howkewlisthat commited on
Commit
47bae96
1 Parent(s): 8e27ffa

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +21 -0
README.md CHANGED
@@ -33,3 +33,24 @@ configs:
33
  - split: validation
34
  path: data/validation-*
35
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
33
  - split: validation
34
  path: data/validation-*
35
  ---
36
+
37
+ Dataset used for training text to sql.
38
+ I've pre-tokenized this for faster loading.
39
+
40
+
41
+ Here is the tokenizer code:
42
+ ```
43
+ def tokenize_function(example):
44
+ start_prompt = "Tables:\n"
45
+ middle_prompt = "\n\nQuestion:\n"
46
+ end_prompt = "\n\nAnswer:\n"
47
+
48
+ data_zip = zip(example['context'], example['question'])
49
+ prompt = [start_prompt + context + middle_prompt + question + end_prompt for context, question in data_zip]
50
+ example['input_ids'] = tokenizer(prompt, padding="max_length", truncation=True, return_tensors="pt").input_ids
51
+ example['labels'] = tokenizer(example['answer'], padding="max_length", truncation=True, return_tensors="pt").input_ids
52
+ # print(prompt[0])
53
+ # print()
54
+
55
+ return example
56
+ ```