Skip to content

README.md

Latest commit

 

History

History
20 lines (10 loc) · 485 Bytes

README.md

File metadata and controls

20 lines (10 loc) · 485 Bytes
Raw

Training and Evaluation

Training

We train our models with 4 H200 GPUs. The models can also be trained with 4xA100s/H100s. The 7B model requires 80G gpu cards, and 3B model requires 48G gpu cards.

We release our trained models on Huggingface.

sh scripts/train.sh

We will add more settings under scripts/.

Evaluation

We released the code, and we will release scripts of different benchmarks soon.