base_model:
- Salesforce/SFR-Iterative-DPO-LLaMA-3-8B-R
library_name: transformers
tags:
- mergekit
- merge
A zero training self-merge test of Salesforce/SFR-Iterative-DPO-LLaMA-3-8B-R using settings mentioned on mistral-11b-slimorca
It's....not dumber I guess 🤷♀️
Simple PPL comparison
perplexity.exe -[MODEL] -f wiki.test.raw -b 512 -ngl 99
SFR-Iterative-DPO-LLaMA-3-8B-R-F16.gguf - Final estimate: PPL = 7.0279 +/- 0.04493
SFR-Iterative-DPO-LLaMA-3-11.5B-R-Q6_K.gguf - Final estimate: PPL = 7.0500 +/- 0.04516
Meta-Llama-3-8B-Instruct-Q6_K - Final estimate: PPL = 8.4727 +/- 0.06308
Tools used / version
Mergekit: c93c9bb
llama.cpp: b2876
merged
This is a merge of pre-trained language models created using mergekit.
Merge Details
Merge Method
This model was merged using the passthrough merge method.
Models Merged
The following models were included in the merge:
Configuration
The following YAML configuration was used to produce this model:
dtype: bfloat16
merge_method: passthrough
slices:
- sources:
- layer_range: [0, 24]
model: Salesforce/SFR-Iterative-DPO-LLaMA-3-8B-R
- sources:
- layer_range: [8, 24]
model: Salesforce/SFR-Iterative-DPO-LLaMA-3-8B-R
parameters:
scale:
- filter: o_proj
value: 0.0
- filter: down_proj
value: 0.0
- value: 1.0
- sources:
- layer_range: [24, 32]
model: Salesforce/SFR-Iterative-DPO-LLaMA-3-8B-R