Nexus One AI 🔔 Basic Tier
Admin · Fine-Tuning

Model Training

Upload datasets, launch QLoRA fine-tuning jobs, and monitor training progress live.

📂 Dataset Manager
Upload JSONL or CSV files. JSONL rows must have a text field, or prompt/completion pair, or instruction/output pair.
📤
Click to upload or drag & drop
.jsonl · .json · .csv — max 500 MB
Uploading…
FilenameRowsSizeUploaded
Loading…
🚀 Launch Fine-Tune Job
Uses QLoRA (4-bit quantization + LoRA adapters). Requires a GPU with ≥ 8 GB VRAM.
Models currently in Ollama
Training passes over dataset
Recommended: 1e-4 to 3e-4
Per-device, lower = less VRAM
Higher = more capacity
Usually 2× LoRA rank
Name for Ollama registration
📊 Training Jobs
Click a job row to expand logs and the live loss curve. Running jobs auto-refresh every 10 seconds.
NameBase ModelDatasetStatusCreated
Loading…