Delete dataset.py

dataset.py

@@ -1,67 +0,0 @@
-import json
-import os
-import re
-import torch
-
-
-def read_jsonl(path: str):
-    with open(path) as f:
-        data = json.load(f)
-        return data
-
-
-def get_data(path):
-    '''
-        If there are many json files to load:
-        path_item = os.path.join("data/", f"{path}.jsonl")
-        data_item = read_jsonl(path_item)
-    ''' 
-    data = read_jsonl(path)
-
-    for ex in data:
-        if ex["options"] != "":
-            ex.update(Problem=ex["Problem"] + "\n Options: " + ex["options"])
-        else:
-            ex.update(Problem=ex["Problem"] + "\n" + "The answer to the question is:")
-        #ex.update(answer=ex["answer"] + "<|endoftext|>")
-        # Update the diagramRef tokenizer method here
-
-    print(f"Loaded {len(data)} examples")
-    return data
-
-class TestDataset(torch.utils.data.Dataset):
-    '''
-        tokenizer: is a text -> token converter for model input
-        data: is the data loaded from the "get_data" function
-    '''
-    def __init__(self, tokenizer, data, loss_on_prefix=True):
-        self.data = data
-        self.problems = [ex["Problem"] for ex in self.data]
-        self.answer = [ex["answer"] for ex in self.data]
-        self.problems = tokenizer(self.problems, padding=False)
-        self.answer = tokenizer(self.answer, padding=False)
-        self.loss_on_prefix = loss_on_prefix
-        self.max_len_of_items = max(
-            [
-                len(self.problems["input_ids"][i]) + len(self.answer["input_ids"][i])
-                for i in range(len(self.data))
-            ]
-        )
-        #print(f"Max tokens: {self.max_len}")
-
-    def __len__(self):
-        return len(self.data)
-
-    def __getitem__(self, idx):
-        qn_tokens = self.problems["input_ids"][idx]
-        answer_tokens = self.answer["input_ids"][idx]
-        pad_tokens = [0] * (self.max_len_of_items - len(qn_tokens) - len(answer_tokens))
-        tokens = qn_tokens + answer_tokens + pad_tokens
-        mask = (
-            ([int(self.loss_on_prefix)] * len(qn_tokens))
-            + ([1] * len(answer_tokens))
-            + ([0] * len(pad_tokens))
-        )
-        tokens = torch.tensor(tokens)
-        mask = torch.tensor(mask)
-        return dict(input_ids=tokens, attention_mask=mask)