Edit model card

DeepSeek-Coder-V2-Instruct-IMat-GGUF

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

Original Model: deepseek-ai/DeepSeek-Coder-V2-Instruct
Original dtype: BF16 (bfloat16)
Quantized by: llama.cpp b3166
IMatrix dataset: here


Files

IMatrix

Status: βœ… Available
Link: here

Common Quants

Filename Quant type File Size Status Uses IMatrix Is Split
DeepSeek-Coder-V2-Instruct.Q8_0/* Q8_0 250.62GB βœ… Available βšͺ Static βœ‚ Yes
DeepSeek-Coder-V2-Instruct.Q6_K/* Q6_K 193.54GB βœ… Available βšͺ Static βœ‚ Yes
DeepSeek-Coder-V2-Instruct.Q4_K/* Q4_K 142.45GB βœ… Available 🟒 IMatrix βœ‚ Yes
DeepSeek-Coder-V2-Instruct.Q3_K/* Q3_K 112.67GB βœ… Available 🟒 IMatrix βœ‚ Yes
DeepSeek-Coder-V2-Instruct.Q2_K/* Q2_K 85.95GB βœ… Available 🟒 IMatrix βœ‚ Yes

All Quants

Filename Quant type File Size Status Uses IMatrix Is Split
DeepSeek-Coder-V2-Instruct.Q8_0/* Q8_0 250.62GB βœ… Available βšͺ Static βœ‚ Yes
DeepSeek-Coder-V2-Instruct.Q6_K/* Q6_K 193.54GB βœ… Available βšͺ Static βœ‚ Yes
DeepSeek-Coder-V2-Instruct.Q5_K/* Q5_K 167.22GB βœ… Available βšͺ Static βœ‚ Yes
DeepSeek-Coder-V2-Instruct.Q5_K_S/* Q5_K_S 162.31GB βœ… Available βšͺ Static βœ‚ Yes
DeepSeek-Coder-V2-Instruct.Q4_K/* Q4_K 142.45GB βœ… Available 🟒 IMatrix βœ‚ Yes
DeepSeek-Coder-V2-Instruct.Q4_K_S/* Q4_K_S 133.88GB βœ… Available 🟒 IMatrix βœ‚ Yes
DeepSeek-Coder-V2-Instruct.IQ4_NL/* IQ4_NL 132.91GB βœ… Available 🟒 IMatrix βœ‚ Yes
DeepSeek-Coder-V2-Instruct.IQ4_XS/* IQ4_XS 125.56GB βœ… Available 🟒 IMatrix βœ‚ Yes
DeepSeek-Coder-V2-Instruct.Q3_K/* Q3_K 112.67GB βœ… Available 🟒 IMatrix βœ‚ Yes
DeepSeek-Coder-V2-Instruct.Q3_K_L/* Q3_K_L 122.37GB βœ… Available 🟒 IMatrix βœ‚ Yes
DeepSeek-Coder-V2-Instruct.Q3_K_S/* Q3_K_S 101.68GB βœ… Available 🟒 IMatrix βœ‚ Yes
DeepSeek-Coder-V2-Instruct.IQ3_M/* IQ3_M 103.37GB βœ… Available 🟒 IMatrix βœ‚ Yes
DeepSeek-Coder-V2-Instruct.IQ3_S/* IQ3_S 101.68GB βœ… Available 🟒 IMatrix βœ‚ Yes
DeepSeek-Coder-V2-Instruct.IQ3_XS/* IQ3_XS 96.30GB βœ… Available 🟒 IMatrix βœ‚ Yes
DeepSeek-Coder-V2-Instruct.IQ3_XXS/* IQ3_XXS 90.85GB βœ… Available 🟒 IMatrix βœ‚ Yes
DeepSeek-Coder-V2-Instruct.Q2_K/* Q2_K 85.95GB βœ… Available 🟒 IMatrix βœ‚ Yes
DeepSeek-Coder-V2-Instruct.Q2_K_S/* Q2_K_S 79.60GB βœ… Available 🟒 IMatrix βœ‚ Yes
DeepSeek-Coder-V2-Instruct.IQ2_M/* IQ2_M 76.92GB βœ… Available 🟒 IMatrix βœ‚ Yes
DeepSeek-Coder-V2-Instruct.IQ2_S/* IQ2_S 69.87GB βœ… Available 🟒 IMatrix βœ‚ Yes
DeepSeek-Coder-V2-Instruct.IQ2_XS/* IQ2_XS 68.71GB βœ… Available 🟒 IMatrix βœ‚ Yes
DeepSeek-Coder-V2-Instruct.IQ2_XXS/* IQ2_XXS 61.50GB βœ… Available 🟒 IMatrix βœ‚ Yes
DeepSeek-Coder-V2-Instruct.IQ1_M/* IQ1_M 52.68GB βœ… Available 🟒 IMatrix βœ‚ Yes
DeepSeek-Coder-V2-Instruct.IQ1_S/* IQ1_S 47.39GB βœ… Available 🟒 IMatrix βœ‚ 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-IMat-GGUF --include "DeepSeek-Coder-V2-Instruct.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-IMat-GGUF --include "DeepSeek-Coder-V2-Instruct.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.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
  2. Locate your GGUF chunks folder (ex: DeepSeek-Coder-V2-Instruct.Q8_0)
  3. Run gguf-split --merge DeepSeek-Coder-V2-Instruct.Q8_0/DeepSeek-Coder-V2-Instruct.Q8_0-00001-of-XXXXX.gguf DeepSeek-Coder-V2-Instruct.Q8_0.gguf
    • Make sure to point gguf-split to the first chunk of the split.

Got a suggestion? Ping me @legraphista!

Downloads last month
1,045
GGUF
Model size
236B params
Architecture
deepseek2
Inference API (serverless) has been turned off for this model.

Quantized from

Collection including legraphista/DeepSeek-Coder-V2-Instruct-IMat-GGUF