Llama-3.2-3B-Instruct-IMat-GGUF

Llama.cpp imatrix quantization of meta-llama/Llama-3.2-3B-Instruct

Original Model: meta-llama/Llama-3.2-3B-Instruct
Original dtype: BF16 (bfloat16)
Quantized by: llama.cpp b3825
IMatrix dataset: here


Files

IMatrix

Status: ✅ Available
Link: here

Common Quants

Filename Quant type File Size Status Uses IMatrix Is Split
Llama-3.2-3B-Instruct.Q8_0.gguf Q8_0 3.42GB ✅ Available ⚪ Static 📦 No
Llama-3.2-3B-Instruct.Q6_K.gguf Q6_K 2.64GB ✅ Available ⚪ Static 📦 No
Llama-3.2-3B-Instruct.Q4_K.gguf Q4_K 2.02GB ✅ Available 🟢 IMatrix 📦 No
Llama-3.2-3B-Instruct.Q3_K.gguf Q3_K 1.69GB ✅ Available 🟢 IMatrix 📦 No
Llama-3.2-3B-Instruct.Q2_K.gguf Q2_K 1.36GB ✅ Available 🟢 IMatrix 📦 No

All Quants

Filename Quant type File Size Status Uses IMatrix Is Split
Llama-3.2-3B-Instruct.BF16.gguf BF16 6.43GB ✅ Available ⚪ Static 📦 No
Llama-3.2-3B-Instruct.FP16.gguf F16 6.43GB ✅ Available ⚪ Static 📦 No
Llama-3.2-3B-Instruct.Q8_0.gguf Q8_0 3.42GB ✅ Available ⚪ Static 📦 No
Llama-3.2-3B-Instruct.Q6_K.gguf Q6_K 2.64GB ✅ Available ⚪ Static 📦 No
Llama-3.2-3B-Instruct.Q5_K.gguf Q5_K 2.32GB ✅ Available ⚪ Static 📦 No
Llama-3.2-3B-Instruct.Q5_K_S.gguf Q5_K_S 2.27GB ✅ Available ⚪ Static 📦 No
Llama-3.2-3B-Instruct.Q4_K.gguf Q4_K 2.02GB ✅ Available 🟢 IMatrix 📦 No
Llama-3.2-3B-Instruct.Q4_K_S.gguf Q4_K_S 1.93GB ✅ Available 🟢 IMatrix 📦 No
Llama-3.2-3B-Instruct.IQ4_NL.gguf IQ4_NL 1.92GB ✅ Available 🟢 IMatrix 📦 No
Llama-3.2-3B-Instruct.IQ4_XS.gguf IQ4_XS 1.83GB ✅ Available 🟢 IMatrix 📦 No
Llama-3.2-3B-Instruct.Q3_K.gguf Q3_K 1.69GB ✅ Available 🟢 IMatrix 📦 No
Llama-3.2-3B-Instruct.Q3_K_L.gguf Q3_K_L 1.82GB ✅ Available 🟢 IMatrix 📦 No
Llama-3.2-3B-Instruct.Q3_K_S.gguf Q3_K_S 1.54GB ✅ Available 🟢 IMatrix 📦 No
Llama-3.2-3B-Instruct.IQ3_M.gguf IQ3_M 1.60GB ✅ Available 🟢 IMatrix 📦 No
Llama-3.2-3B-Instruct.IQ3_S.gguf IQ3_S 1.54GB ✅ Available 🟢 IMatrix 📦 No
Llama-3.2-3B-Instruct.IQ3_XS.gguf IQ3_XS 1.48GB ✅ Available 🟢 IMatrix 📦 No
Llama-3.2-3B-Instruct.IQ3_XXS.gguf IQ3_XXS 1.35GB ✅ Available 🟢 IMatrix 📦 No
Llama-3.2-3B-Instruct.Q2_K.gguf Q2_K 1.36GB ✅ Available 🟢 IMatrix 📦 No
Llama-3.2-3B-Instruct.Q2_K_S.gguf Q2_K_S 1.27GB ✅ Available 🟢 IMatrix 📦 No
Llama-3.2-3B-Instruct.IQ2_M.gguf IQ2_M 1.23GB ✅ Available 🟢 IMatrix 📦 No
Llama-3.2-3B-Instruct.IQ2_S.gguf IQ2_S 1.15GB ✅ Available 🟢 IMatrix 📦 No
Llama-3.2-3B-Instruct.IQ2_XS.gguf IQ2_XS 1.10GB ✅ Available 🟢 IMatrix 📦 No
Llama-3.2-3B-Instruct.IQ2_XXS.gguf IQ2_XXS 1.02GB ✅ Available 🟢 IMatrix 📦 No
Llama-3.2-3B-Instruct.IQ1_M.gguf IQ1_M 924.19MB ✅ Available 🟢 IMatrix 📦 No
Llama-3.2-3B-Instruct.IQ1_S.gguf IQ1_S 868.16MB ✅ 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/Llama-3.2-3B-Instruct-IMat-GGUF --include "Llama-3.2-3B-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/Llama-3.2-3B-Instruct-IMat-GGUF --include "Llama-3.2-3B-Instruct.Q8_0/*" --local-dir ./
# see FAQ for merging GGUF's

Inference

Simple chat template

<|begin_of_text|><|start_header_id|>system<|end_header_id|>

Cutting Knowledge Date: December 2023
Today Date: 26 Jul 2024

<|eot_id|><|start_header_id|>user<|end_header_id|>

{user_prompt}<|eot_id|><|start_header_id|>assistant<|end_header_id|>

{assistant_response}<|eot_id|><|start_header_id|>user<|end_header_id|>

{next_user_prompt}<|eot_id|>

Chat template with system prompt

<|begin_of_text|><|start_header_id|>system<|end_header_id|>

Cutting Knowledge Date: December 2023
Today Date: 26 Jul 2024

{system_prompt}<|eot_id|><|start_header_id|>user<|end_header_id|>

{user_prompt}<|eot_id|><|start_header_id|>assistant<|end_header_id|>

{assistant_response}<|eot_id|><|start_header_id|>user<|end_header_id|>

{next_user_prompt}<|eot_id|>

Llama.cpp

llama.cpp/main -m Llama-3.2-3B-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: Llama-3.2-3B-Instruct.Q8_0)
  3. Run gguf-split --merge Llama-3.2-3B-Instruct.Q8_0/Llama-3.2-3B-Instruct.Q8_0-00001-of-XXXXX.gguf Llama-3.2-3B-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
271
GGUF
Model size
3.21B params
Architecture
llama

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/Llama-3.2-3B-Instruct-IMat-GGUF

Quantized
(195)
this model