âšī¸ Fine-tune models using SageMaker serverless (Monarch) API. Llama 3.2 3B Instruct is fine-tuned with SFT + LoRA using the Bitext intent dataset already in S3. Jobs typically take ~30 minutes. After training, the model is auto-imported to Bedrock.
1. Choose a Dataset
Select a preset or enter a custom HuggingFace dataset below.
2. Custom Dataset (Optional)
Override with any HuggingFace dataset. Fields map to the Converse API format.