Fine-tune an open model for your taskAmish Bemelkhedkar
For when prompting a general model is not accurate enough, or costs too much per call.
I build the training set, fine-tune an open model with QLoRA, and evaluate it against your current baseline on held-out data. That includes a check that the model has not gotten worse at everything else, which is the failure most fine-tuning work misses.
You get the trained weights, the training and evaluation code, and a before and after comparison. I need examples of the task, or the raw material to build them from.
Turnaround: 3 weeks.
Starting at$1,200
Duration3 weeks
Tags
PyTorch
AI Engineer
Data Scientist
Service provided by
Amish Bemelkhedkar Bengaluru, India
Fine-tune an open model for your taskAmish Bemelkhedkar
Starting at$1,200
Duration3 weeks
Tags
PyTorch
AI Engineer
Data Scientist
For when prompting a general model is not accurate enough, or costs too much per call.
I build the training set, fine-tune an open model with QLoRA, and evaluate it against your current baseline on held-out data. That includes a check that the model has not gotten worse at everything else, which is the failure most fine-tuning work misses.
You get the trained weights, the training and evaluation code, and a before and after comparison. I need examples of the task, or the raw material to build them from.
Turnaround: 3 weeks.
$1,200