File size: 3,091 Bytes
6d446a9 5a93818 e1364fa 4f6bbf5 229c036 3c5e719 998a2d3 229c036 1ba6990 6d446a9 5a93818 6d446a9 a3cc44d 6d446a9 229c036 09ef8f6 229c036 09ef8f6 afcf4b8 c06880d afcf4b8 1f24039 c06880d afcf4b8 c06880d afcf4b8 09ef8f6 411b5df |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 |
---
license: mit
datasets:
- b-mc2/sql-create-context
- gretelai/synthetic_text_to_sql
language:
- en
base_model: google-t5/t5-base
metrics:
- exact_match
model-index:
- name: juanfra218/text2sql
results:
- task:
type: text-to-sql
metrics:
- name: exact_match
type: exact_match
value: 0.4326836917562724
- name: bleu
type: bleu
value: 0.6687
tags:
- sql
library_name: transformers
---
# Fine-Tuned Google T5 Model for Text to SQL Translation
A fine-tuned version of the Google T5 model, trained for the task of translating natural language queries into SQL statements.
## Model Details
- **Architecture**: Google T5 Base (Text-to-Text Transfer Transformer)
- **Task**: Text to SQL Translation
- **Fine-Tuning Datasets**:
- [sql-create-context Dataset](https://huggingface.co/datasets/b-mc2/sql-create-context)
- [Synthetic-Text-To-SQL Dataset](https://huggingface.co/datasets/gretelai/synthetic-text-to-sql)
## Training Parameters
```
training_args = Seq2SeqTrainingArguments(
output_dir="./results",
evaluation_strategy="epoch",
learning_rate=2e-5,
per_device_train_batch_size=8,
per_device_eval_batch_size=8,
weight_decay=0.01,
save_total_limit=3,
num_train_epochs=3,
predict_with_generate=True,
fp16=True,
push_to_hub=False,
)
```
## Usage
```
import torch
from transformers import T5Tokenizer, T5ForConditionalGeneration
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
# Load the tokenizer and model
model_path = 'juanfra218/text2sql'
tokenizer = T5Tokenizer.from_pretrained(model_path)
model = T5ForConditionalGeneration.from_pretrained(model_path)
model.to(device)
# Function to generate SQL queries
def generate_sql(prompt, schema):
input_text = "translate English to SQL: " + prompt + " " + schema
inputs = tokenizer(input_text, return_tensors="pt", max_length=512, truncation=True, padding="max_length")
inputs = {key: value.to(device) for key, value in inputs.items()}
max_output_length = 1024
outputs = model.generate(**inputs, max_length=max_output_length)
return tokenizer.decode(outputs[0], skip_special_tokens=True)
# Interactive loop
print("Enter 'quit' to exit.")
while True:
prompt = input("Insert prompt: ")
schema = input("Insert schema: ")
if prompt.lower() == 'quit':
break
sql_query = generate_sql(prompt, schema)
print(f"Generated SQL query: {sql_query}")
print()
```
## Files
- `optimizer.pt`: State of the optimizer.
- `training_args.bin`: Training arguments and hyperparameters.
- `tokenizer.json`: Tokenizer vocabulary and settings.
- `spiece.model`: SentencePiece model file.
- `special_tokens_map.json`: Special tokens mapping.
- `tokenizer_config.json`: Tokenizer configuration settings.
- `model.safetensors`: Trained model weights.
- `generation_config.json`: Configuration for text generation.
- `config.json`: Model architecture configuration.
- `test_results.csv`: Results on the testing set, contains: prompt, context, true_answer, predicted_answer, exact_match |