metadata

license: apache-2.0
language:
  - en
pipeline_tag: text-generation

Falcon-40b-oasst1-en

Falcon-40b-oasst1-en is finedtuned on Falcon-40B using OpenAssistant/oasst1 dataset. Model was only trained on 'en' language dataset which took about 03:46:23 for one Epoch.

Approach

Model was finetuned using QLoRA. Here is the Github.

CUDA

CUDA Version 12.0
Hardware: 1x NVIDIA A100-SXM
Memory: {0: '37GB'}