RoGemma-7b-Instruct-IMat-GGUF

Llama.cpp imatrix quantization of OpenLLM-Ro/RoGemma-7b-Instruct

Original Model: OpenLLM-Ro/RoGemma-7b-Instruct
Original dtype: BF16 (bfloat16)
Quantized by: llama.cpp b3246
IMatrix dataset: here


Files

IMatrix

Status: βœ… Available
Link: here

Common Quants

Filename Quant type File Size Status Uses IMatrix Is Split
RoGemma-7b-Instruct.Q8_0.gguf Q8_0 9.08GB βœ… Available βšͺ Static πŸ“¦ No
RoGemma-7b-Instruct.Q6_K.gguf Q6_K 7.01GB βœ… Available βšͺ Static πŸ“¦ No
RoGemma-7b-Instruct.Q4_K.gguf Q4_K 5.33GB βœ… Available 🟒 IMatrix πŸ“¦ No
RoGemma-7b-Instruct.Q3_K.gguf Q3_K 4.37GB βœ… Available 🟒 IMatrix πŸ“¦ No
RoGemma-7b-Instruct.Q2_K.gguf Q2_K 3.48GB βœ… Available 🟒 IMatrix πŸ“¦ No

All Quants

Filename Quant type File Size Status Uses IMatrix Is Split
RoGemma-7b-Instruct.BF16.gguf BF16 17.08GB βœ… Available βšͺ Static πŸ“¦ No
RoGemma-7b-Instruct.FP16.gguf F16 17.08GB βœ… Available βšͺ Static πŸ“¦ No
RoGemma-7b-Instruct.Q8_0.gguf Q8_0 9.08GB βœ… Available βšͺ Static πŸ“¦ No
RoGemma-7b-Instruct.Q6_K.gguf Q6_K 7.01GB βœ… Available βšͺ Static πŸ“¦ No
RoGemma-7b-Instruct.Q5_K.gguf Q5_K 6.14GB βœ… Available βšͺ Static πŸ“¦ No
RoGemma-7b-Instruct.Q5_K_S.gguf Q5_K_S 5.98GB βœ… Available βšͺ Static πŸ“¦ No
RoGemma-7b-Instruct.Q4_K.gguf Q4_K 5.33GB βœ… Available 🟒 IMatrix πŸ“¦ No
RoGemma-7b-Instruct.Q4_K_S.gguf Q4_K_S 5.05GB βœ… Available 🟒 IMatrix πŸ“¦ No
RoGemma-7b-Instruct.IQ4_NL.gguf IQ4_NL 5.01GB βœ… Available 🟒 IMatrix πŸ“¦ No
RoGemma-7b-Instruct.IQ4_XS.gguf IQ4_XS 4.77GB βœ… Available 🟒 IMatrix πŸ“¦ No
RoGemma-7b-Instruct.Q3_K.gguf Q3_K 4.37GB βœ… Available 🟒 IMatrix πŸ“¦ No
RoGemma-7b-Instruct.Q3_K_L.gguf Q3_K_L 4.71GB βœ… Available 🟒 IMatrix πŸ“¦ No
RoGemma-7b-Instruct.Q3_K_S.gguf Q3_K_S 3.98GB βœ… Available 🟒 IMatrix πŸ“¦ No
RoGemma-7b-Instruct.IQ3_M.gguf IQ3_M 4.11GB βœ… Available 🟒 IMatrix πŸ“¦ No
RoGemma-7b-Instruct.IQ3_S.gguf IQ3_S 3.98GB βœ… Available 🟒 IMatrix πŸ“¦ No
RoGemma-7b-Instruct.IQ3_XS.gguf IQ3_XS 3.80GB βœ… Available 🟒 IMatrix πŸ“¦ No
RoGemma-7b-Instruct.IQ3_XXS.gguf IQ3_XXS 3.49GB βœ… Available 🟒 IMatrix πŸ“¦ No
RoGemma-7b-Instruct.Q2_K.gguf Q2_K 3.48GB βœ… Available 🟒 IMatrix πŸ“¦ No
RoGemma-7b-Instruct.Q2_K_S.gguf Q2_K_S 3.25GB βœ… Available 🟒 IMatrix πŸ“¦ No
RoGemma-7b-Instruct.IQ2_M.gguf IQ2_M 3.13GB βœ… Available 🟒 IMatrix πŸ“¦ No
RoGemma-7b-Instruct.IQ2_S.gguf IQ2_S 2.92GB βœ… Available 🟒 IMatrix πŸ“¦ No
RoGemma-7b-Instruct.IQ2_XS.gguf IQ2_XS 2.81GB βœ… Available 🟒 IMatrix πŸ“¦ No
RoGemma-7b-Instruct.IQ2_XXS.gguf IQ2_XXS 2.59GB βœ… Available 🟒 IMatrix πŸ“¦ No
RoGemma-7b-Instruct.IQ1_M.gguf IQ1_M 2.32GB βœ… Available 🟒 IMatrix πŸ“¦ No
RoGemma-7b-Instruct.IQ1_S.gguf IQ1_S 2.16GB βœ… Available 🟒 IMatrix πŸ“¦ 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/RoGemma-7b-Instruct-IMat-GGUF --include "RoGemma-7b-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/RoGemma-7b-Instruct-IMat-GGUF --include "RoGemma-7b-Instruct.Q8_0/*" --local-dir ./
# see FAQ for merging GGUF's

Inference

Simple chat template

<bos><start_of_turn>user
{user_prompt}<end_of_turn>
<start_of_turn>model
{assistant_response}<end_of_turn>
<start_of_turn>user
{next_user_prompt}<end_of_turn>
<start_of_turn>model

Chat template with system prompt

<bos>{system_prompt}<start_of_turn>user
{user_prompt}<end_of_turn>
<start_of_turn>model
{assistant_response}<end_of_turn>
<start_of_turn>user
{next_user_prompt}<end_of_turn>
<start_of_turn>model

Llama.cpp

llama.cpp/main -m RoGemma-7b-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: RoGemma-7b-Instruct.Q8_0)
  3. Run gguf-split --merge RoGemma-7b-Instruct.Q8_0/RoGemma-7b-Instruct.Q8_0-00001-of-XXXXX.gguf RoGemma-7b-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
106
GGUF
Model size
8.54B params
Architecture
gemma

1-bit

2-bit

3-bit

4-bit

5-bit

6-bit

8-bit

16-bit

Inference Examples
Inference API (serverless) has been turned off for this model.

Model tree for legraphista/RoGemma-7b-Instruct-IMat-GGUF

Base model

google/gemma-7b
Quantized
(2)
this model