NVIDIA
Fine-tune NVIDIA Nemotron 3.5 on your data
NVIDIA's Nemotron 3.5 mixture-of-experts models with Mamba-2 layers; the text-only Lightning trains on Tensorant. Supported by the training and serving software Tensorant runs. It hasn't run on Tensorant yet, so start with a trial.
- Status
- Supported
- Input
- Text
- Kind
- Mixture of experts
- Trains with
- LoRA
- Suggested sizes
- LoRA from1 GPU, about 81 GB
Sizes
Sizes and GPUs
Each size with the GPUs one machine needs to train and to serve it. The figures are the estimates Tensorant checks a model against; a trial measures the real use. See GPU memory.
| Model | Parameters | Train with LoRA | Train with QLoRA | Serve | Status |
|---|---|---|---|---|---|
| nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16 | 30B, 3B active | 1 GPUabout 81 GB | - | 1, 2, 4 or 8 GPUsabout 79 GB in all | Supported |
| nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16 | 120B, 12B active | - | - | 2, 4 or 8 GPUsabout 303 GB in all | SupportedServing onlyNeeds its experts split across 8 GPUs; Tensorant hasn't set up the trainer's recipe for it yet. |
- Training needs the memory shown on each GPU: more GPUs on a machine train faster, but each holds the whole model.
- Serve: the numbers of GPUs of one machine the base model can be split across, and the memory it needs in all.
Training
How it trains
Train NVIDIA Nemotron 3.5 with LoRA, by supervised fine-tuning (SFT), DPO on preference pairs or reinforcement learning (RL).
- QLoRA isn't offered for this family: choose LoRA.
- RL trains with LoRA, on one GPU. RL on mixture-of-experts models isn't proven yet, so run an RL trial first.
- The Lightning model trains. Adapters train its attention layers and its Mamba-2 layers, not its experts.
Every training run, trials included, pays a training-token fee for the tokens it trains on, by the model's size.
Serving
How it serves
Publish a trained model under a name and call it through the OpenAI-compatible API, with the OpenAI SDK or any HTTP client. The name is what your application sends as model.
- A LoRA adapter you train serves on top of its base model, on the GPUs in the Serve column.
- The base model serves on its own too, in a test session or for a baseline. A deployment splits it across the GPUs of one machine: the Serve column lists the numbers each size allows.
nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16serves as a base model but doesn't train: the table says why.
Steps
Fine-tune NVIDIA Nemotron 3.5 on Tensorant
Create a project
Create a project with
nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16as its base model. The model check reads the model before anything starts.Import your data
Upload files, paste text or import a Hugging Face dataset as sources.
Review and freeze a version
Review the examples and approve the ones to train on, then freeze them into a version with training, validation and test splits.
Score the base model
Evaluate the base model on the version's test split. This baseline is what a training run is compared with, and a full run needs one.
Run a trial, then train
A trial trains a small sample to check the settings, the GPU memory and the speed. Then start the full run. LoRA on
nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16needs 1 GPU, about 81 GB.Compare and publish
Evaluate the trained model on the same test split and compare it with the baseline. Then publish it and call it through the OpenAI-compatible API.
Docs
Supported models
Every family's architecture, status and notes, and what a base model needs.
Model family
LFM2 MoE
Liquid AI's small mixture-of-experts text models, with short-convolution layers between their attention layers.
Model family
Laguna 2.1
Poolside's Laguna 2.1 mixture-of-experts text models.
Start with one dataset.
Create an account and import your data, or book a demo and we will show you the product.