legraphista's picture
Update README.md
a5a8b76 verified
metadata
base_model: failspy/Meta-Llama-3-8B-Instruct-abliterated-v3
inference: false
library_name: gguf
license: llama3
pipeline_tag: text-generation
quantized_by: legraphista
tags:
  - quantized
  - GGUF
  - imatrix
  - quantization
  - imat
  - imatrix
  - static
  - 16bit
  - 8bit
  - 6bit
  - 5bit
  - 4bit
  - 3bit
  - 2bit
  - 1bit

Meta-Llama-3-8B-Instruct-abliterated-v3-IMat-GGUF

Llama.cpp imatrix quantization of failspy/Meta-Llama-3-8B-Instruct-abliterated-v3

Original Model: failspy/Meta-Llama-3-8B-Instruct-abliterated-v3
Original dtype: BF16 (bfloat16)
Quantized by: llama.cpp b3058
IMatrix dataset: here


Files

IMatrix

Status: βœ… Available
Link: here

Common Quants

Filename Quant type File Size Status Uses IMatrix Is Split
Meta-Llama-3-8B-Instruct-abliterated-v3.Q8_0.gguf Q8_0 8.54GB βœ… Available βšͺ Static πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.Q6_K.gguf Q6_K 6.60GB βœ… Available βšͺ Static πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.Q4_K.gguf Q4_K 4.92GB βœ… Available 🟒 IMatrix πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.Q3_K.gguf Q3_K 4.02GB βœ… Available 🟒 IMatrix πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.Q2_K.gguf Q2_K 3.18GB βœ… Available 🟒 IMatrix πŸ“¦ No

All Quants

Filename Quant type File Size Status Uses IMatrix Is Split
Meta-Llama-3-8B-Instruct-abliterated-v3.BF16.gguf BF16 16.07GB βœ… Available βšͺ Static πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.FP16.gguf F16 16.07GB βœ… Available βšͺ Static πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.Q8_0.gguf Q8_0 8.54GB βœ… Available βšͺ Static πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.Q6_K.gguf Q6_K 6.60GB βœ… Available βšͺ Static πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.Q5_K.gguf Q5_K 5.73GB βœ… Available βšͺ Static πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.Q5_K_S.gguf Q5_K_S 5.60GB βœ… Available βšͺ Static πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.Q4_K.gguf Q4_K 4.92GB βœ… Available 🟒 IMatrix πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.Q4_K_S.gguf Q4_K_S 4.69GB βœ… Available 🟒 IMatrix πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.IQ4_NL.gguf IQ4_NL 4.68GB βœ… Available 🟒 IMatrix πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.IQ4_XS.gguf IQ4_XS 4.45GB βœ… Available 🟒 IMatrix πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.Q3_K.gguf Q3_K 4.02GB βœ… Available 🟒 IMatrix πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.Q3_K_L.gguf Q3_K_L 4.32GB βœ… Available 🟒 IMatrix πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.Q3_K_S.gguf Q3_K_S 3.66GB βœ… Available 🟒 IMatrix πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.IQ3_M.gguf IQ3_M 3.78GB βœ… Available 🟒 IMatrix πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.IQ3_S.gguf IQ3_S 3.68GB βœ… Available 🟒 IMatrix πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.IQ3_XS.gguf IQ3_XS 3.52GB βœ… Available 🟒 IMatrix πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.IQ3_XXS.gguf IQ3_XXS 3.27GB βœ… Available 🟒 IMatrix πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.Q2_K.gguf Q2_K 3.18GB βœ… Available 🟒 IMatrix πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.Q2_K_S.gguf Q2_K_S 2.99GB βœ… Available 🟒 IMatrix πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.IQ2_M.gguf IQ2_M 2.95GB βœ… Available 🟒 IMatrix πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.IQ2_S.gguf IQ2_S 2.76GB βœ… Available 🟒 IMatrix πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.IQ2_XS.gguf IQ2_XS 2.61GB βœ… Available 🟒 IMatrix πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.IQ2_XXS.gguf IQ2_XXS 2.40GB βœ… Available 🟒 IMatrix πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.IQ1_M.gguf IQ1_M 2.16GB βœ… Available 🟒 IMatrix πŸ“¦ No
Meta-Llama-3-8B-Instruct-abliterated-v3.IQ1_S.gguf IQ1_S 2.02GB 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/Meta-Llama-3-8B-Instruct-abliterated-v3-IMat-GGUF --include "Meta-Llama-3-8B-Instruct-abliterated-v3.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/Meta-Llama-3-8B-Instruct-abliterated-v3-IMat-GGUF --include "Meta-Llama-3-8B-Instruct-abliterated-v3.Q8_0/*" --local-dir ./
# see FAQ for merging GGUF's

Inference

Simple chat template

<|begin_of_text|><|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|>

{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 Meta-Llama-3-8B-Instruct-abliterated-v3.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: Meta-Llama-3-8B-Instruct-abliterated-v3.Q8_0)
  3. Run gguf-split --merge Meta-Llama-3-8B-Instruct-abliterated-v3.Q8_0/Meta-Llama-3-8B-Instruct-abliterated-v3.Q8_0-00001-of-XXXXX.gguf Meta-Llama-3-8B-Instruct-abliterated-v3.Q8_0.gguf
    • Make sure to point gguf-split to the first chunk of the split.

Got a suggestion? Ping me @legraphista!