| pipeline_tag: text-generation | |
| tags: | |
| - llama | |
| - safetensors | |
| - fineweb-edu | |
| - dclm-edu | |
| # cosmo.0 | |
| `DedeProGames/cosmo.0` is a Llama-architecture language model with 34.6 million parameters. | |
| ## Architecture | |
| | Property | Value | | |
| | --- | --- | | |
| | Architecture | Llama | | |
| | Parameters | 34.6M | | |
| | Weight format | Safetensors | | |
| ## Training | |
| cosmo.0 was trained on **5 billion tokens** from **FineWeb-Edu** and **DCLM-Edu**. | |
| ## Limitations | |
| At 34.6M parameters, cosmo.0 is a very small language model, with limited capacity for complex tasks compared with larger models. | |
| ## Repository | |
| [DedeProGames/cosmo.0](https://huggingface.co/DedeProGames/cosmo.0) |