File size: 3,354 Bytes
0d2c6af e6328b6 fea7a04 e6328b6 | 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 | ---
license: bsd-3-clause
datasets:
- pedrodev2026/microcoder-dataset-1024-tokens
base_model:
- unsloth/Qwen2.5-Coder-1.5B-Instruct
pipeline_tag: text-generation
tags:
- coder
- code
- microcoder
---
# Microcoder 1.5B
**Microcoder 1.5B** is a code-focused language model fine-tuned from [Qwen 2.5 Coder 1.5B Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-1.5B-Instruct) using LoRA (Low-Rank Adaptation) on curated code datasets. It is designed for code generation, completion, and instruction-following tasks in a lightweight, efficient package.
---
## Model Details
| Property | Value |
|------------------|--------------------------------------------|
| **Base Model** | Qwen 2.5 Coder 1.5B Instruct |
| **Fine-tuning** | LoRA |
| **Parameters** | ~1.5B |
| **License** | BSD 3-Clause |
| **Language** | English (primary), multilingual code |
| **Task** | Code generation, completion, instruction following |
---
## Benchmarks
| Benchmark | Metric | Score |
|--------------------|----------|--------------|
| HumanEval | pass@1 | **59.15%** |
| MBPP+ | pass@1 | **52.91%** |
> HumanEval and MBPP+ results were obtained using the model in **GGUF format** with **Q5_K_M quantization**. Results may vary slightly with other formats or quantization levels.
---
## Usage
> **Important:** You must use `apply_chat_template` when formatting inputs. Passing raw text directly to the tokenizer will produce incorrect results.
```python
from transformers import AutoTokenizer, AutoModelForCausalLM
model_id = "your-org/microcoder-1.5b"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id)
messages = [
{
"role": "user",
"content": "Write a Python function that returns the nth Fibonacci number."
}
]
input_text = tokenizer.apply_chat_template(
messages,
tokenize=False,
add_generation_prompt=True
)
inputs = tokenizer(input_text, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
```
---
## Training Details
Microcoder 1.5B was fine-tuned using LoRA on top of Qwen 2.5 Coder 1.5B Instruct. The training focused on code-heavy datasets covering multiple programming languages and problem-solving scenarios, aiming to improve instruction-following and code correctness at a small model scale.
---
## Credits
- **Model credits** — see [`MODEL_CREDITS.md`](./MODEL_CREDITS.md)
- **Dataset credits** — see [`DATASET_CREDITS.md`](./DATASET_CREDITS.md)
---
## License
The Microcoder 1.5B model weights and associated code in this repository are released under the **BSD 3-Clause License**. See [`LICENSE`](./LICENSE) for details.
Note that the base model (Qwen 2.5 Coder 1.5B Instruct) and the datasets used for fine-tuning are subject to their own respective licenses, as detailed in the credit files above.
---
## Notice
The documentation files in this repository (including `README.md`, `MODEL_CREDITS.md`, `DATASET_CREDITS.md`, and other `.md` files) were generated with the assistance of an AI language model. |