Soma-1M / README.md
Oxpose's picture
Duplicate from Martico2432/srlm-1m
60b63b7
|
Raw
History Blame Contribute Delete
682 Bytes
---
datasets:
- HuggingFaceTB/smollm-corpus
language:
- en
pipeline_tag: text-generation
tags:
- SLM
- ROSA
license: mit
---
# SRLM-1M
A **S**mall **R**OSA based **L**anguage **M**odel, of 900k parameters.
## Training inforation
Trained on smollm-corpus fineweb-edu-dedup subset, over 16M tokens.
## Evaluations
- Wikitext v2 byte_perplexity: 4.4553
- Arc_easy acc_norm: 28.66%
- Arc_easy acc: 26.39%
- Blimp acc: 53.1%
- Hellaswag acc: 26.65%
- Hellaswag acc_norm: 27.07%
## How to run it
To run this model, you have to:
1. Download the files
2. Compile rosa.pyx using `python setup.py build_ext --inplace`
3. Load the model with the correct configuration
4. Call the model