Based on Meta-Llama-3-8b-Instruct, and is governed by Meta Llama 3 License agreement: https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct
Experimental model trying to make Llama 3 8B better in general overall. Pretty difficult to do.
Base model: https://huggingface.co/failspy/Meta-Llama-3-8B-Instruct-abliterated-v3
Training:
- 4096 sequence length
- Training duration is around 2 days on 2x3090Ti
- 1 epoch training with a massive dataset for minimized repetition sickness.
- LORA with 64-rank 128-alpha resulting in ~2% trainable weights.
Llama 3 Instruct format:
<|begin_of_text|><|start_header_id|>system<|end_header_id|>
{{ system_prompt }}<|eot_id|><|start_header_id|>user<|end_header_id|>
{{ user_message_1 }}<|eot_id|><|start_header_id|>assistant<|end_header_id|>
{{ model_answer_1 }}<|eot_id|><|start_header_id|>user<|end_header_id|>
{{ user_message_2 }}<|eot_id|><|start_header_id|>assistant<|end_header_id|>
Quants:
FP16: https://huggingface.co/OwenArli/ArliAI-Llama-3-8B-Argon-v1.0
GGUF: https://huggingface.co/OwenArli/ArliAI-Llama-3-8B-Argon-v1.0-GGUF