BoyBarley v32 (Recommended)

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained('BoyBarley/BoyBarley-v32')
tokenizer = AutoTokenizer.from_pretrained('BoyBarley/BoyBarley-v32')

Training

• Params: ~494M, bfloat16
• Epochs: 1, LR: 5e-6
• Train loss: 0.0751, Eval loss: 0.0665
• Dataset: BoyBarley/BoyBarley-v32-dataset

Features

• DevOps CLI (Docker, Kubernetes, Linux)
• Tool JSON output for agents
• Refuses destructive commands
• Bilingual Indonesian/English