QwOwO-1.5b but bigger & better in every way :3
It uses ChatML/Qwen formatting
This model is trained on a dataset generated by Mistral Small 3, It's goal is to give qwen less robotic and less boring answers.
The dataset is 10k examples from alpaca-cleaned regenerated using owo-speak/uwu-speak
Experiments were conducted using a private infrastructure, which has a carbon efficiency of 0.432 kgCO2eq/kWh. A cumulative of 1 hours of computation was performed on hardware of type RTX 3090 (TDP of 350W).
Total emissions are estimated to be 0.15 kgCO2eq of which 0 percents were directly offset. This is equivalent to 0.61 Km driven by an average ICE car [1].
Estimations were conducted using the MachineLearning Impact calculator presented in Lacoste et al. (2019).
References
