|
--- |
|
license: mit |
|
base_model: |
|
- Qwen/Qwen3-1.7B |
|
tags: |
|
- code |
|
- qwen3 |
|
--- |
|
# π» Qwen-1.7B Coder β XformAI Fine-Tuned |
|
|
|
**Model:** `XformAI-india/qwen-1.7b-coder` |
|
**Base Model:** [`Qwen/Qwen3-1.7B`](https://huggingface.co/Qwen/Qwen3-1.7B) |
|
**Architecture:** Transformer decoder (GPT-style) |
|
**Size:** 1.7 Billion Parameters |
|
**Fine-Tuned By:** [XformAI](https://xformai.in) |
|
**Release Date:** May 2025 |
|
**License:** MIT |
|
|
|
--- |
|
|
|
## π Overview |
|
|
|
`qwen-1.7b-coder` is a **purpose-built code generation model**, fine-tuned from Qwen3 1.7B by XformAI to deliver highly usable Python, JS, and Bash snippets with low-latency inference. |
|
|
|
Designed to help: |
|
- π§βπ» Developers |
|
- π§ AI agents |
|
- βοΈ Backend toolchains |
|
Generate and complete code reliably β both in IDEs and on edge devices. |
|
|
|
--- |
|
|
|
## π§ Training Highlights |
|
|
|
| Aspect | Value | |
|
|---------------------|--------------------| |
|
| Fine-Tuning Type | Instruction-tuned on code corpus | |
|
| Target Domains | Python, Bash, HTML, JavaScript | |
|
| Style | Docstring-to-code, prompt-to-app | |
|
| Tuning Technique | LoRA (8-bit) + PEFT | |
|
| Framework | π€ Transformers | |
|
| Precision | bfloat16 | |
|
| Epochs | 3 | |
|
| Max Tokens | 2048 | |
|
|
|
--- |
|
|
|
## π§ Use Cases |
|
|
|
- VSCode-like autocomplete agents |
|
- Shell command assistants |
|
- Backend logic & API template generation |
|
- Code-aware chatbots |
|
- On-device copilots |
|
|
|
--- |
|
|
|
## βοΈ Example Prompt + Usage |
|
|
|
```python |
|
from transformers import AutoTokenizer, AutoModelForCausalLM |
|
|
|
model = AutoModelForCausalLM.from_pretrained("XformAI-india/qwen-1.7b-coder") |
|
tokenizer = AutoTokenizer.from_pretrained("XformAI-india/qwen-1.7b-coder") |
|
|
|
prompt = "Write a Python script that takes a directory path and prints all .txt file names inside it." |
|
inputs = tokenizer(prompt, return_tensors="pt") |
|
outputs = model.generate(**inputs, max_new_tokens=200) |
|
print(tokenizer.decode(outputs[0], skip_special_tokens=True)) |