init

experiments/main.sh

@@ -12,16 +12,16 @@ MODEL="jhu-clsp/bernice"
 
 
 # NER
-python model_finetuning_ner.py -m "${MODEL}" -d "ner_temporal"
+python model_finetuning_ner.py -m "${MODEL}" -d "ner_temporal" --skip-train --skip-test
 python model_finetuning_ner.py -m "${MODEL}" -d "ner_random0_seed0"
 python model_finetuning_ner.py -m "${MODEL}" -d "ner_random1_seed0"
 python model_finetuning_ner.py -m "${MODEL}" -d "ner_random2_seed0"
 python model_finetuning_ner.py -m "${MODEL}" -d "ner_random3_seed0"
-python model_finetuning_ner.py -m "${MODEL}" -d "ner_random0_seed1"
-python model_finetuning_ner.py -m "${MODEL}" -d "ner_random1_seed1"
-python model_finetuning_ner.py -m "${MODEL}" -d "ner_random2_seed1"
+python model_finetuning_ner.py -m "${MODEL}" -d "ner_random0_seed1" --skip-train --skip-test
+python model_finetuning_ner.py -m "${MODEL}" -d "ner_random1_seed1" --skip-train --skip-test
+python model_finetuning_ner.py -m "${MODEL}" -d "ner_random2_seed1" --skip-train --skip-test
 python model_finetuning_ner.py -m "${MODEL}" -d "ner_random3_seed1"
-python model_finetuning_ner.py -m "${MODEL}" -d "ner_random0_seed2"
+python model_finetuning_ner.py -m "${MODEL}" -d "ner_random0_seed2" --skip-train --skip-test
 python model_finetuning_ner.py -m "${MODEL}" -d "ner_random1_seed2"
 python model_finetuning_ner.py -m "${MODEL}" -d "ner_random2_seed2"
 python model_finetuning_ner.py -m "${MODEL}" -d "ner_random3_seed2"

process/tweet_nerd.py

@@ -23,8 +23,13 @@ while True:
     if dist_date[:n].sum() > total_n/2:
         break
 split_date = dist_date.index[n]
+input(split_date)
+
 train = df[df["date_dt"] <= split_date]
 test = df[df["date_dt"] > split_date]
+print(train.date_dt.min(), train.date_dt.max())
+print(test.date_dt.min(), test.date_dt.max())
+input()
 train.pop("date_dt")
 test.pop("date_dt")
 train = list(train.T.to_dict().values())

statistics.py

@@ -5,23 +5,25 @@ from transformers import AutoTokenizer
 
 tokenizer = AutoTokenizer.from_pretrained("roberta-base")
 stats = []
-for i in ["topic_temporal", "nerd_temporal", "ner_temporal"]:
+for i in ["nerd_temporal", "ner_temporal", "topic_temporal"]:
     # for s in ["train", "validation", "test", "test_1", "test_2", "test_3", "test_4"]:
     for s in ["train", "validation", "test"]:
-        dataset = load_dataset("tweettemposhift/tweet_temporal_shift", i, split=s, download_mode="force_redownload")
+        dataset = load_dataset("tweettemposhift/tweet_temporal_shift", i, split=s)
         df = dataset.to_pandas()
         if i != "nerd_temporal":
             token_length = [len(tokenizer.tokenize(t)) for t in dataset['text']]
         else:
             token_length = [len(tokenizer.tokenize(f"{d['target']} {tokenizer.sep_token} {d['definition']} {tokenizer.sep_token} {d['text']}")) for d in dataset]
         token_length_in = [i for i in token_length if i <= 126]
+        date = pd.to_datetime(df.date).sort_values().values
         stats.append({
             "data": i,
             "split": s,
             "size": len(dataset),
             "size (token length < 128)": len(token_length_in),
             "mean_token_length": sum(token_length)/len(token_length),
-            "date": f'{str(pd.to_datetime(df.date).min()).split(" ")[0]} / {str(pd.to_datetime(df.date).max()).split(" ")[0]}',
+            "date": f'{str(date[0]).split(" ")[0]} / {str(date[-1]).split(" ")[0]}',
          })
+    break
 df = pd.DataFrame(stats)
 print(df)