• Method Continued pretraining
• Epochs 3
• Batch size 4
• Grad Accum 2
• Learning rate 0.00005
• Embedding Learning Rate 0.000005
• Optimizer AdamW 8-bit
• Context length 2048
• Warmup steps 4
• Packing True
• weight decay 0.001
• LR scheduler Cosine
• seed 3407
• LoRA Rank 32
• Alpha 64
• Dropout 0.0
• Variant RS-lora
• Dataset 2-Pass CPT Corpus

Uploaded finetuned model

• Developed by: ChesterProgrammer
• License: apache-2.0
• Finetuned from model : DreamFast/Qwen3.5-9B-Uncensored-HauhauCS-Aggressive-Safetensor-Benchmark