DeepSeek
Fine-tune DeepSeek V2 Lite on your data
DeepSeek's V2 Lite mixture-of-experts text models, for chat and for code. 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, QLoRA
- Suggested sizes
- LoRA from1 GPU, about 42 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 |
|---|---|---|---|---|---|
| deepseek-ai/DeepSeek-V2-Lite-ChatMixture of experts; a 48 GB GPU to train | 16B | 1 GPUabout 42 GB | 1 GPUabout 40 GB | 1, 2, 4 or 8 GPUsabout 40 GB in all | Supported |
| deepseek-ai/DeepSeek-Coder-V2-Lite-InstructMixture of experts, for code | 16B | 1 GPUabout 42 GB | 1 GPUabout 40 GB | 1, 2, 4 or 8 GPUsabout 40 GB in all | Supported |
- 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 DeepSeek V2 Lite with LoRA or QLoRA, by supervised fine-tuning (SFT), DPO on preference pairs or reinforcement learning (RL).
- RL trains with LoRA, on one GPU. RL on mixture-of-experts models isn't proven yet, so run an RL trial first.
- Their repositories declare code of their own, but Transformers has DeepSeek V2 built in, so they pass the model check.
- Adapters train the attention projections except
kv_b_proj, and the dense and shared-expert layers, not the routed experts. QLoRA leaves the experts in 16-bit, so it saves little memory.
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.
Steps
Fine-tune DeepSeek V2 Lite on Tensorant
Create a project
Create a project with
deepseek-ai/DeepSeek-V2-Lite-Chatas its base model, or pick another size from the list. 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
deepseek-ai/DeepSeek-V2-Lite-Chatneeds 1 GPU, about 42 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
DeepSeek R1 distills
Qwen and Llama models that DeepSeek trained on DeepSeek R1's answers, so they reason before they answer.
Model family
NVIDIA Nemotron 3.5
NVIDIA's Nemotron 3.5 mixture-of-experts models with Mamba-2 layers; the text-only Lightning trains on Tensorant.
Start with one dataset.
Create an account and import your data, or book a demo and we will show you the product.