# DeepSeek-Coder-V2-Instruct --- ## README([From Huggingface](https://huggingface.co/deepseek-ai/DeepSeek-Coder-V2-Instruct))

API Platform | How to Use | License |

Paper Link👁️

# DeepSeek-Coder-V2: Breaking the Barrier of Closed-Source Models in Code Intelligence ## 1. Introduction We present DeepSeek-Coder-V2, an open-source Mixture-of-Experts (MoE) code language model that achieves performance comparable to GPT4-Turbo in code-specific tasks. Specifically, DeepSeek-Coder-V2 is further pre-trained from an intermediate checkpoint of DeepSeek-V2 with additional 6 trillion tokens. Through this continued pre-training, DeepSeek-Coder-V2 substantially enhances the coding and mathematical reasoning capabilities of DeepSeek-V2, while maintaining comparable performance in general language tasks. Compared to DeepSeek-Coder-33B, DeepSeek-Coder-V2 demonstrates significant advancements in various aspects of code-related tasks, as well as reasoning and general capabilities. Additionally, DeepSeek-Coder-V2 expands its support for programming languages from 86 to 338, while extending the context length from 16K to 128K.

In standard benchmark evaluations, DeepSeek-Coder-V2 achieves superior performance compared to closed-source models such as GPT4-Turbo, Claude 3 Opus, and Gemini 1.5 Pro in coding and math benchmarks. The list of supported programming languages can be found [here](https://github.com/deepseek-ai/DeepSeek-Coder-V2/blob/main/supported_langs.txt). ## 2. Model Downloads We release the DeepSeek-Coder-V2 with 16B and 236B parameters based on the [DeepSeekMoE](https://arxiv.org/pdf/2401.06066) framework, which has actived parameters of only 2.4B and 21B , including base and instruct models, to the public.

| **Model** | **#Total Params** | **#Active Params** | **Context Length** | **Download** | | :-----------------------------: | :---------------: | :----------------: | :----------------: | :----------------------------------------------------------: | | DeepSeek-Coder-V2-Lite-Base | 16B | 2.4B | 128k | [🤗 HuggingFace](https://huggingface.co/deepseek-ai/DeepSeek-Coder-V2-Lite-Base) | | DeepSeek-Coder-V2-Lite-Instruct | 16B | 2.4B | 128k | [🤗 HuggingFace](https://huggingface.co/deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct) | | DeepSeek-Coder-V2-Base | 236B | 21B | 128k | [🤗 HuggingFace](https://huggingface.co/deepseek-ai/DeepSeek-Coder-V2-Base) | | DeepSeek-Coder-V2-Instruct | 236B | 21B | 128k | [🤗 HuggingFace](https://huggingface.co/deepseek-ai/DeepSeek-Coder-V2-Instruct) |

## 3. Chat Website You can chat with the DeepSeek-Coder-V2 on DeepSeek's official website: [coder.deepseek.com](https://coder.deepseek.com/sign_in) ## 4. API Platform We also provide OpenAI-Compatible API at DeepSeek Platform: [platform.deepseek.com](https://platform.deepseek.com/), and you can also pay-as-you-go at an unbeatable price.

