Implement relative importing of data

.gitignore

@@ -136,3 +136,6 @@ fingpt/FinGPT_sentiment/instruct-FinGPT/run.sh
 fingpt/FinGPT_sentiment/instruct-FinGPT/checkpoints
 fingpt/FinGPT_sentiment/instruct-FinGPT/ds_results_all_10_v2_1.*
 FinGPT_Training_LoRA_with_Chatglm2_6b_for_beginners.ipynb
+
+# Benchmark data
+fingpt/FinGPT_Benchmark/data/*/**

fingpt/FinGPT_Benchmark/benchmarks/convfinqa.py

@@ -8,9 +8,9 @@ from functools import partial
 import re
 import sys
 import numpy as np
+from fingpt.FinGPT_Benchmark.utils import *
+from pathlib import Path
 sys.path.append('../')
-from utils import *
-    
 
 def cvt_text_to_pred(text):
     if not text:
@@ -32,9 +32,8 @@ def map_output(feature):
     return {'label': label, 'pred': pred}
 
 
-def test_convfinqa(args, model, tokenizer):
-
-    dataset = load_from_disk('../data/fingpt-convfinqa')['test']#.select(range(30))
+    
+    dataset = load_from_disk(Path(__file__).parent.parent / 'data/fingpt-convfinqa')['test']
     dataset = dataset.map(partial(test_mapping, args), load_from_cache_file=False)
     
     def collate_fn(batch):

fingpt/FinGPT_Benchmark/benchmarks/fineval.py

@@ -8,8 +8,9 @@ from functools import partial
 import re
 import sys
 import numpy as np
+from fingpt.FinGPT_Benchmark.utils import *
+from pathlib import Path
 sys.path.append('../')
-from utils import *
     
 
 def cvt_text_to_pred(text):
@@ -33,7 +34,7 @@ def map_output(feature):
 
 def test_fineval(args, model, tokenizer):
 
-    dataset = load_from_disk('../data/fingpt-fineval')['test']#.select(range(30))
+    dataset = load_from_disk(Path(__file__).parent.parent / 'data/fingpt-fineval')['test']
     dataset = dataset.map(partial(test_mapping, args), load_from_cache_file=False)
     
     def collate_fn(batch):

fingpt/FinGPT_Benchmark/benchmarks/finred.py

@@ -8,8 +8,9 @@ from functools import partial
 import re
 import sys
 import numpy as np
+from fingpt.FinGPT_Benchmark.utils import *
+from pathlib import Path
 sys.path.append('../')
-from utils import *
     
 
 relations = [
@@ -102,7 +103,7 @@ def calc_metric(gt_list, pred_list):
 
 def test_re(args, model, tokenizer):
 
-    dataset = load_from_disk('../data/fingpt-finred-re')['test']#.select(range(50))
+    dataset = load_from_disk(Path(__file__).parent.parent / 'data/fingpt-finred-re')['test']
     dataset = dataset.train_test_split(0.2, seed=42)['test']
     dataset = dataset.map(partial(test_mapping, args), load_from_cache_file=False)
     

fingpt/FinGPT_Benchmark/benchmarks/fiqa.py

@@ -9,9 +9,10 @@ import torch
 
 from torch.utils.data import DataLoader
 from functools import partial
+from pathlib import Path
 
 
-with open('sentiment_templates.txt') as f:
+with open(Path(__file__).parent / 'sentiment_templates.txt') as f:
     templates = [l.strip() for l in f.readlines()]
     
 
@@ -58,7 +59,7 @@ def vote_output(x):
 def test_fiqa(args, model, tokenizer, prompt_fun=add_instructions):
     batch_size = args.batch_size
     # dataset = load_dataset('pauri32/fiqa-2018')
-    dataset = load_from_disk('../data/fiqa-2018/')
+    dataset = load_from_disk(Path(__file__).parent.parent / 'data/fiqa-2018/')
     dataset = datasets.concatenate_datasets([dataset["train"], dataset["validation"] ,dataset["test"] ])
     dataset = dataset.train_test_split(0.226, seed = 42)['test']
     dataset = dataset.to_pandas()
@@ -112,7 +113,7 @@ def test_fiqa(args, model, tokenizer, prompt_fun=add_instructions):
 def test_fiqa_mlt(args, model, tokenizer):
     batch_size = args.batch_size
     # dataset = load_dataset('pauri32/fiqa-2018')
-    dataset = load_from_disk('../data/fiqa-2018/')
+    dataset = load_from_disk(Path(__file__).parent.parent / 'data/fiqa-2018/')
     dataset = datasets.concatenate_datasets([dataset["train"], dataset["validation"] ,dataset["test"] ])
     dataset = dataset.train_test_split(0.226, seed=42)['test']
     dataset = dataset.to_pandas()

