Prox 7B DPO
By OpenVoid AI
Discord: https://discord.gg/CBDAbKkgNV
Model description
This model is based on Mistral-7b-v0.2 a new base model released by MistralAI on March 23, 2024 but they have not yet published on HuggingFace. It has been fine-tuned on a dataset containing information related to hacking and coding, with the aim of enhancing its performance on tasks within these domains.
Intended uses & limitations
This model is designed to assist with a variety of natural language processing tasks related to hacking and coding, such as:
- Code generation
- Code explanation and documentation
- Answering questions about hacking techniques and cybersecurity
- Providing insights and suggestions for coding projects
However, it is important to note that while the model has been fine-tuned on hacking and coding data, it should not be relied upon for critical security applications or used to engage in illegal activities. The model's outputs should be carefully reviewed and verified by experts in the relevant fields before being put into practice.
Training data
The prox-7b model was fine-tuned on a proprietary dataset curated by OpenVoid AI, which consists of a diverse range of hacking and coding-related content.
Training hyperparameters
The following hyperparameters were used during training:
- Learning rate: 2e-05
- Train batch size: 4
- Eval batch size: 8
- Seed: 42
- Distributed type: multi-GPU
- Number of devices: 2
- Gradient accumulation steps: 4
- Total train batch size: 32
- Total eval batch size: 16
- Optimizer: Adam with betas=(0.9,0.95) and epsilon=1e-08
- LR scheduler type: cosine
- LR scheduler warmup steps: 100
- Training Steps: 414
The training was performed using a distributed multi-GPU setup to accelerate the process and handle the large model size.