metadata
license: bigscience-bloom-rail-1.0
The model is based on bigscience/bloom-1b1.
To reduce GPU memory usage, we pruned its vocabulary from 250880 to 46145 with Chinese corpus. So the total parameter is 800m now.
license: bigscience-bloom-rail-1.0
The model is based on bigscience/bloom-1b1.
To reduce GPU memory usage, we pruned its vocabulary from 250880 to 46145 with Chinese corpus. So the total parameter is 800m now.