--- title: CodeFuse-CodeLlama-34B-4bits canonical_url: "https://www.modelscope.cn/models/codefuse-ai/CodeFuse-CodeLlama-34B-4bits" md_url: "https://www.modelscope.cn/models/codefuse-ai/CodeFuse-CodeLlama-34B-4bits.md" repository: codefuse-ai/CodeFuse-CodeLlama-34B-4bits last_updated: 2024-04-18 license: other pipeline_tag: text-generation tasks: - text-generation model_type: - llama architectures: - LlamaForCausalLM library_name: - pytorch frameworks: - Pytorch inference_backends: - "deploy_task text/emb" - "lmdeploy 0.9.1" - "lmdeploy_turbomind 0.9.1" - "sglang 0.5.2" - "vllm 0.9.2" downloads: 131419 stars: 39 --- # CodeFuse-CodeLlama-34B-4bits > CodeFuse-CodeLlama-34B-4bits - codefuse-ai 在 ModelScope 开源的模型。CodeFuse-CodeLlama-34B-4bits是CodeFuse-CodeLlama-34B模型的4bits量化版本，后者是通过QLoRA对基座模型CodeLlama-34b-Python进行多代码任务微调而得到的代码大模型，模型输入长度为4K。经4bits量化后，CodeFuse-CodeLlama-34B-4bits可用单张A10 (24GB显存)或者4090… codefuse-ai/CodeFuse-CodeLlama-34B-4bits 是 ModelScope 魔搭社区上的text-generation模型，采用 other 许可，可用 deploy_task text/emb、lmdeploy 0.9.1、lmdeploy_turbomind 0.9.1 部署。 - **Repository**: codefuse-ai/CodeFuse-CodeLlama-34B-4bits - **License**: other - **Tasks**: text-generation - **Inference backends**: deploy_task text/emb, lmdeploy 0.9.1, lmdeploy_turbomind 0.9.1, sglang 0.5.2, vllm 0.9.2 - **Downloads**: 131419 - **Stars**: 39 - **Last updated**: 2024-04-18 Source: https://www.modelscope.cn/models/codefuse-ai/CodeFuse-CodeLlama-34B-4bits --- # Model Card for CodeFuse-CodeLlama-34B-4bits

[[中文]](#chinese) [[English]](#english) ## Model Description CodeFuse-CodeLlama-34B-4bits is the 4-bit quantized version of CodeFuse-CodeLlama-34B, which is a 34B Code-LLM fine-tuned over multiple code tasks（600k instrunctions/answers）on the base model CodeLlama-34b-Python. After undergoing 4-bit quantization, the CodeFuse-CodeLlama-34B-4bits model can be loaded on either a single A10 (24GB VRAM) or a RTX 4090 (24GB VRAM). Moreover, the quantized model still achives an impressive accuracy of 73.8% on the Humaneval pass@1 metric.
## News and Updates 🔥🔥🔥 2023-09-28 CodeFuse-CodeLlama-34B 4-bit technical documentation has been released. If you are interested, please click the provided link to view it on the CodeFuse WeChat official account.(https://mp.weixin.qq.com/s/QLycLdgPGQjF7JE_YF466Q) 🔥🔥🔥 2023-09-26 We are pleased to announce the release of the 4-bit quantized version of CodeFuse-CodeLlama-34B. Despite the quantization process, the model still achieves a remarkable 73.8% accuracy (greedy decoding) on the HumanEval pass@1 metric. 🔥🔥🔥 2023-09-11 CodeFuse-CodeLlama34B has achived 74.4% of pass@1 (greedy decoding) on HumanEval, which is SOTA results for openspurced LLMs at present.
