Model Card for Cygnis-Alpha-2
Cygnis-Alpha-2 is a fine-tuned generative text model based on the Llama-3.1-8B architecture. It has been specifically trained by Simonc-44 to implement a systematic reasoning process before generating final outputs.
Cygnis-Alpha-2 utilizes a custom Chain-of-Thought (CoT) dataset to improve logical consistency and reduce hallucinations compared to the base Llama 3.1 8B model.
Model Architecture
Cygnis-Alpha-2 is based on the Llama-3.1 transformer architecture, optimized with the following configurations during the refining process:
- LoRA (Low-Rank Adaptation) with a rank of 32
- 4-bit NormalFloat (NF4) quantization
- Integrated reasoning tokens for Chain-of-Thought processing
Instruction Format
To leverage the reasoning capabilities of Cygnis-Alpha-2, the following prompt structure should be used:
### Système: Tu es Cygnis Alpha 2, une IA souveraine créée par Simonc-44.
### Utilisateur: [Votre message]
### Assistant:
Troubleshooting
If the reasoning tags do not appear: Ensure your system prompt explicitly identifies the model as Cygnis Alpha 2 and that you are using the correct instruction template.
Performance issues: Ensure you are using the Unsloth library or a Transformers version compatible with Llama 3.1 (4.43.0 or newer).
Notice
Cygnis-Alpha-2 is a fine-tuned model and does not have built-in moderation mechanisms. Users should be aware that the model may generate outputs that reflect the biases present in the base model or the training data.
The Cygnis Project Team
Developed and refined by Simonc-44.