AI Agent Hub
Back to models
deepseek-coder-7b-instruct-v1.5 logo

deepseek-coder-7b-instruct-v1.5

Code deepseek-ai Released 2024-01-25
-- 7.0B params 4.1K context Open Source

About this model

DeepSeek Coder

[🏠Homepage] | [🤖 Chat with DeepSeek Coder] | [Discord] | [Wechat(微信)]

1. Introduction of Deepseek-Coder-7B-Instruct v1.5

Deepseek-Coder-7B-Instruct-v1.5 is continue pre-trained from Deepseek-LLM 7B on 2T tokens by employing a window size of 4K and next token prediction objective, and then fine-tuned on 2B tokens of instruction data.

2. Evaluation Results

DeepSeek Coder

3. How to Use

Here give some examples of how to use our model.

Chat Model Inference

from transformers import AutoTokenizer, AutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("deepseek-ai/deepseek-coder-7b-instruct-v1.5", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("deepseek-ai/deepseek-coder-7b-instruct-v1.5", trust_remote_code=True).cuda()
messages=[
    { 'role': 'user', 'content': "write a quick sort algorithm in python."}
]
inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt").to(model.device)

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)
print(tokenizer.decode(outputs[0][len(inputs[0]):], skip_special_tokens=True))

4. License

This code repository is licensed under the MIT License. The use of DeepSeek Coder models is subject to the Model License. DeepSeek Coder supports commercial use.

See the LICENSE-MODEL for more details.

5. Contact

If you have any questions, please raise an issue or contact us at service@deepseek.com.

Benchmark Scores

ARC
48.1
BBH
53.3
MATH
34.1
MBPP
64.6
MMLU
49.5
GSM8K
72.6
HellaSwag
72.2
HumanEval
64.1

Technical Specs

  • Parameters: 7.0B
  • Architecture: Llama
  • Context Window: 4,096 tokens
  • Input Modalities: text

Hardware Requirements

  • VRAM: 14.0 GB
  • Compute: Single GPU with ~14GB VRAM for BF16 inference; ~6GB with Q4 quantization