## 5. How to run locally **Here, we provide some examples of how to use DeepSeek-Coder-V2-Lite model. If you want to utilize DeepSeek-Coder-V2 in BF16 format for inference, 80GB*8 GPUs are required.** ### Inference with Huggingface's Transformers You can directly employ [Huggingface's Transformers](https://github.com/huggingface/transformers) for model inference. #### Code Completion ```python from paddlenlp.transformers import AutoTokenizer, AutoModelForCausalLM import paddle tokenizer = AutoTokenizer.from_pretrained("deepseek-ai/DeepSeek-Coder-V2-Lite-Base", trust_remote_code=True) model = AutoModelForCausalLM.from_pretrained("deepseek-ai/DeepSeek-Coder-V2-Lite-Base", trust_remote_code=True, dtype=paddle.bfloat16).cuda() input_text = "#write a quick sort algorithm" inputs = tokenizer(input_text, return_tensors="pd") outputs = model.generate(**inputs, max_length=128)[0] print(tokenizer.decode(outputs[0], skip_special_tokens=True)) ``` #### Code Insertion ```python from paddlenlp.transformers import AutoTokenizer, AutoModelForCausalLM import paddle tokenizer = AutoTokenizer.from_pretrained("deepseek-ai/DeepSeek-Coder-V2-Lite-Base", trust_remote_code=True) model = AutoModelForCausalLM.from_pretrained("deepseek-ai/DeepSeek-Coder-V2-Lite-Base", trust_remote_code=True, dtype=paddle.bfloat16).cuda() input_text = """<｜fim▁begin｜>def quick_sort(arr): if len(arr) <= 1: return arr pivot = arr[0] left = [] right = [] <｜fim▁hole｜> if arr[i] < pivot: left.append(arr[i]) else: right.append(arr[i]) return quick_sort(left) + [pivot] + quick_sort(right)<｜fim▁end｜>""" inputs = tokenizer(input_text, return_tensors="pd") outputs = model.generate(**inputs, max_length=128)[0] print(tokenizer.decode(outputs[0], skip_special_tokens=True)[len(input_text):]) ``` #### Chat Completion ```python from paddlenlp.transformers import AutoTokenizer, AutoModelForCausalLM import paddle tokenizer = AutoTokenizer.from_pretrained("deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct", trust_remote_code=True) model = AutoModelForCausalLM.from_pretrained("deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct", trust_remote_code=True, dtype=paddle.bfloat16).cuda() messages=[ { 'role': 'user', 'content': "write a quick sort algorithm in python."} ] inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pd") # tokenizer.eos_token_id is the id of <｜end▁of▁sentence｜> token outputs = model.generate(inputs, max_new_tokens=512, do_sample=False, top_k=50, top_p=0.95, num_return_sequences=1, eos_token_id=tokenizer.eos_token_id)[0] print(tokenizer.decode(outputs[0][len(inputs[0]):], skip_special_tokens=True)) ``` The complete chat template can be found within `tokenizer_config.json` located in the huggingface model repository. An example of chat template is as belows: ```bash <｜begin▁of▁sentence｜>User: {user_message_1} Assistant: {assistant_message_1}<｜end▁of▁sentence｜>User: {user_message_2} Assistant: ``` You can also add an optional system message: ```bash <｜begin▁of▁sentence｜>{system_message} User: {user_message_1} Assistant: {assistant_message_1}<｜end▁of▁sentence｜>User: {user_message_2} Assistant: ``` ### Inference with vLLM (recommended) To utilize [vLLM](https://github.com/vllm-project/vllm) for model inference, please merge this Pull Request into your vLLM codebase: https://github.com/vllm-project/vllm/pull/4650. ```python from paddlenlp.transformers import AutoTokenizer from vllm import LLM, SamplingParams max_model_len, tp_size = 8192, 1 model_name = "deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct" tokenizer = AutoTokenizer.from_pretrained(model_name) llm = LLM(model=model_name, tensor_parallel_size=tp_size, max_model_len=max_model_len, trust_remote_code=True, enforce_eager=True) sampling_params = SamplingParams(temperature=0.3, max_tokens=256, stop_token_ids=[tokenizer.eos_token_id]) messages_list = [ [{"role": "user", "content": "Who are you?"}], [{"role": "user", "content": "write a quick sort algorithm in python."}], [{"role": "user", "content": "Write a piece of quicksort code in C++."}], ] prompt_token_ids = [tokenizer.apply_chat_template(messages, add_generation_prompt=True) for messages in messages_list] outputs = llm.generate(prompt_token_ids=prompt_token_ids, sampling_params=sampling_params) generated_text = [output.outputs[0].text for output in outputs] print(generated_text) ``` ## 6. License This code repository is licensed under [the MIT License](https://github.com/deepseek-ai/DeepSeek-Coder-V2/blob/main/LICENSE-CODE). The use of DeepSeek-Coder-V2 Base/Instruct models is subject to [the Model License](https://github.com/deepseek-ai/DeepSeek-Coder-V2/blob/main/LICENSE-MODEL). DeepSeek-Coder-V2 series (including Base and Instruct) supports commercial use. ## 7. Contact If you have any questions, please raise an issue or contact us at [service@deepseek.com](service@deepseek.com). ## Model Files - [README.md](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/README.md) (11.1 KB) - [config.json](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/config.json) (1.5 KB) - [generation_config.json](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/generation_config.json) (145.0 B) - [model-00001-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00001-of-000055.safetensors) (8.0 GB) - [model-00002-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00002-of-000055.safetensors) (8.0 GB) - [model-00003-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00003-of-000055.safetensors) (8.0 GB) - [model-00004-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00004-of-000055.safetensors) (8.0 GB) - [model-00005-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00005-of-000055.safetensors) (8.0 