# obtain the official LLaMA model weights and place them in ./models
ls ./models
llama-2-7b tokenizer_checklist.chk tokenizer.model
# [Optional] for models using BPE tokenizers
ls ./models
<folder containing weights and tokenizer json> vocab.json
# [Optional] for PyTorch .bin models like Mistral-7B
ls ./models
<folder containing weights and tokenizer json>

# install Python dependencies
python3 -m pip install -r requirements.txt

# convert the model to ggml FP16 format
python3 convert_hf_to_gguf.py models/mymodel/

# quantize the model to 4-bits (using Q4_K_M method)
./llama-quantize ./models/mymodel/ggml-model-f16.gguf ./models/mymodel/ggml-model-Q4_K_M.gguf Q4_K_M

# update the gguf filetype to current version if older version is now unsupported
./llama-quantize ./models/mymodel/ggml-model-Q4_K_M.gguf ./models/mymodel/ggml-model-Q4_K_M-v2.gguf COPY

Model Details

Original model: https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B

Summary models 📋

Filename	Quant type	Description
DeepSeek-R1-Distill-Qwen-1.5B-F16.gguf	F16	Half precision, no quantization applied
DeepSeek-R1-Distill-Qwen-1.5B-Q8_0.gguf	Q8_0	8-bit quantization, highest quality, largest size
DeepSeek-R1-Distill-Qwen-1.5B-Q6_K.gguf	Q6_K	6-bit quantization, very high quality
DeepSeek-R1-Distill-Qwen-1.5B-Q5_1.gguf	Q5_1	5-bit quantization, good balance of quality and size
DeepSeek-R1-Distill-Qwen-1.5B-Q5_K_M.gguf	Q5_K_M	5-bit quantization, good balance of quality and size
DeepSeek-R1-Distill-Qwen-1.5B-Q5_K_S.gguf	Q5_K_S	5-bit quantization, good balance of quality and size
DeepSeek-R1-Distill-Qwen-1.5B-Q5_0.gguf	Q5_0	5-bit quantization, good balance of quality and size
DeepSeek-R1-Distill-Qwen-1.5B-Q4_1.gguf	Q4_1	4-bit quantization, balanced quality and size
DeepSeek-R1-Distill-Qwen-1.5B-Q4_K_M.gguf	Q4_K_M	4-bit quantization, balanced quality and size
DeepSeek-R1-Distill-Qwen-1.5B-Q4_K_S.gguf	Q4_K_S	4-bit quantization, balanced quality and size
DeepSeek-R1-Distill-Qwen-1.5B-Q4_0.gguf	Q4_0	4-bit quantization, balanced quality and size
DeepSeek-R1-Distill-Qwen-1.5B-Q3_K_L.gguf	Q3_K_L	3-bit quantization, smaller size, lower quality
DeepSeek-R1-Distill-Qwen-1.5B-Q3_K_M.gguf	Q3_K_M	3-bit quantization, smaller size, lower quality
DeepSeek-R1-Distill-Qwen-1.5B-Q3_K_S.gguf	Q3_K_S	3-bit quantization, smaller size, lower quality
DeepSeek-R1-Distill-Qwen-1.5B-Q2_K.gguf	Q2_K	2-bit quantization, smallest size, lowest quality

Usage with Ollama 🦙

Direct from Ollama

ollama run hdnh2006/DeepSeek-R1-Distill-Qwen-1.5B

Download Models Using huggingface-cli 🤗

Installation of `huggingface_hub[cli]`

pip install -U "huggingface_hub[cli]"

Downloading Specific Model Files

huggingface-cli download hdnh2006/DeepSeek-R1-Distill-Qwen-1.5B --include "DeepSeek-R1-Distill-Qwen-1.5B-Q4_K_M.gguf" --local-dir ./

Which File Should I Choose? 📈

A comprehensive analysis with performance charts is provided by Artefact2 here.

Assessing System Capabilities

Determine Your Model Size: Start by checking the amount of RAM and VRAM available in your system. This will help you decide the largest possible model you can run.
Optimizing for Speed:
- GPU Utilization: To run your model as quickly as possible, aim to fit the entire model into your GPU's VRAM. Pick a version that’s 1-2GB smaller than the total VRAM.
Maximizing Quality:
- Combined Memory: For the highest possible quality, sum your system RAM and GPU's VRAM. Then choose a model that's 1-2GB smaller than this combined total.

Deciding Between 'I-Quant' and 'K-Quant'

Simplicity:
- K-Quant: If you prefer a straightforward approach, select a K-quant model. These are labeled as 'QX_K_X', such as Q5_K_M.
Advanced Configuration:
- Feature Chart: For a more nuanced choice, refer to the llama.cpp feature matrix.
- I-Quant Models: Best suited for configurations below Q4 and for systems running cuBLAS (Nvidia) or rocBLAS (AMD). These are labeled 'IQX_X', such as IQ3_M, and offer better performance for their size.
- Compatibility Considerations:
  - I-Quant Models: While usable on CPU and Apple Metal, they perform slower compared to their K-quant counterparts. The choice between speed and performance becomes a significant tradeoff.
  - AMD Cards: Verify if you are using the rocBLAS build or the Vulkan build. I-quants are not compatible with Vulkan.
  - Current Support: At the time of writing, LM Studio offers a preview with ROCm support, and other inference engines provide specific ROCm builds.

By following these guidelines, you can make an informed decision on which file best suits your system and performance needs.

Contact 🌐

Website: henrynavarro.org

Email: public.contact.rerun407@simplelogin.com

hdnh2006/DeepSeek-R1-Distill-Qwen-1.5B-GGUF

DeepSeek-R1-Distill-Qwen-1.5B GGUF llama.cpp quantization by Henry Navarro 🧠🤖