jpacifico / chocolatine-3b

Quantized 4-bit version of the original Chocolatine LLM, best performing 3B model on the OpenLLM Leaderboard.

3B

174 Pulls Updated 6 weeks ago

6 weeks ago

7cd747b38a14 · 2.4GB

model
phi3
·
3.82B
·
Q4_K_M
template
{{ if .System }}<|system|> {{ .System }}<|end|> {{ end }}{{ if .Prompt }}<|user|> {{ .Prompt }}<|end|> {{ end }}<|assistant|> {{ .Response }}<|end|>
system
You are a friendly assistant called Chocolatine.
params
{"stop":["{\"stop\": [\"<|end|>\",\"<|user|>\",\"<|assistant|>\"]}"]}

Readme

Chocolatine-3b

Quantized Q4_K_M version of the Chocolatine-3B-Instruct-DPO-Revised original model.

Chocolatine-3b is a DPO fine-tuning of microsoft/Phi-3-mini-4k-instruct (3.82B params)
using the jpacifico/french-orca-dpo-pairs-revised rlhf dataset.
Training in French also improves the model in English, surpassing the performances of its base model.
Window context = 4k tokens

Benchmarks

Chocolatine is the best-performing 3B model on the OpenLLM Leaderboard (july 2024)
5th best < 30B params (average benchmarks).

Limitations

The Chocolatine model is a quick demonstration that a base model can be easily fine-tuned to achieve compelling performance.
It does not have any moderation mechanism.

  • Developed by: Jonathan Pacifico, 2024
  • Model type: LLM
  • Language(s) (NLP): French, English
  • License: MIT