## Code Community **Homepage**: 🏡 https://github.com/codefuse-ai (**Please give us your support with a Star🌟 + Fork🚀 + Watch👀**) + If you wish to fine-tune the model yourself, you can visit ✨[MFTCoder](https://github.com/codefuse-ai/MFTCoder)✨✨ + If you wish to deploy the model yourself, you can visit ✨[FasterTransformer4CodeFuse](https://github.com/codefuse-ai/FasterTransformer4CodeFuse)✨✨ + If you wish to see a demo of the model, you can visit ✨[CodeFuse Demo](https://github.com/codefuse-ai/codefuse)✨✨
## Performance ### Code | Model | HumanEval(pass@1) | Date | |:--------------------------------|:-----------------:|:-------:| | **CodeFuse-CodeLlama-34B** | **74.4%** | 2023.9 | |**CodeFuse-CodeLlama-34B-4bits** | **73.8%** | 2023.9 | | WizardCoder-Python-34B-V1.0 | 73.2% | 2023.8 | | GPT-4(zero-shot) | 67.0% | 2023.3 | | PanGu-Coder2 15B | 61.6% | 2023.8 | | CodeLlama-34b-Python | 53.7% | 2023.8 | | CodeLlama-34b | 48.8% | 2023.8 | | GPT-3.5(zero-shot) | 48.1% | 2022.11 | | OctoCoder | 46.2% | 2023.8 | | StarCoder-15B | 33.6% | 2023.5 | | LLaMA 2 70B(zero-shot) | 29.9% | 2023.7 |
## GPU Memory Usage We measured the GPU memory usage after loading the model, as well as the memory usage when encoding 2048/1024 tokens and generating 1024/2048 tokens. The results are presented in the table below. | Precision | Idle Model | Encoding 2048 tokens and Generating 1024 tokens | Encoding 1024 tokens and Generating 2048 tokens | |:--------------------------------|:-------------------|:------------------------:|:------------:| |bfloat16 | 64.89GB | 69.31GB | 66.41GB | |int4 | 19.09GB | 22.19GB | 20.78GB |
## Requirements * python>=3.8 * pytorch>=2.0.0 * transformers==4.32.0 * auto_gptq==0.4.2 * Sentencepiece * CUDA 11.4
## Inference String Format The inference string is a concatenated string formed by combining conversation data (human and bot contents) in the training data format. It is used as input during the inference process. Here is an example format of the concatenated string: ```python """ <|role_start|>human<|role_end|>Human 1st round input <|role_start|>bot<|role_end|>Bot 1st round output <|role_start|>human<|role_end|>Human 2nd round input <|role_start|>bot<|role_end|>Bot 2nd round output ... ... ... <|role_start|>human<|role_end|>Human nth round input <|role_start|>bot<|role_end|>{Bot output to be genreated} """ ``` When applying inference, you always make your input string end with "<|role_start|>bot<|role_end|>" to ask the model generating answers.
## Quickstart ```bash git clone https://www.modelscope.cn/codefuse-ai/CodeFuse-CodeLlama-34B-4bits.git ``` ```bash pip install -r requirements.txt ``` ```python import os import torch import time from modelscope import AutoTokenizer, snapshot_download from auto_gptq import AutoGPTQForCausalLM os.environ["TOKENIZERS_PARALLELISM"] = "false" def load_model_tokenizer(model_path): """ Load model and tokenizer based on the given model name or local path of downloaded model. """ tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True, use_fast=False, lagecy=False) tokenizer.padding_side = "left" tokenizer.pad_token_id = tokenizer.convert_tokens_to_ids("") tokenizer.eos_token_id = tokenizer.convert_tokens_to_ids("") model = AutoGPTQForCausalLM.from_quantized(model_path, inject_fused_attention=False, inject_fused_mlp=False, use_safetensors=False, use_cuda_fp16=True, disable_exllama=False, device_map='auto' # Support multi-gpus ) return model, tokenizer def inference(model, tokenizer, prompt): """ Uset the given model and tokenizer to generate an answer for the speicifed prompt. """ st = time.time() prompt = prompt if prompt.endswith('\n') else f'{prompt}\n' inputs = f"<|role_start|>human<|role_end|>{prompt}<|role_start|>bot<|role_end|>" input_ids = tokenizer.encode(inputs, return_tensors="pt", padding=True, add_special_tokens=False).to("cuda") with torch.no_grad(): generated_ids = model.generate( input_ids=input_ids, top_p=0.95, temperature=0.1, do_sample=True, max_new_tokens=512, eos_token_id=tokenizer.eos_token_id, pad_token_id=tokenizer.pad_token_id ) print(f'generated tokens num is {len(generated_ids[0][input_ids.size(1):])}') outputs = tokenizer.batch_decode(generated_ids, skip_special_tokens=True) print(f'generate text is {outputs[0][len(inputs): ]}') latency = time.time() - st print('latency is {} seconds'.format(latency)) if __name__ == "__main__": model_dir = snapshot_download('codefuse-ai/CodeFuse-CodeLlama-34B-4bits', revision='v1.0.0') prompt = 'Please write a QuickSort program in Python' model, tokenizer = load_model_tokenizer(model_dir) inference(model, tokenizer, prompt) ``` **The current inference example code is based on [AutoGPTQ](https://github.com/PanQiWei/AutoGPTQ). If you want to achieve higher inference speed, it is recommended to combine it with [TensorRT-LLM (Early Access)](https://developer.nvidia.com/tensorrt-llm-early-access).**
## Consistency Check Here, SHA256 values are provided for the model-related files for consistency check during the download. | File | SHA256 | |-------------------------------:|:--------------------------------:| |config.json | bd1b92f942549f76d7e02e65fd346b39903943912d6d6a2ff8ff345e43e1115b | |generation_config.json | b625bd13a52d0685313c32919324b9bdc9e75a4f1338ca5c28226d1693e130a3 | |gptq_model-4bit-64g.bin | 79441bad1d5ab852d0238ed7e113b9912f31189cf9181d7119dd297c4beb454a | |pytorch_model.bin.index.json | 9a714170172282cfbcaa120af13c0df08b06d040ff24dab30229d8a010821d3d | |quantize_config.json | 3c1744a928e9d6c3f9a2cbb1bb5a89539077e7d456948bf5aee0deed6a7b8028 | |special_tokens_map.json | ff3b4a612c4e447acb02d40071bddd989fe0da87eb5b7fe0dbadfc4f74de7531 | |tokenizer.json | f7b50bcf6d6672eade5e43514d48e9c1e4e63a56aef7b14acdaca94ce93436f7 | |tokenizer.model | 9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347 | |tokenizer_config.json | c12441e82f2dce0baff87cf5948e82d6e9b51cc0b5266369c30c319fb771eeb2 |

## 模型简介 CodeFuse-CodeLlama-34B-4bits是CodeFuse-CodeLlama-34B模型的4bits量化版本，后者是通过QLoRA对基座模型CodeLlama-34b-Python进行多代码任务微调而得到的代码大模型，模型输入长度为4K。 