Hermes Mamba 2.8B SlimPajama cDPO
After OpenHermes 2.5 supervised fine-tuning, this model received conservative direct preference optimization. The card reports beta 0.1 and label smoothing 0.1, using a mixture of public and private preference data from Nous Research.
Research release
Select the cDPO repository explicitly
The example loader in the model card names the supervised SlimPajama repository. That example does not select this cDPO checkpoint. Use the full repository identifier for this release when the preference-trained weights are required.Hermes-mamba-2.8b-slimpj-cDPO model cardHermes-mamba-2.8b-slimpj-cDPO model configuration
Use the released chat markers
These adaptations replace the usual ChatML start and end markers with ASCII control characters 0x16 and 0x17. The card describes this choice for RWKV tokenizer compatibility. Literal ChatML tokens are not interchangeable with the published format.Hermes-mamba-2.8b-slimpj-cDPO model card
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Research release |
| Access | Downloadable |
| Model identifiers | EleutherAI/Hermes-mamba-2.8b-slimpj-cDPO |
| Base architecture | Mamba; SlimPajama base plus preference trainingHermes-mamba-2.8b-slimpj-cDPO model cardMamba state-space architecture and original models |
| Nominal parameters | 2.8BHermes-mamba-2.8b-slimpj-cDPO model card |
| License | Apache 2.0Hermes-mamba-2.8b-slimpj-cDPO model card |
Related models
- Hermes Mamba 2.8B
EleutherAI · Hermes Mamba
A chat adaptation of the Pile-trained Mamba 2.8B base.
Launch date not confirmedResearch releaseDownloadable - Hermes Mamba 2.8B SlimPajama
EleutherAI · Hermes Mamba
A Mamba chat model whose base was trained on SlimPajama.
Launch date not confirmedResearch releaseDownloadable