---
tags:
- GGUF
- iMat
- Llama3
- conversational
---
```
e88 88e d8
d888 888b 8888 8888 ,"Y88b 888 8e d88
C8888 8888D 8888 8888 "8" 888 888 88b d88888
Y888 888P Y888 888P ,ee 888 888 888 888
"88 88" "88 88" "88 888 888 888 888
b
8b,
e88'Y88 d8 888
d888 'Y ,"Y88b 888,8, d88 ,e e, 888
C8888 "8" 888 888 " d88888 d88 88b 888
Y888 ,d ,ee 888 888 888 888 , 888
"88,d88 "88 888 888 888 "YeeP" 888
PROUDLY PRESENTS
```
## experiment_2_8b-iMat-GGUF
Quantization Notes: Quantized from 3500 checkpoint. Use repetition penalty (--repeat-penalty on llama.cpp) of ~1.15 with Q6_K and lower and ~1.18 with IQ3_M and lower for best results.
Quantized from fp16 with love.
* Weighted quantizations were created using fp16 GGUF and [groups_merged-enhancedV2-TurboMini.txt](https://github.com/ggerganov/llama.cpp/discussions/5263#discussioncomment-9432658) in 189 chunks and n_ctx=512
* This method of calculating the importance matrix showed improvements in some areas for Mistral 7b and Llama3 8b models, see above post for details
* The enhancedv2-turbomini file appends snippets from turboderp's calibration data to the standard groups_merged.txt file
For a brief rundown of iMatrix quant performance please see this [PR](https://github.com/ggerganov/llama.cpp/pull/5747)
All quants are verified working prior to uploading to repo for your safety and convenience.
Original model card [here](https://huggingface.co/rAIfle/experiment_2_8b-fp16) and below
---
# experiment_2_8b-fp16
Another experimental train w/ unsloth. This time, roughly 0.6 epochs of the cleaned c2-logs. My metaparams are probably bad, since the loss-value was super weird at the end. Also uploaded another version in the `checkpoint-3500`-branch that may mitigate some of that.