the1ullneversee
commited on
Commit
•
1104b15
1
Parent(s):
74bd794
Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,68 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language:
|
3 |
+
- en
|
4 |
+
tags:
|
5 |
+
- llama
|
6 |
+
- instruct
|
7 |
+
- conversational
|
8 |
+
- api
|
9 |
+
- code-generation
|
10 |
+
- lora
|
11 |
+
license: apache-2.0
|
12 |
+
---
|
13 |
+
|
14 |
+
# LLaMA-7B-Instruct-API-Coder
|
15 |
+
|
16 |
+
## Model Description
|
17 |
+
|
18 |
+
This model is a fine-tuned version of the LLaMA-7B-Instruct model, specifically trained on conversational data related to RESTful API usage and code generation. The training data was generated by LLaMA-70B-Instruct, focusing on API interactions and code creation based on user queries and JSON REST schemas.
|
19 |
+
|
20 |
+
## Intended Use
|
21 |
+
|
22 |
+
This model is designed to assist developers and API users in:
|
23 |
+
|
24 |
+
1. Understanding and interacting with RESTful APIs
|
25 |
+
2. Generating code snippets to call APIs based on user questions
|
26 |
+
3. Interpreting JSON REST schemas
|
27 |
+
4. Providing conversational guidance on API usage
|
28 |
+
|
29 |
+
## Training Data
|
30 |
+
|
31 |
+
The model was fine-tuned on a dataset of conversational interactions generated by LLaMA-70B-Instruct. This dataset includes:
|
32 |
+
|
33 |
+
- Discussions about RESTful API concepts
|
34 |
+
- Examples of API usage
|
35 |
+
- Code generation based on API schemas
|
36 |
+
- Q&A sessions about API integration
|
37 |
+
|
38 |
+
## Training Procedure
|
39 |
+
|
40 |
+
1. Base Model: LLaMA-7B-Instruct
|
41 |
+
2. Quantization: The base model was loaded in 4-bit precision using Unsloth for efficient training
|
42 |
+
3. Fine-tuning Method: SFTTrainer (Supervised Fine-Tuning Trainer) was used for the fine-tuning process
|
43 |
+
4. LoRA (Low-Rank Adaptation): The model was fine-tuned using LoRA to generate an adapter
|
44 |
+
5. Merging: The LoRA adapter was merged back with the original model to create the final fine-tuned version
|
45 |
+
|
46 |
+
This approach allows for efficient fine-tuning while maintaining model quality and reducing computational requirements.
|
47 |
+
|
48 |
+
## Limitations
|
49 |
+
|
50 |
+
- The model's knowledge is limited to the APIs and schemas present in the training data
|
51 |
+
- It may not be up-to-date with the latest API standards or practices
|
52 |
+
- The generated code should be reviewed and tested before use in production environments
|
53 |
+
- Performance may vary compared to the full-precision model due to 4-bit quantization
|
54 |
+
|
55 |
+
## Ethical Considerations
|
56 |
+
|
57 |
+
- The model should not be used to access or manipulate APIs without proper authorization
|
58 |
+
- Users should be aware of potential biases in the generated code or API usage suggestions
|
59 |
+
|
60 |
+
## Additional Information
|
61 |
+
|
62 |
+
- Model Type: Causal Language Model
|
63 |
+
- Language: English
|
64 |
+
- License: Apache 2.0
|
65 |
+
- Fine-tuning Technique: LoRA (Low-Rank Adaptation)
|
66 |
+
- Quantization: 4-bit precision
|
67 |
+
|
68 |
+
For any questions or issues, please open an issue in the GitHub repository.
|