fingpt/FinGPT_Benchmark/benchmarks/fpb.py

@@ -9,6 +9,7 @@ import torch
 
 from torch.utils.data import DataLoader
 from functools import partial
+from pathlib import Path
 
 dic = {
         0:"negative",
@@ -16,7 +17,7 @@ dic = {
         2:'positive',
     }
 
-with open('sentiment_templates.txt') as f:
+with open(Path(__file__).parent / 'sentiment_templates.txt') as f:
     templates = [l.strip() for l in f.readlines()]
     
 
@@ -52,7 +53,7 @@ def vote_output(x):
 def test_fpb(args, model, tokenizer, prompt_fun=None):
     batch_size = args.batch_size
     # instructions = load_dataset("financial_phrasebank", "sentences_50agree")
-    instructions = load_from_disk("../data/financial_phrasebank-sentences_50agree/")
+    instructions = load_from_disk(Path(__file__).parent.parent / "data/financial_phrasebank-sentences_50agree/")
     instructions = instructions["train"]
     instructions = instructions.train_test_split(seed = 42)['test']
     instructions = instructions.to_pandas()
@@ -105,8 +106,8 @@ def test_fpb(args, model, tokenizer, prompt_fun=None):
 
 def test_fpb_mlt(args, model, tokenizer):
     batch_size = args.batch_size
-    # instructions = load_dataset("financial_phrasebank", "sentences_50agree")
-    dataset = load_from_disk('../data/financial_phrasebank-sentences_50agree/')
+    # dataset = load_dataset("financial_phrasebank", "sentences_50agree")
+    dataset = load_from_disk(Path(__file__).parent.parent / 'data/financial_phrasebank-sentences_50agree/')
     dataset = dataset["train"]#.select(range(300))
     dataset = dataset.train_test_split(seed=42)['test']
     dataset = dataset.to_pandas()

fingpt/FinGPT_Benchmark/benchmarks/headline.py

@@ -5,11 +5,11 @@ import datasets
 import torch
 from torch.utils.data import DataLoader
 from functools import partial
-
+from pathlib import Path
+from fingpt.FinGPT_Benchmark.utils import *
 
 import sys
 sys.path.append('../')
-from utils import *
     
     
     
@@ -34,8 +34,8 @@ def map_output(feature):
 
 def test_headline(args, model, tokenizer):
     
-    # dataset = load_from_disk('../data/fingpt-headline')['test']#.select(range(300))
-    dataset = load_from_disk('../data/fingpt-headline-instruct')['test']#.select(range(300))
+    # dataset = load_from_disk('../data/fingpt-headline')['test']
+    dataset = load_from_disk(Path(__file__).parent.parent / 'data/fingpt-headline-instruct')['test']
     dataset = dataset.map(partial(test_mapping, args), load_from_cache_file=False)
     
     def collate_fn(batch):

fingpt/FinGPT_Benchmark/benchmarks/ner.py

@@ -8,9 +8,9 @@ from functools import partial
 import re
 import sys
 import numpy as np
+from fingpt.FinGPT_Benchmark.utils import *
+from pathlib import Path
 sys.path.append('../')
-from utils import *
-    
 
 ent_dict = {
     'PER': 'person',
@@ -53,7 +53,7 @@ def map_output(feature):
 
 def test_ner(args, model, tokenizer):
 
-    dataset = load_from_disk('../data/fingpt-ner')['test']#.select(range(30))
+    dataset = load_from_disk(Path(__file__).parent.parent / 'data/fingpt-ner')['test']
     dataset = dataset.map(partial(test_mapping, args), load_from_cache_file=False)
     
     def collate_fn(batch):

fingpt/FinGPT_Benchmark/benchmarks/nwgi.py

@@ -6,6 +6,7 @@ from datasets import load_dataset, load_from_disk
 from tqdm import tqdm
 import datasets
 import torch
+from pathlib import Path
 
