Skip to main content
LLM
Toolkit
Search
/
Home
/
Training
Training Methods
Choose the right training approach for your goal and hardware.
WHAT ARE YOU TRYING TO CHANGE?
Model Knowledge
Add new facts, languages, domains
Behaviour / Instructions
Teach the model to follow formats
Limited GPU Memory
Work around hardware constraints
Preference Alignment
Make outputs safer, more helpful
Size / Efficiency
Distill or compress the model
New Foundation Model
Pretrain from scratch
TRAINING METHODS REFERENCE
PLANNER
Training Configuration
Goal
Pretraining
Continued Pretraining
SFT
LoRA
QLoRA
DPO
Model Size (B)
Dataset Size (tokens)
GPU VRAM (GB)
GPU Count
Precision
BF16
FP16
FP32
Generate Training Plan
Next actions
Open Training Planner →
Estimate VRAM →
Learn LoRA →