QuantFactory/OpenCoder-1.5B-Instruct-GGUF
4811
1 2---3 4license: other5license_name: inf6license_link: https://huggingface.co/infly/OpenCoder-1.5B-Instruct/blob/main/LICENSE7language:8- en9- zh10base_model:11- infly/OpenCoder-1.5B-Base12pipeline_tag: text-generation13library_name: transformers14datasets:15- OpenCoder-LLM/opencoder-sft-stage116- OpenCoder-LLM/opencoder-sft-stage217 18---19 20[](https://hf.co/QuantFactory)21 22 23# QuantFactory/OpenCoder-1.5B-Instruct-GGUF24This is quantized version of [infly/OpenCoder-1.5B-Instruct](https://huggingface.co/infly/OpenCoder-1.5B-Instruct) created using llama.cpp25 26# Original Model Card27 28 29 30 31<div align="center">32 <img src="https://github.com/OpenCoder-llm/opencoder-llm.github.io/blob/main/static/images/opencoder_icon.jpg?raw=true" width="50%" alt="OpenCoder-Icon" />33</div>34 35 36 37<p align="center">38 <a href="https://arxiv.org/pdf/2411.04905"><b>Paper Link</b>👁️</a>39</p>40 41 42## 1. Introduction43 44**OpenCoder** is an open and reproducible code LLM family which includes 1.5B and 8B base and chat models, supporting both English and Chinese languages. Starting from scratch, OpenCoder is pretrained on 2.5 trillion tokens composed of 90% raw code and 10% code-related web data, and supervised finetuned on over 4.5M high-quality SFT examples, finally reaching the performance of top-tier code LLMs. We provide not only model weights and inference code, but also the reproducible training data, the complete data processing pipeline, rigorous experimental ablation results, and detailed training protocols. Empowering researchers to build and innovate, OpenCoder is your open foundation for advancing code AI. 45 46- **Complete Open Source**: OpenCoder ensures full transparency by releasing not only the model weights and forthcoming inference code but also the complete data-cleaning code for training. This release includes high-quality synthetic data, an extensive set of checkpoints, and a dataset of over 4.5 million supervised fine-tuning (SFT) entries, making OpenCoder one of the most comprehensively open-sourced models available.47- **Comprehensive Experimental Analysis**: OpenCoder is rigorously tested through extensive ablation studies on various data-cleaning strategies and training processes, including file-level and repository-level deduplication experiments, ensuring thorough exploration and validation of the model’s performance.48- **High-Quality Synthetic Data**: OpenCoder provides a fully developed synthetic data generation process and over 4.5 million SFT data entries, establishing a robust data foundation for model training and evaluation.49- **Exceptional Performance**: OpenCoder achieves high performance across multiple language model benchmarks, positioning it among the leading open-source models for code.50 51 52## 2. Models53 54| Model | Sequence Length | Download |55|:---------------------:|:---------------:|:-----------------------------------------------------------------------:|56| OpenCoder-1.5B-Base | 4K | 🤗 [HuggingFace](https://huggingface.co/infly/OpenCoder-1.5B-Base) |57| OpenCoder-8B-Base | 8K | 🤗 [HuggingFace](https://huggingface.co/infly/OpenCoder-8B-Base) |58| OpenCoder-1.5B-Instruct | 4K | 🤗 [HuggingFace](https://huggingface.co/infly/OpenCoder-1.5B-Instruct) |59| OpenCoder-8B-Instruct | 8K | 🤗 [HuggingFace](https://huggingface.co/infly/OpenCoder-8B-Instruct) |60 61## 3. Datasets62 63| Dataset | Num | Download |64|:---------------------:|:---------------:|:-----------------------------------------------------------------------:|65| OpenCoder-SFT-Stage1 | 4.21 M | 🤗 [HuggingFace](https://huggingface.co/datasets/OpenCoder-LLM/opencoder-sft-stage1) |66| OpenCoder-SFT-Stage2 | 375 K | 🤗 [HuggingFace](https://huggingface.co/datasets/OpenCoder-LLM/opencoder-sft-stage2) |67 68 69 70## 4. Benchmarks71 72**Note:** For the detailed evaluation results, please refer to [our paper](https://arxiv.org/pdf/2411.04905).73 74<!-- ### Base Model -->75<!-- | model | OpenCoder-1.5B-Base | OpenCoder-8B-Base |76|:---------------:|:-------------:|:------------:|77| HumanEval(+) | 54.3 (49.4) | 66.5 (63.4) |78| MBPP(+) | 70.6 (58.7) | 79.9 (70.4) |79| BigCodeBench | 24.5 | 40.5 |80| BigCodeBench-Hard | 5.4 | 9.5 | -->81 82 83<!-- ### Chat Model -->84| model | OpenCoder-1.5B-Instruct | OpenCoder-8B-Instruct |85|:---------------:|:-------------:|:------------:|86| HumanEval(+) | 72.5 (67.7) | 83.5 (78.7) |87| MBPP(+) | 72.7 (61.9) | 79.1 (69.0) |88| BigCodeBench | 33.3 | 40.3 |89| BigCodeBench-Hard | 11.5 | 16.9 |90| LiveCodeBench | 12.8 | 23.2 |91| MultiPL-E (AVG) | 57.5 | 71.0 |92 93 94## 5. Inference95 96### Inference with Huggingface's Transformers97 98```python99import torch100from transformers import AutoTokenizer, AutoModelForCausalLM101 102model_name = "infly/OpenCoder-1.5B-Instruct"103model = AutoModelForCausalLM.from_pretrained(model_name,104 torch_dtype=torch.bfloat16,105 device_map="auto",106 trust_remote_code=True)107tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)108 109messages=[110 { 'role': 'user', 'content': "write a quick sort algorithm in python."}111]112 113inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt")114 115outputs = model.generate(inputs, max_new_tokens=512, do_sample=False)116 117result = tokenizer.decode(outputs[0][len(inputs[0]):], skip_special_tokens=True)118print(result)119```120 121<!-- ### Inference with vLLM (recommended) -->122 123## 6. License124 125OpenCoder series (including Base and Chat) support commercial applications under a permissive [License](https://huggingface.co/infly/OpenCoder-1.5B-Instruct/blob/main/LICENSE).126 127## 7. Citation128```129@inproceedings{Huang2024OpenCoderTO,130 title={OpenCoder: The Open Cookbook for Top-Tier Code Large Language Models},131 author={Siming Huang and Tianhao Cheng and Jason Klein Liu and Jiaran Hao and Liuyihan Song and Yang Xu and J. Yang and J. H. Liu and Chenchen Zhang and Linzheng Chai and Ruifeng Yuan and Zhaoxiang Zhang and Jie Fu and Qian Liu and Ge Zhang and Zili Wang and Yuan Qi and Yinghui Xu and Wei Chu},132 year={2024},133 url={https://arxiv.org/pdf/2411.04905}134}135```136