GB) - [model-00006-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00006-of-000055.safetensors) (8.0 GB) - [model-00007-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00007-of-000055.safetensors) (8.0 GB) - [model-00008-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00008-of-000055.safetensors) (8.0 GB) - [model-00009-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00009-of-000055.safetensors) (8.0 GB) - [model-00010-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00010-of-000055.safetensors) (8.0 GB) - [model-00011-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00011-of-000055.safetensors) (8.0 GB) - [model-00012-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00012-of-000055.safetensors) (8.0 GB) - [model-00013-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00013-of-000055.safetensors) (8.0 GB) - [model-00014-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00014-of-000055.safetensors) (8.0 GB) - [model-00015-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00015-of-000055.safetensors) (8.0 GB) - [model-00016-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00016-of-000055.safetensors) (8.0 GB) - [model-00017-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00017-of-000055.safetensors) (8.0 GB) - [model-00018-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00018-of-000055.safetensors) (8.0 GB) - [model-00019-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00019-of-000055.safetensors) (8.0 GB) - [model-00020-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00020-of-000055.safetensors) (8.0 GB) - [model-00021-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00021-of-000055.safetensors) (8.0 GB) - [model-00022-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00022-of-000055.safetensors) (8.0 GB) - [model-00023-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00023-of-000055.safetensors) (8.0 GB) - [model-00024-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00024-of-000055.safetensors) (8.0 GB) - [model-00025-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00025-of-000055.safetensors) (8.0 GB) - [model-00026-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00026-of-000055.safetensors) (8.0 GB) - [model-00027-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00027-of-000055.safetensors) (8.0 GB) - [model-00028-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00028-of-000055.safetensors) (8.0 GB) - [model-00029-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00029-of-000055.safetensors) (8.0 GB) - [model-00030-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00030-of-000055.safetensors) (8.0 GB) - [model-00031-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00031-of-000055.safetensors) (8.0 GB) - [model-00032-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00032-of-000055.safetensors) (8.0 GB) - [model-00033-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00033-of-000055.safetensors) (8.0 GB) - [model-00034-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00034-of-000055.safetensors) (8.0 GB) - [model-00035-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00035-of-000055.safetensors) (8.0 GB) - [model-00036-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00036-of-000055.safetensors) (8.0 GB) - [model-00037-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00037-of-000055.safetensors) (8.0 GB) - [model-00038-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00038-of-000055.safetensors) (8.0 GB) - [model-00039-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00039-of-000055.safetensors) (8.0 GB) - [model-00040-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00040-of-000055.safetensors) (8.0 GB) - [model-00041-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00041-of-000055.safetensors) (8.0 GB) - [model-00042-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00042-of-000055.safetensors) (8.0 GB) - [model-00043-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00043-of-000055.safetensors) (8.0 GB) - [model-00044-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00044-of-000055.safetensors) (8.0 GB) - [model-00045-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00045-of-000055.safetensors) (8.0 GB) - [model-00046-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00046-of-000055.safetensors) (8.0 GB) - [model-00047-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00047-of-000055.safetensors) (8.0 GB) - [model-00048-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00048-of-000055.safetensors) (8.0 GB) - [model-00049-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00049-of-000055.safetensors) (8.0 GB) - [model-00050-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00050-of-000055.safetensors) (8.0 GB) - [model-00051-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00051-of-000055.safetensors) (8.0 GB) - [model-00052-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00052-of-000055.safetensors) (8.0 GB) - [model-00053-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00053-of-000055.safetensors) (8.0 GB) - [model-00054-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00054-of-000055.safetensors) (8.0 GB) - [model-00055-of-000055.safetensors](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model-00055-of-000055.safetensors) (6.4 GB) - [model.safetensors.index.json](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/model.safetensors.index.json) (2.7 MB) - [tokenizer.json](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/tokenizer.json) (4.4 MB) - [tokenizer_config.json](https://paddlenlp.bj.bcebos.com/models/community/deepseek-ai/DeepSeek-Coder-V2-Instruct/tokenizer_config.json) (1.2 KB) [Back to Main](../../)