This model converted from DeepSeek-R1 to BF16.
Here we simply provide the conversion command and related information about ollama.
If needed, we can upload the bf16 version.

FP8 to BF16

Download deepseek-ai/DeepSeek-R1 model, requires approximately 641GB of space.

cd /home/admin/models
huggingface-cli download deepseek-ai/DeepSeek-R1 --local-dir ./deepseek-ai/DeepSeek-R1

Create the environment.

conda create -yn DeepSeek-V3 python=3.12
conda activate DeepSeek-V3
pip install -r requirements.txt

Convert to BF16, requires an additional approximately 1.3 TB of space.
Here, you need to download the transformation code from the "inference" folder of deepseek-ai/DeepSeek-V3

cd deepseek-ai/DeepSeek-V3/inference
python fp8_cast_bf16.py --input-fp8-hf-path /home/admin/models/deepseek-ai/DeepSeek-R1/ --output-bf16-hf-path /home/admin/models/deepseek-ai/DeepSeek-R1-bf16

BF16 to f16.gguf

Use the llama.cpp conversion program to convert DeepSeek-R1-bf16 to gguf format, requires an additional approximately 1.3 TB of space.

python convert_hf_to_gguf.py /home/admin/models/deepseek-ai/DeepSeek-R1-bf16 --outfile /home/admin/models/deepseek-ai/DeepSeek-R1-bf16/ggml-model-f16.gguf --outtype f16

Use the llama.cpp quantitative program to quantitative model (llama-quantize needs to be compiled.), other quant option. Convert first Q2_K, requires an additional approximately 227 GB of space.

llama-quantize /home/admin/models/deepseek-ai/DeepSeek-R1-bf16/ggml-model-f16.gguf  /home/admin/models/deepseek-ai/DeepSeek-R1-bf16/ggml-model-Q2_K.gguf Q2_K

Use llama-cli to test.

llama-cli -m /home/admin/models/deepseek-ai/DeepSeek-R1-bf16/ggml-model-Q2_K.gguf -n 2048

Use with ollama

Note: this model requires Ollama 0.5.5

You can use huihui_ai/deepseek-r1:671b-q2_K directly

ollama run huihui_ai/deepseek-r1:671b-q2_K

or huihui_ai/deepseek-r1:671b-q3_K

ollama run huihui_ai/deepseek-r1:671b-q3_K

huihui-ai/DeepSeek-R1-bf16

huihui-ai/DeepSeek-R1

FP8 to BF16

BF16 to f16.gguf

Use with ollama

huihui-ai/
DeepSeek-R1-bf16