OPEN-CORE • DEV PREVIEW

Fine-tune LLMs in 10 minutes — no CLI required.

FineForge is the UI for parameter-efficient training. LoRA • QLoRA • SLoTH — upload data, set hyperparams, visualize training, and export adapters.

LLaMAMistralFalconLoRAQLoRASLoTH

Why FineForge?

Fine-tune in 10 minutes

Upload data, pick a base model, choose LoRA/QLoRA/SLoTH, hit Run.

Visual training dashboard

Loss curves, tokens/sec, GPU memory — all in one place.

Infra-agnostic

Run locally or burst to GPU clouds like RunPod/Lambda/HF Spaces.

Export & deploy fast

Adapters (HF) & GGUF for local inference — one click deploy.

From dataset to deploy — in one place

  • • Drag-and-drop dataset (CSV/JSONL/HF Datasets)
  • • Choose base model (LLaMA/Mistral/Falcon)
  • • Configure LoRA/QLoRA/SLoTH with guardrails
  • • Live metrics: loss, tokens/sec, GPU mem
  • • Export adapters (HF) or GGUF for local inference
training run • demo
loss: 1.82 → 1.21
tokens/sec: 2,450
GPU mem: 19.4 / 24 GB