DeepSeek-Coder-V2-Instruct-0724-IMat-GGUF

Llama.cpp imatrix quantization of deepseek-ai/DeepSeek-Coder-V2-Instruct-0724

Original Model: deepseek-ai/DeepSeek-Coder-V2-Instruct-0724
Original dtype: BF16 (bfloat16)
Quantized by: llama.cpp b3687
IMatrix dataset: here

• Files
  • IMatrix
  • Common Quants
  • All Quants
• Downloading using huggingface-cli
• Inference
  • Simple chat template
  • Chat template with system prompt
  • Llama.cpp
• FAQ
  • Why is the IMatrix not applied everywhere?
  • How do I merge a split GGUF?

---

Files

IMatrix

Status: ✅ Available
Link: here

Common Quants

Filename	Quant type	File Size	Status	Uses IMatrix	Is Split
DeepSeek-Coder-V2-Instruct-0724.Q8_0/*	Q8_0	250.62GB	✅ Available	⚪ Static	✂ Yes
DeepSeek-Coder-V2-Instruct-0724.Q6_K/*	Q6_K	193.54GB	✅ Available	⚪ Static	✂ Yes
DeepSeek-Coder-V2-Instruct-0724.Q4_K/*	Q4_K	142.45GB	✅ Available	🟢 IMatrix	✂ Yes
DeepSeek-Coder-V2-Instruct-0724.Q3_K/*	Q3_K	112.67GB	✅ Available	🟢 IMatrix	✂ Yes
DeepSeek-Coder-V2-Instruct-0724.Q2_K/*	Q2_K	85.95GB	✅ Available	🟢 IMatrix	✂ Yes

All Quants

Filename	Quant type	File Size	Status	Uses IMatrix	Is Split
DeepSeek-Coder-V2-Instruct-0724.Q8_0/*	Q8_0	250.62GB	✅ Available	⚪ Static	✂ Yes
DeepSeek-Coder-V2-Instruct-0724.Q6_K/*	Q6_K	193.54GB	✅ Available	⚪ Static	✂ Yes
DeepSeek-Coder-V2-Instruct-0724.Q5_K/*	Q5_K	167.22GB	✅ Available	⚪ Static	✂ Yes
DeepSeek-Coder-V2-Instruct-0724.Q5_K_S/*	Q5_K_S	162.31GB	✅ Available	⚪ Static	✂ Yes
DeepSeek-Coder-V2-Instruct-0724.Q4_K/*	Q4_K	142.45GB	✅ Available	🟢 IMatrix	✂ Yes
DeepSeek-Coder-V2-Instruct-0724.Q4_K_S/*	Q4_K_S	133.88GB	✅ Available	🟢 IMatrix	✂ Yes
DeepSeek-Coder-V2-Instruct-0724.IQ4_NL/*	IQ4_NL	132.91GB	✅ Available	🟢 IMatrix	✂ Yes
DeepSeek-Coder-V2-Instruct-0724.IQ4_XS/*	IQ4_XS	125.56GB	✅ Available	🟢 IMatrix	✂ Yes
DeepSeek-Coder-V2-Instruct-0724.Q3_K/*	Q3_K	112.67GB	✅ Available	🟢 IMatrix	✂ Yes
DeepSeek-Coder-V2-Instruct-0724.Q3_K_L/*	Q3_K_L	122.37GB	✅ Available	🟢 IMatrix	✂ Yes
DeepSeek-Coder-V2-Instruct-0724.Q3_K_S/*	Q3_K_S	101.68GB	✅ Available	🟢 IMatrix	✂ Yes
DeepSeek-Coder-V2-Instruct-0724.IQ3_M/*	IQ3_M	103.37GB	✅ Available	🟢 IMatrix	✂ Yes
DeepSeek-Coder-V2-Instruct-0724.IQ3_S/*	IQ3_S	101.68GB	✅ Available	🟢 IMatrix	✂ Yes
DeepSeek-Coder-V2-Instruct-0724.IQ3_XS/*	IQ3_XS	96.30GB	✅ Available	🟢 IMatrix	✂ Yes
DeepSeek-Coder-V2-Instruct-0724.IQ3_XXS/*	IQ3_XXS	90.85GB	✅ Available	🟢 IMatrix	✂ Yes
DeepSeek-Coder-V2-Instruct-0724.Q2_K/*	Q2_K	85.95GB	✅ Available	🟢 IMatrix	✂ Yes
DeepSeek-Coder-V2-Instruct-0724.Q2_K_S/*	Q2_K_S	79.60GB	✅ Available	🟢 IMatrix	✂ Yes
DeepSeek-Coder-V2-Instruct-0724.IQ2_M/*	IQ2_M	76.92GB	✅ Available	🟢 IMatrix	✂ Yes
DeepSeek-Coder-V2-Instruct-0724.IQ2_S/*	IQ2_S	69.87GB	✅ Available	🟢 IMatrix	✂ Yes
DeepSeek-Coder-V2-Instruct-0724.IQ2_XS/*	IQ2_XS	68.71GB	✅ Available	🟢 IMatrix	✂ Yes
DeepSeek-Coder-V2-Instruct-0724.IQ2_XXS/*	IQ2_XXS	61.50GB	✅ Available	🟢 IMatrix	✂ Yes
DeepSeek-Coder-V2-Instruct-0724.IQ1_M/*	IQ1_M	52.68GB	✅ Available	🟢 IMatrix	✂ Yes
DeepSeek-Coder-V2-Instruct-0724.IQ1_S/*	IQ1_S	47.39GB	✅ Available	🟢 IMatrix	✂ Yes