经4bits量化后，CodeFuse-CodeLlama-34B-4bits可用单张A10 (24GB显存)或者RTX 4090 (24GB显存)加载，同时，量化后的模型在Humaneval pass@1指标上仍取得了73.8%的表现。
## 新闻 🔥🔥🔥 2023-09-28 CodeFuse-CodeLlama-34B 4bits技术文档公布，感兴趣请点击微信公众号CodeFuse链接查看: https://mp.weixin.qq.com/s/QLycLdgPGQjF7JE_YF466Q 🔥🔥🔥 2023-09-26 CodeFuse-CodeLlama-34B 4bits量化版本发布，量化后模型在HumanEval pass@1指标为73.8% (贪婪解码)。 🔥🔥🔥 2023-09-11 CodeFuse-CodeLlama-34B发布，HumanEval pass@1指标达到74.4% (贪婪解码), 为当前开源SOTA。
## 代码社区 **大本营**： 🏡 https://github.com/codefuse-ai （**请支持我们的项目Star🌟 + Fork🚀 + Watch👀**） + 如果您想自己微调该模型，可以访问 ✨[MFTCoder](https://github.com/codefuse-ai/MFTCoder)✨✨ + 如果您想自己部署该模型，可以访问 ✨[FasterTransformer4CodeFuse](https://github.com/codefuse-ai/FasterTransformer4CodeFuse)✨✨ + 如果您想观看该模型示例，可以访问 ✨[CodeFuse Demo](https://github.com/codefuse-ai/codefuse)✨✨
## 评测表现(代码) | 模型 | HumanEval(pass@1) | 日期 | |:--------------------------------|:-----------------:|:-------:| | **CodeFuse-CodeLlama-34B** | **74.4%** | 2023.9 | |**CodeFuse-CodeLlama-34B-4bits** | **73.8%** | 2023.9 | | WizardCoder-Python-34B-V1.0 | 73.2% | 2023.8 | | GPT-4(zero-shot) | 67.0% | 2023.3 | | PanGu-Coder2 15B | 61.6% | 2023.8 | | CodeLlama-34b-Python | 53.7% | 2023.8 | | CodeLlama-34b | 48.8% | 2023.8 | | GPT-3.5(zero-shot) | 48.1% | 2022.11 | | OctoCoder | 46.2% | 2023.8 | | StarCoder-15B | 33.6% | 2023.5 | | LLaMA 2 70B(zero-shot) | 29.9% | 2023.7 |
## 显存使用 我们测量了模型加载后占用的显存占用情况，以及输入2048/1024 tokens并输出1024/2048 tokens时的显存使用情况，如下表所示 | 精度 | 模型空载 | 输入2048 tokens + 输出1024 tokens | 输入1024 tokens + 输出2048 tokens | |:--------------------------------|:-------------------|:------------------------:|:------------:| |bfloat16 | 64.89GB | 69.31GB | 66.41GB | |int4 | 19.09GB | 22.19GB | 20.78GB |
## 依赖要求 * python>=3.8 * pytorch>=2.0.0 * transformers==4.32.0 * auto_gptq==0.4.2 * Sentencepiece * CUDA 11.4
## 推理数据格式 推理数据为模型在训练数据格式下拼接的字符串形式，它也是推理时输入prompt拼接的方式： ```python """ <|role_start|>human<|role_end|>Human 1st round input <|role_start|>bot<|role_end|>Bot 1st round output <|role_start|>human<|role_end|>Human 2nd round input <|role_start|>bot<|role_end|>Bot 2nd round output ... ... ... <|end|><|role_start|>human<|role_end|>Human nth round input <|end|><|role_start|>bot<|role_end|>{Bot output to be genreated} """ ``` 推理时，请确保拼接的prompt字符串以"<|role_start|>bot<|role_end|>"结尾，引导模型生成回答。
## 快速使用 ```bash git clone https://www.modelscope.cn/codefuse-ai/CodeFuse-CodeLlama-34B-4bits.git ``` ```bash pip install -r requirements.txt ``` ```python import os import torch import time from modelscope import AutoTokenizer, snapshot_download from auto_gptq import AutoGPTQForCausalLM os.environ["TOKENIZERS_PARALLELISM"] = "false" def load_model_tokenizer(model_path): """ Load model and tokenizer based on the given model name or local path of downloaded model. """ tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True, use_fast=False, lagecy=False) tokenizer.padding_side = "left" tokenizer.pad_token_id = tokenizer.convert_tokens_to_ids("") tokenizer.eos_token_id = tokenizer.convert_tokens_to_ids("") model = AutoGPTQForCausalLM.from_quantized(model_path, inject_fused_attention=False, inject_fused_mlp=False, use_safetensors=False, use_cuda_fp16=True, disable_exllama=False, device_map='auto' # 支持多卡 ) return model, tokenizer def inference(model, tokenizer, prompt): """ Uset the given model and tokenizer to generate an answer for the speicifed prompt. """ st = time.time() prompt = prompt if prompt.endswith('\n') else f'{prompt}\n' inputs = f"<|role_start|>human<|role_end|>{prompt}<|role_start|>bot<|role_end|>" input_ids = tokenizer.encode(inputs, return_tensors="pt", padding=True, add_special_tokens=False).to("cuda") with torch.no_grad(): generated_ids = model.generate( input_ids=input_ids, top_p=0.95, temperature=0.1, do_sample=True, max_new_tokens=512, eos_token_id=tokenizer.eos_token_id, pad_token_id=tokenizer.pad_token_id ) print(f'generated tokens num is {len(generated_ids[0][input_ids.size(1):])}') outputs = tokenizer.batch_decode(generated_ids, skip_special_tokens=True) print(f'generate text is {outputs[0][len(inputs): ]}') latency = time.time() - st print('latency is {} seconds'.format(latency)) if __name__ == "__main__": model_dir = snapshot_download('codefuse-ai/CodeFuse-CodeLlama-34B-4bits', revision='v1.0.0') prompt = '请用Python实现一个快速排序算法' model, tokenizer = load_model_tokenizer(model_dir) inference(model, tokenizer, prompt) ``` **目前的推理示例代码是基于[AutoGPTQ](https://github.com/PanQiWei/AutoGPTQ)的，如果你想获取更高的推理速度，建议结合使用[TensorRT-LLM (Early Access)](https://developer.nvidia.com/tensorrt-llm-early-access)。**
## 一致性校验 这里提供了模型相关文件的SHA256值，用于下载一致性校验。 | 文件 | SHA256 | |-------------------------------:|:--------------------------------:| |config.json | bd1b92f942549f76d7e02e65fd346b39903943912d6d6a2ff8ff345e43e1115b | |generation_config.json | b625bd13a52d0685313c32919324b9bdc9e75a4f1338ca5c28226d1693e130a3 | |gptq_model-4bit-64g.bin | 79441bad1d5ab852d0238ed7e113b9912f31189cf9181d7119dd297c4beb454a | |pytorch_model.bin.index.json | 9a714170172282cfbcaa120af13c0df08b06d040ff24dab30229d8a010821d3d | |quantize_config.json | 3c1744a928e9d6c3f9a2cbb1bb5a89539077e7d456948bf5aee0deed6a7b8028 | |special_tokens_map.json | ff3b4a612c4e447acb02d40071bddd989fe0da87eb5b7fe0dbadfc4f74de7531 | |tokenizer.json | f7b50bcf6d6672eade5e43514d48e9c1e4e63a56aef7b14acdaca94ce93436f7 | |tokenizer.model | 9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347 | |tokenizer_config.json | c12441e82f2dce0baff87cf5948e82d6e9b51cc0b5266369c30c319fb771eeb2 | ## 加入我们 我们是平台技术事业群AI Native团队，负责蚂蚁蚂蚁集团平台工程的智能化，团队成立3年多以来，支持了蚂蚁集团云计算基础设施智能化运维的升级改造。团队的Mission是，通过世界级的技术创新和影响，构建有广泛用户的算法服务和平台，支撑内外部产品和业务落地。团队秉承创新基因，在支撑业务落地的同时，推动技术影响。3年以来在ICLR、NeurIPS、KDD、ACL等顶会发表论文20余篇，创新业务结果获得两次蚂蚁技术最高奖T-Star，1次蚂蚁集团最高奖SuperMA。开源项目CodeFuse获得4K点赞(2024年2月)，Huggingface和modelscope上模型累积下载量超过150万次。 **我们正在寻找行业中的佼佼者加入我们的团队！如果您希望在一个充满活力、创新和卓越文化的环境中发展您的职业生涯，欢迎您查看我们的社招&校招机会，加入我们，一起创造下一个行业里程碑。** 校招：https://hrrecommend.antgroup.com/guide.html?code=8uoP5mlus5DqQYbE_EnqcE2FD5JZH21MwvMUIb9mb6X3osXPuBraG54SyM8GLn_7 社招：https://talent.antgroup.com/off-campus-position?positionId=1933830