fix some typo

README.md

@@ -47,7 +47,8 @@ Then, use the PTMs as the following example, where `MODEL_NAME` is the correspon
 
 For CPT:
 ```python
-from modeling_cpt import BertTokenizer, CPTForConditionalGeneration
+from modeling_cpt import CPTForConditionalGeneration
+from transformers import BertTokenizer
 tokenizer = BertTokenizer.from_pretrained("MODEL_NAME")
 model = CPTForConditionalGeneration.from_pretrained("MODEL_NAME")
 print(model)
@@ -63,9 +64,9 @@ print(model)
 
 After initializing the model, you can use the following lines to generate text.
 ```python
->>> inputs = tokenizer.encode("北京是[MASK]的首都", return_tensors='pt')
+>>> input_ids = tokenizer.encode("北京是[MASK]的首都", return_tensors='pt')
 >>> pred_ids = model.generate(input_ids, num_beams=4, max_length=20)
->>> print(tokenizer.convert_ids_to_tokens(pred_ids[i]))
+>>> print(tokenizer.convert_ids_to_tokens(pred_ids[0]))
     ['[SEP]', '[CLS]', '北', '京', '是', '中', '国', '的', '首', '都', '[SEP]']
 ```
 

finetune/README.md

@@ -5,6 +5,8 @@ This repo contains the fine-tuning code for CPT on multiple NLU and NLG tasks, s
 ## Requirements
 - pytorch==1.8.1
 - transformers==4.4.1
+- fitlog
+- fastNLP
 
 ## Run
 The code and running examples are listed in the corresponding folders of the fine-tuning tasks.
@@ -18,7 +20,8 @@ The code and running examples are listed in the corresponding folders of the fin
 You can also fine-tuning CPT on other tasks by adding `modeling_cpt.py` into your project and use the following code to use CPT.
 
 ```python
-from modeling_cpt import BertTokenizer, CPTForConditionalGeneration
+from modeling_cpt import CPTForConditionalGeneration
+from transformers import BertTokenizer
 tokenizer = BertTokenizer.from_pretrained("MODEL_NAME")
 model = CPTForConditionalGeneration.from_pretrained("MODEL_NAME")
 print(model)

finetune/mrc/README.md

@@ -9,7 +9,7 @@ To train and evaluate **CPT$_u$**, **CPT$_g$** and **CPT$_{ug}$**, run the pytho
 ```bash
 export MODEL_TYPE=cpt-base
 export MODEL_NAME=fnlp/cpt-base
-export CLUE_DATA_DIR=/path/to/mrc_data_dir
+export CLUE_DATA_DIR=~/workdir/datasets/CLUEdatasets/
 export TASK_NAME=drcd
 export CLS_MODE=1
 python run_mrc.py \
@@ -22,7 +22,7 @@ python run_mrc.py \
   --gradient_accumulation_steps 4 \
   --lr=3e-5 \
   --dropout=0.2 \
-  --CLS_MODE=$CLS_MODE \
+  --cls_mode=$CLS_MODE \
   --warmup_rate=0.1 \
   --weight_decay_rate=0.01 \
   --max_seq_length=512 \

finetune/mrc/run_mrc.py

@@ -210,7 +210,7 @@ def test(model, args, eval_examples, eval_features, device, name):
         torch.distributed.barrier()
 
     # load the bert setting
-    if 'bert' == args.model_type:
+    if 'bert' in args.model_type or 'cpt' in args.model_type:
         if 'large' in args.init_restore_dir or '24' in args.init_restore_dir:
             config_path = 'hfl/chinese-roberta-wwm-ext-large'
         else:
@@ -219,7 +219,7 @@ def test(model, args, eval_examples, eval_features, device, name):
         tokenizer = BertTokenizer.from_pretrained(config_path)
         bert_config.hidden_dropout_prob = args.dropout
         bert_config.attention_probs_dropout_prob = args.dropout
-        if 'arch' in args.init_restore_dir:
+        if 'cpt' in args.init_restore_dir:
             config = CPTConfig.from_pretrained(args.init_restore_dir)
             config.cls_mode = args.cls_mode
             config.attention_dropout = args.dropout