Downloading using huggingface-cli

If you do not have hugginface-cli installed:

pip install -U "huggingface_hub[cli]"

Download the specific file you want:

huggingface-cli download legraphista/DeepSeek-Coder-V2-Instruct-0724-IMat-GGUF --include "DeepSeek-Coder-V2-Instruct-0724.Q8_0.gguf" --local-dir ./

If the model file is big, it has been split into multiple files. In order to download them all to a local folder, run:

huggingface-cli download legraphista/DeepSeek-Coder-V2-Instruct-0724-IMat-GGUF --include "DeepSeek-Coder-V2-Instruct-0724.Q8_0/*" --local-dir ./
# see FAQ for merging GGUF's

---

Inference

Simple chat template

<｜begin▁of▁sentence｜><｜User｜>{user_prompt}<｜Assistant｜>{assistant_response}<｜end▁of▁sentence｜><｜User｜>{next_user_prompt}

Chat template with system prompt

        <｜begin▁of▁sentence｜>{system_prompt}<｜User｜>{user_prompt}<｜Assistant｜>{assistant_response}<｜end▁of▁sentence｜><｜User｜>{next_user_prompt}

Llama.cpp

llama.cpp/main -m DeepSeek-Coder-V2-Instruct-0724.Q8_0.gguf --color -i -p "prompt here (according to the chat template)"

---

FAQ

Why is the IMatrix not applied everywhere?

According to this investigation, it appears that lower quantizations are the only ones that benefit from the imatrix input (as per hellaswag results).

How do I merge a split GGUF?

1. Make sure you have gguf-split available
   • To get hold of gguf-split, navigate to https://github.com/ggerganov/llama.cpp/releases
   • Download the appropriate zip for your system from the latest release
   • Unzip the archive and you should be able to find gguf-split
2. Locate your GGUF chunks folder (ex: DeepSeek-Coder-V2-Instruct-0724.Q8_0)
3. Run gguf-split --merge DeepSeek-Coder-V2-Instruct-0724.Q8_0/DeepSeek-Coder-V2-Instruct-0724.Q8_0-00001-of-XXXXX.gguf DeepSeek-Coder-V2-Instruct-0724.Q8_0.gguf
   • Make sure to point gguf-split to the first chunk of the split.

---

Got a suggestion? Ping me @legraphista!