Instructions to use AaronTekle/SQLQwen with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use AaronTekle/SQLQwen with PEFT:
from peft import PeftModel from transformers import AutoModelForCausalLM base_model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-Coder-0.5B-Instruct") model = PeftModel.from_pretrained(base_model, "AaronTekle/SQLQwen") - Notebooks
- Google Colab
- Kaggle
Update README.md
Browse files
README.md
CHANGED
|
@@ -73,12 +73,6 @@ Only **2,162,688 of 496,195,456 parameters**, or approximately **0.4359%**, were
|
|
| 73 |
|
| 74 |
* **Base model:** [`Qwen/Qwen2.5-Coder-0.5B-Instruct`](https://huggingface.co/Qwen/Qwen2.5-Coder-0.5B-Instruct)
|
| 75 |
* **Fine-tuning:** PEFT LoRA
|
| 76 |
-
* **Training:** TRL `SFTTrainer`
|
| 77 |
-
* **LoRA rank:** `16`
|
| 78 |
-
* **LoRA alpha:** `32`
|
| 79 |
-
* **LoRA dropout:** `0.05`
|
| 80 |
-
* **Target modules:** `q_proj`, `k_proj`, `v_proj`, `o_proj`
|
| 81 |
-
* **Training objective:** Completion-only supervised fine-tuning
|
| 82 |
|
| 83 |
Qwen2.5-Coder was selected because SQL generation is fundamentally a structured code-generation task rather than conventional natural-language classification.
|
| 84 |
|
|
|
|
| 73 |
|
| 74 |
* **Base model:** [`Qwen/Qwen2.5-Coder-0.5B-Instruct`](https://huggingface.co/Qwen/Qwen2.5-Coder-0.5B-Instruct)
|
| 75 |
* **Fine-tuning:** PEFT LoRA
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 76 |
|
| 77 |
Qwen2.5-Coder was selected because SQL generation is fundamentally a structured code-generation task rather than conventional natural-language classification.
|
| 78 |
|