Let's finetune and deploy DeepSeek R1 (8B) for under $10
Small language models are super capable and don't have to break the bank! In this tutorial, I show how to finetune and deploy DeepSeek R1 (8B) parameters (which is a small LM), in under $10 and 2 hours.
The longer you finetune, the better the model gets, but I have found that 2 hours gets you pretty good results.
People shouldn't be so scared to finetune models, it can be cheap and fast if done well.
00:00 - intro
01:00 - enable autoscale to zero (serverless)
01:20 - upload JSONL file
02:00 - deploy the model (no-code)
02:30 - monitor cold start
03:10 - monitor logs
03:20 - audit events
03:…
Watch on YouTube ↗
(saves to browser)
Chapters (13)
intro
1:00
enable autoscale to zero (serverless)
1:20
upload JSONL file
2:00
deploy the model (no-code)
2:30
monitor cold start
3:10
monitor logs
3:20
audit events
3:30
wait for deployment
4:50
add your own serving engine (LitServe)
5:30
training replica
5:40
deployment replica
7:46
test the model with the API playground
8:00
cost discussing ($0.45)
DeepCamp AI