aashish1904
commited on
Commit
•
d98d5ea
1
Parent(s):
1034578
Upload README.md with huggingface_hub
Browse files
README.md
ADDED
@@ -0,0 +1,134 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
|
2 |
+
---
|
3 |
+
|
4 |
+
language:
|
5 |
+
- en
|
6 |
+
- it
|
7 |
+
license: apache-2.0
|
8 |
+
tags:
|
9 |
+
- text-generation-inference
|
10 |
+
- transformers
|
11 |
+
- unsloth
|
12 |
+
- llama
|
13 |
+
- trl
|
14 |
+
- sft
|
15 |
+
|
16 |
+
---
|
17 |
+
|
18 |
+
[![QuantFactory Banner](https://lh7-rt.googleusercontent.com/docsz/AD_4nXeiuCm7c8lEwEJuRey9kiVZsRn2W-b4pWlu3-X534V3YmVuVc2ZL-NXg2RkzSOOS2JXGHutDuyyNAUtdJI65jGTo8jT9Y99tMi4H4MqL44Uc5QKG77B0d6-JfIkZHFaUA71-RtjyYZWVIhqsNZcx8-OMaA?key=xt3VSDoCbmTY7o-cwwOFwQ)](https://huggingface.co/QuantFactory)
|
19 |
+
|
20 |
+
|
21 |
+
# QuantFactory/Meta-Llama-3.1-8B-Text-to-SQL-GGUF
|
22 |
+
This is quantized version of [ruslanmv/Meta-Llama-3.1-8B-Text-to-SQL](https://huggingface.co/ruslanmv/Meta-Llama-3.1-8B-Text-to-SQL) created using llama.cpp
|
23 |
+
|
24 |
+
# Original Model Card
|
25 |
+
|
26 |
+
|
27 |
+
# Meta LLaMA 3.1 8B 4-bit Finetuned Model
|
28 |
+
|
29 |
+
This model is a fine-tuned version of `Meta-Llama-3.1-8B`, developed by **ruslanmv** for text generation tasks. It leverages 4-bit quantization, making it more efficient for inference while maintaining strong performance in natural language generation.
|
30 |
+
|
31 |
+
---
|
32 |
+
|
33 |
+
## Model Details
|
34 |
+
|
35 |
+
- **Base Model**: `unsloth/meta-llama-3.1-8b-bnb-4bit`
|
36 |
+
- **Finetuned by**: ruslanmv
|
37 |
+
- **Language**: English
|
38 |
+
- **License**: [Apache 2.0](https://www.apache.org/licenses/LICENSE-2.0)
|
39 |
+
- **Tags**:
|
40 |
+
- text-generation-inference
|
41 |
+
- transformers
|
42 |
+
- unsloth
|
43 |
+
- llama
|
44 |
+
- trl
|
45 |
+
- sft
|
46 |
+
|
47 |
+
---
|
48 |
+
|
49 |
+
## Model Usage
|
50 |
+
|
51 |
+
### Installation
|
52 |
+
|
53 |
+
To use this model, you will need to install the necessary libraries:
|
54 |
+
|
55 |
+
```bash
|
56 |
+
pip install transformers accelerate
|
57 |
+
```
|
58 |
+
|
59 |
+
### Loading the Model in Python
|
60 |
+
|
61 |
+
Here’s an example of how to load this fine-tuned model using Hugging Face's `transformers` library:
|
62 |
+
|
63 |
+
```python
|
64 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
65 |
+
import torch
|
66 |
+
|
67 |
+
# Load the model and tokenizer
|
68 |
+
model_name = "ruslanmv/Meta-Llama-3.1-8B-Text-to-SQL"
|
69 |
+
|
70 |
+
# Ensure you have the right device setup
|
71 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
72 |
+
|
73 |
+
# Load the model and tokenizer from the Hugging Face Hub
|
74 |
+
model = AutoModelForCausalLM.from_pretrained(model_name, device_map="auto", torch_dtype=torch.float16)
|
75 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
76 |
+
|
77 |
+
# Initialize the tokenizer (adjust the model name as needed)
|
78 |
+
# Define EOS token for terminating the sequences
|
79 |
+
EOS_TOKEN = tokenizer.eos_token
|
80 |
+
|
81 |
+
# Define Alpaca-style prompt template
|
82 |
+
alpaca_prompt = """Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.
|
83 |
+
|
84 |
+
### Instruction:
|
85 |
+
{}
|
86 |
+
|
87 |
+
### Input:
|
88 |
+
{}
|
89 |
+
|
90 |
+
### Response:
|
91 |
+
"""
|
92 |
+
|
93 |
+
# Format the prompt without the response part
|
94 |
+
prompt = alpaca_prompt.format(
|
95 |
+
"Provide the SQL query",
|
96 |
+
"Seleziona tutte le colonne della tabella table1 dove la colonna anni è uguale a 2020"
|
97 |
+
)
|
98 |
+
# Tokenize the prompt and generate text
|
99 |
+
inputs = tokenizer([prompt], return_tensors="pt").to("cuda")
|
100 |
+
outputs = model.generate(**inputs, max_new_tokens=64, use_cache=True)
|
101 |
+
|
102 |
+
# Decode the generated text
|
103 |
+
generated_text = tokenizer.batch_decode(outputs, skip_special_tokens=True)[0]
|
104 |
+
|
105 |
+
# Extract the generated response only (remove the prompt part)
|
106 |
+
response_start = generated_text.find("### Response:") + len("### Response:\n")
|
107 |
+
response = generated_text[response_start:].strip()
|
108 |
+
|
109 |
+
# Print the response (excluding the prompt)
|
110 |
+
print(response)
|
111 |
+
|
112 |
+
|
113 |
+
|
114 |
+
```
|
115 |
+
and the answer is
|
116 |
+
|
117 |
+
```
|
118 |
+
SELECT * FROM table1 WHERE anni = 2020
|
119 |
+
```
|
120 |
+
### Model Features
|
121 |
+
|
122 |
+
- **Text Generation**: This model is fine-tuned to generate coherent and contextually accurate text based on the provided input.
|
123 |
+
|
124 |
+
|
125 |
+
### License
|
126 |
+
|
127 |
+
This model is licensed under the [Apache 2.0 License](https://www.apache.org/licenses/LICENSE-2.0). You are free to use, modify, and distribute this model, provided that you comply with the license terms.
|
128 |
+
|
129 |
+
### Acknowledgments
|
130 |
+
|
131 |
+
This model was fine-tuned by **ruslanmv** based on the original work of `unsloth` and the `meta-llama-3.1-8b-bnb-4bit` model.
|
132 |
+
|
133 |
+
|
134 |
+
|