Compute & FlatkeyMarketplace
Templates, ready to run
Each template is a reviewed image, command and GPU under one name. Point it at your data and submit it with your key. You pay for the GPU time the job uses, nothing more.
4 templates · published by Unify Compute
batch-embedAvailableBatch embeddings with vLLM
Embeds an OpenAI batch file of /v1/embeddings requests with vLLM's batch runner and writes one result line per request.
- GPU
- NVIDIA A10G · 24 GB
- Default runtime
- 1 h
batch-infer-loraAvailableBatch chat completions with your LoRA adapter
Runs an OpenAI batch file of /v1/chat/completions requests through vLLM with a LoRA adapter applied (for example the adapter.tar.gz a finetune-lora job produced) and writes one result line per request, like batch-infer-vllm.
- GPU
- NVIDIA A10G · 24 GB
- Default runtime
- 1 h
batch-infer-vllmAvailableBatch chat completions with vLLM
Runs an OpenAI batch file of /v1/chat/completions requests through vLLM's batch runner and writes one result line per request.
- GPU
- NVIDIA A10G · 24 GB
- Default runtime
- 1 h
finetune-loraAvailableLoRA fine-tune with evaluation
The reference fine-tune: zero-shot and 3-shot baselines, a gate that stops if few-shot is already good enough, LoRA SFT with TRL and PEFT, and an eval with the adapter.
- GPU
- NVIDIA A10G · 24 GB
- Default runtime
- 2 h
GPUs
Choose the GPU for the job
| GPU | Memory | vCPU · RAM | Tier | Billing | Status |
|---|---|---|---|---|---|
NVIDIA A10Ga10g-24gb-x1 | 24 GB | 4 · 16 GB | Hyperscaler | On-demand | Available |
NVIDIA A100 80GBa100-80gb-x1-neo | 80 GB | 30 · 200 GB | Neocloud | On-demand | Coming soon |
NVIDIA A10Ga10g-24gb-x1-spot | 24 GB | 4 · 16 GB | Hyperscaler | Spot | Coming soon |
NVIDIA L4l4-24gb-x1 | 24 GB | 4 · 16 GB | Hyperscaler | On-demand | Coming soon |
Running a template needs an account with GPU job execution enabled. It is turned on per account while it rolls out; the quick start says how to ask.