 dic = {
     'strong negative':"negative",
@@ -36,8 +37,7 @@ def change_target(x):
 def test_nwgi(args, model, tokenizer, prompt_fun=None):
     batch_size = args.batch_size
     # dataset = load_dataset('oliverwang15/news_with_gpt_instructions')
-    dataset = load_from_disk('../data/news_with_gpt_instructions/')
-    dataset = dataset['test'].to_pandas()
+    dataset = load_from_disk(Path(__file__).parent.parent / 'data/news_with_gpt_instructions/')
     dataset['output'] = dataset['label'].apply(lambda x:dic[x])
 
     if prompt_fun is None:

fingpt/FinGPT_Benchmark/benchmarks/tfns.py

@@ -6,6 +6,7 @@ from datasets import load_dataset, load_from_disk
 from tqdm import tqdm
 import datasets
 import torch
+from pathlib import Path
 
 dic = {
     0:"negative",
@@ -32,7 +33,7 @@ def change_target(x):
 def test_tfns(args, model, tokenizer, prompt_fun=None):
     batch_size = args.batch_size
     # dataset = load_dataset('zeroshot/twitter-financial-news-sentiment')
-    dataset = load_from_disk('../data/twitter-financial-news-sentiment')
+    dataset = load_from_disk(Path(__file__).parent.parent / 'data/twitter-financial-news-sentiment')
     dataset = dataset['validation']
     dataset = dataset.to_pandas()
     dataset['label'] = dataset['label'].apply(lambda x:dic[x])

fingpt/FinGPT_Benchmark/data/download.py

@@ -0,0 +1,57 @@
+import datasets
+from pathlib import Path
+
+def download():
+    """Downloads all datasets to where the FinGPT library is located."""
+    data_dir = Path(__file__).parent
+    
+    dataset = datasets.load_dataset('pauri32/fiqa-2018')
+    dataset.save_to_disk(data_dir / 'fiqa-2018')
+    
+    dataset = datasets.load_dataset('FinGPT/fingpt-finred')
+    dataset.save_to_disk(data_dir / 'fingpt-finred')
+    
+    dataset = datasets.load_dataset('zeroshot/twitter-financial-news-sentiment')
+    dataset.save_to_disk(data_dir / 'twitter-financial-news-sentiment')
+    
+    dataset = datasets.load_dataset('oliverwang15/news_with_gpt_instructions')
+    dataset.save_to_disk(data_dir / 'news_with_gpt_instructions')
+
+    dataset = datasets.load_dataset("financial_phrasebank", "sentences_50agree")
+    dataset.save_to_disk(data_dir / 'financial_phrasebank-sentences_50agree')
+
+    dataset = datasets.load_dataset('FinGPT/fingpt-fiqa_qa')
+    dataset.save_to_disk(data_dir / 'fingpt-fiqa_qa')
+
+    dataset = datasets.load_dataset('FinGPT/fingpt-headline-cls')
+    dataset.save_to_disk(data_dir / 'fingpt-headline-cls')
+
+    dataset = datasets.load_dataset('FinGPT/fingpt-finred')
+    dataset.save_to_disk(data_dir / 'fingpt-finred')
+
+    dataset = datasets.load_dataset('FinGPT/fingpt-convfinqa')
+    dataset.save_to_disk(data_dir / 'fingpt-convfinqa')
+
+    dataset = datasets.load_dataset('FinGPT/fingpt-finred-cls')
+    dataset.save_to_disk(data_dir / 'fingpt-finred-cls')
+
+    dataset = datasets.load_dataset('FinGPT/fingpt-ner')
+    dataset.save_to_disk(data_dir / 'fingpt-ner')
+
+    dataset = datasets.load_dataset('FinGPT/fingpt-headline')
+    dataset.save_to_disk(data_dir / 'fingpt-headline-instruct')
+
+    dataset = datasets.load_dataset('FinGPT/fingpt-finred-re')
+    dataset.save_to_disk(data_dir / 'fingpt-finred-re')
+
+    dataset = datasets.load_dataset('FinGPT/fingpt-ner-cls')
+    dataset.save_to_disk(data_dir / 'fingpt-ner-cls')
+
+    dataset = datasets.load_dataset('FinGPT/fingpt-fineval')
+    dataset.save_to_disk(data_dir / 'fingpt-fineval')
+
+    dataset = datasets.load_dataset('FinGPT/fingpt-sentiment-cls')
+    dataset.save_to_disk(data_dir / 'fingpt-sentiment-cls')
+
+if __name__ == "__main__":
+    download()