---

base_model: google/gemma-2-9b-it extragatedbutton_content: Acknowledge license extragatedheading: Access Gemma on Hugging Face extragatedprompt: "To access Gemma on Hugging Face, you\u2019re required to review\ \ and agree to Google\u2019s usage license. To do this, please ensure you\u2019\ re logged in to Hugging Face and click below. Requests are processed immediately." inference: false library_name: gguf license: gemma pipeline_tag: text-generation quantized_by: legraphista tags:

• conversational
• quantized
• GGUF
• quantization
• imat
• imatrix
• static
• 16bit
• 8bit
• 6bit
• 5bit
• 4bit
• 3bit
• 2bit
• 1bit

---

gemma-2-9b-it-IMat-GGUF

Llama.cpp imatrix quantization of google/gemma-2-9b-it

Original Model: google/gemma-2-9b-it
Original dtype: BF16 (bfloat16)
Quantized by: llama.cpp b3266
IMatrix dataset: here

• Files

- IMatrix - Common Quants - All Quants

• Downloading using huggingface-cli
• Inference

- Simple chat template - 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
gemma-2-9b-it.Q80.gguf	Q80	9.83GB	✅ Available	⚪ Static

📦 No | gemma-2-9b-it.Q6K.gguf | Q6K | 7.59GB | ✅ Available | ⚪ Static | 📦 No | gemma-2-9b-it.Q4K.gguf | Q4K | 5.76GB | ✅ Available | 🟢 IMatrix | 📦 No | gemma-2-9b-it.Q3K.gguf | Q3K | 4.76GB | ✅ Available | 🟢 IMatrix | 📦 No | gemma-2-9b-it.Q2K.gguf | Q2K | 3.81GB | ✅ Available | 🟢 IMatrix | 📦 No

All Quants

Filename	Quant type	File Size	Status	Uses IMatrix	Is Split
gemma-2-9b-it.BF16.gguf	BF16	18.49GB	✅ Available	⚪ Static

📦 No | gemma-2-9b-it.FP16.gguf | F16 | 18.49GB | ✅ Available | ⚪ Static | 📦 No | gemma-2-9b-it.Q80.gguf | Q80 | 9.83GB | ✅ Available | ⚪ Static | 📦 No | gemma-2-9b-it.Q6K.gguf | Q6K | 7.59GB | ✅ Available | ⚪ Static | 📦 No | gemma-2-9b-it.Q5K.gguf | Q5K | 6.65GB | ✅ Available | ⚪ Static | 📦 No | gemma-2-9b-it.Q5KS.gguf | Q5K_S | 6.48GB | ✅ Available | ⚪ Static | 📦 No | gemma-2-9b-it.Q4K.gguf | Q4K | 5.76GB | ✅ Available | 🟢 IMatrix | 📦 No | gemma-2-9b-it.Q4KS.gguf | Q4K_S | 5.48GB | ✅ Available | 🟢 IMatrix | 📦 No | gemma-2-9b-it.IQ4NL.gguf | IQ4NL | 5.44GB | ✅ Available | 🟢 IMatrix | 📦 No | gemma-2-9b-it.IQ4XS.gguf | IQ4XS | 5.18GB | ✅ Available | 🟢 IMatrix | 📦 No | gemma-2-9b-it.Q3K.gguf | Q3K | 4.76GB | ✅ Available | 🟢 IMatrix | 📦 No | gemma-2-9b-it.Q3KL.gguf | Q3K_L | 5.13GB | ✅ Available | 🟢 IMatrix | 📦 No | gemma-2-9b-it.Q3KS.gguf | Q3K_S | 4.34GB | ✅ Available | 🟢 IMatrix | 📦 No | gemma-2-9b-it.IQ3M.gguf | IQ3M | 4.49GB | ✅ Available | 🟢 IMatrix | 📦 No | gemma-2-9b-it.IQ3S.gguf | IQ3S | 4.34GB | ✅ Available | 🟢 IMatrix | 📦 No | gemma-2-9b-it.IQ3XS.gguf | IQ3XS | 4.14GB | ✅ Available | 🟢 IMatrix | 📦 No | gemma-2-9b-it.IQ3XXS.gguf | IQ3XXS | 3.80GB | ✅ Available | 🟢 IMatrix | 📦 No | gemma-2-9b-it.Q2K.gguf | Q2K | 3.81GB | ✅ Available | 🟢 IMatrix | 📦 No | gemma-2-9b-it.Q2KS.gguf | Q2K_S | 3.55GB | ✅ Available | 🟢 IMatrix | 📦 No | gemma-2-9b-it.IQ2M.gguf | IQ2M | 3.43GB | ✅ Available | 🟢 IMatrix | 📦 No | gemma-2-9b-it.IQ2S.gguf | IQ2S | 3.21GB | ✅ Available | 🟢 IMatrix | 📦 No | gemma-2-9b-it.IQ2XS.gguf | IQ2XS | 3.07GB | ✅ Available | 🟢 IMatrix | 📦 No | gemma-2-9b-it.IQ2XXS.gguf | IQ2XXS | 2.82GB | ✅ Available | 🟢 IMatrix | 📦 No | gemma-2-9b-it.IQ1M.gguf | IQ1M | 2.55GB | ✅ Available | 🟢 IMatrix | 📦 No | gemma-2-9b-it.IQ1S.gguf | IQ1S | 2.38GB | ✅ Available | 🟢 IMatrix | 📦 No

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/gemma-2-9b-it-IMat-GGUF --include "gemma-2-9b-it.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/gemma-2-9b-it-IMat-GGUF --include "gemma-2-9b-it.Q8_0/*" --local-dir ./
see FAQ for merging GGUF's

---

Inference

Simple chat template

<bos><startofturn>user
{userprompt}<endof_turn>
<startofturn>model
{assistantresponse}<endof_turn>
<startofturn>user
{nextuserprompt}<endofturn>

Llama.cpp

llama.cpp/main -m gemma-2-9b-it.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

1. Locate your GGUF chunks folder (ex: gemma-2-9b-it.Q80)
2. Run gguf-split --merge gemma-2-9b-it.Q80/gemma-2-9b-it.Q80-00001-of-XXXXX.gguf gemma-2-9b-it.Q80.gguf

- Make sure to point gguf-split to the first chunk of the split.

---

Got a suggestion? Ping me @legraphista!