Fine-tuning

Multi-GPU training in repeatable environments, as easy as a function call away.

"Training with multiple GPUs"

Refine models using up to 8 GPUs employing your preferred sharding method. Utilize A100 80 GB nodes with access to 640 GB of VRAM.

Code-defined environments

Specify environments in code to ensure the repeatability of your fine-tuning runs across your entire team, eliminating the need for cumbersome Jupyter notebooks!

Resources as needed

"Initiate fine-tuning runs on-demand from your application or terminal, and incur charges only for GPUs when they are actively utilized. Simplify the definition and execution of hyperparameter sweeps

Storage for models

Store fine-tuned weights or LoRA adapters in Dibtrun. Volume, as effortlessly as writing to local disk. Volumes are optimized for high read throughput, ensuring cold-start times .