LLaMA-Factory Online

Editor Review

The docs walk through registering, opening the console, choosing Start fine-tuning, picking a card count, and then opening the web trainer. Low-rank adaptation is configured on the training tab, where you can preview the dataset and write weights to the output directory. On the chat tab you load the adapter, then unload it and ask the same question of the original model. Data can be a preset or an upload, and the docs name two instruction formats. Evaluation and export sit on the same set of tabs. After the shutdown notice, those steps cannot be run again on the official site. It fits researchers who need to read old experiment notes and reconstruct the training steps.

AI Tools Navigator Editorial TeamUpdated: 2026-10-09

What is LLaMA-Factory Online

LlamaFactory Online is a cloud training platform built with the open-source fine-tuning project. Its docs address research labs, company R&D teams, and individual developers. A visual interface covers data preparation, fine-tuning, evaluation, and chat tests, including full-parameter and low-rank methods. Training can be watched in experiment monitors, and an inference API can load the fine-tuned result. After you start an instance from the console, you can open the web trainer, a code editor, or a notebook. Datasets can be preset or uploaded. A shutdown notice on the same domain says the platform has stopped operating, the site no longer opens, and you cannot sign in to keep training.

LLaMA-Factory Online Key Features

Configure fine-tuning in the browser

After the instance starts, the web trainer lets you pick a model and dataset, fill training settings, and preview the command that will run.

Preset and uploaded datasets

Fine-tuning can use a dataset the platform already provides, or a file you upload. The docs name two instruction-data formats.

Training monitors and evaluation

Training can be connected to common experiment monitors, and the evaluation view checks a chosen model and adapter on a dataset.

Chat with the fine-tuned model

On the chat tab, point the adapter path at the output directory and load the model to talk with it. Unloading the adapter lets you compare answers from before fine-tuning.

Prepare data inside the instance

The same instance can open a code editor or notebook to download a model and clean data before you return to the trainer.

LLaMA-Factory Online Key Advantages

The web trainer can select a model, preview data, and show the loss curve, and the same settings can be copied to a command line

Both preset datasets and uploads work, and the docs name two common instruction formats

The chat tab can load an adapter and then unload it so one question compares answers before and after fine-tuning

One instance includes the web trainer, a code editor, and a notebook, so data prep and training stay together

LLaMA-Factory Online Use Cases

Researchers trying a fine-tune

A lab starts an instance from the docs, runs a low-rank train and a chat comparison on preset data, then switches to its own files.

A training demo in class

An instructor uses the web UI to show model selection, the loss curve, and a chat with the adapter, so students skip local setup.

Reading an old experiment log

Someone who trained here before can use the docs on output folders, evaluation, and export to interpret weight notes they still have.

Frequently Asked Questions

▶Can you still register for LlamaFactory Online and fine-tune?

No. A shutdown notice on the original domain says the platform has stopped and users can no longer sign in. Opening the site now does not show a working page, so Start fine-tuning, card selection, and adapter loading cannot be run. Old weight notes can still be read against the docs, but training cannot continue on the platform.

▶How did the documented training flow work?

The docs say to register, open the console, click Start fine-tuning, choose a card count, and enter the web trainer. A low-rank run can preview the dataset and the command. Afterward, the chat tab points the adapter at the output directory and loads it. Data can be preset or uploaded, and evaluation, export, and chat share that interface. The docs also described a new-user trial, but that path closed when the service stopped.

LLaMA-Factory Online

Editor Review

What is LLaMA-Factory Online

Basic Info

Best For