyea)
 
 After downloaded the customized dataset, you can set the data root path [here](minigpt4/configs/datasets/cc_sbu/align.yaml#L7) and the indexing file [here](minigpt4/datasets/builders/image_text_pair_builder.py#L121)
 
@@ -48,6 +48,8 @@ python train.py --cfg-path train_configs/minigpt4_stage2_finetune.yaml
 ## Testing
 After training, you will get a model checkpoint of the last continual learning stage. put the path to scipts in eval_all.sh and specify a results directory.
 
+Then set the results path in the [get_score_all.py](https://vscode.dev/github/DoubleClass/GMM/get_score_all.py#L1)
+
 Run the script:
 
 ```bash 

batch_eval.py

@@ -22,6 +22,7 @@
 from clip_base.datasets import build_cl_scenarios
 from torch.utils.data import DataLoader
 import clip
+from tqdm import tqdm
 
 def parse_args():
     parser = argparse.ArgumentParser(description="Demo")
@@ -43,11 +44,9 @@ def parse_args():
 
 def setup_seeds(config):
     seed = config.run_cfg.seed + get_rank()
-    # import pdb; pdb.set_trace()
     random.seed(seed)
     np.random.seed(seed)
     torch.manual_seed(seed)
-
     cudnn.benchmark = False
     cudnn.deterministic = True
 
@@ -68,9 +67,6 @@ def setup_seeds(config):
 
 
 model_config = cfg.model_cfg
-# import pdb; pdb.set_trace()
-
-
 model_config.device_8bit = args.gpu_id
 model_config.ckpt = args.ckpt_path
 model_cls = registry.get_model_class(model_config.arch)
@@ -99,46 +95,20 @@ def get_ordered_class_name(class_order, class_name):
         cfg_o, is_train=False, transforms=transforms
     )
 
-# import pdb; pdb.set_trace()
 new_class_name = get_ordered_class_name(cfg_o.class_order, classes_names)
 
-# import pdb; pdb.set_trace()
-
-
-
-from tqdm import tqdm
-# import warnings
-# warnings.filterwarnings("ignore")
-
 with open(args.txt_path, 'w') as f:
-
     eval_loader = DataLoader(eval_dataset[:args.task_id+1], batch_size=cfg_o.batch)
     names = new_class_name[:cfg_o.initial_increment + args.task_id * cfg_o.increment]
     for inputs, targets, task_ids in tqdm(eval_loader):
 
-
-
-
-        # import pdb; pdb.set_trace()
         chat_state = CONV_VISION.copy()
         img_list = []
-        # gr_img = torch.randn(2,3,224,224)
-        # gr_img = item[0]
         llm_message = chat.upload_img(inputs, chat_state, img_list)
-        # llm_message = chat.upload_img(gr_img, chat_state, img_list)
-        # import pdb; pdb.set_trace()
         chat.ask('what is this photo of?', chat_state)
         llm_message = chat.answer(conv=chat_state,img_list=img_list,num_beams=1,temperature=0.01,max_new_tokens=300,max_length=2000)[0]
-
-        # llm_message = chat.answer(conv=chat_state,img_list=img_list,num_beams=1,temperature=0.01, max_new_tokens=300,max_length=2000)[0]
-        # print('the label is ', new_class_name[item[1]])
-        # print('the message is ', llm_message)
         for i in range(inputs.shape[0]):
-
             str1 = 'the label is '  + new_class_name[targets[i]] + '\n'
             str2 = 'msg: '  + llm_message[i] + '\n'
-            # print(str1)
-            # print(str2)
-        # import pdb; pdb.set_trace()
             f.write(str1)
             f.write(str2)