tjellm's picture
Update README.md
e9457e8 verified
|
raw
history blame
523 Bytes
metadata
license: apache-2.0
tags:
  - medusa

Model Description

This is a Medusa model for Mistral 7B Instruct v0.2. This is trained using the latest Medusa 2 commit.

Training:

  • Dataset used is the self distillation dataset from Mistral 7B Instruct v0.2, temperature 0.3 with output token of 2048.
  • It has been trained using axolotl fork as describe in Medusa 2 README.md

Inference:

  • To load the model please follow the instruction found in Github