MCPcopy
hub / github.com/FareedKhan-dev/train-llm-from-scratch / train_sft.py

File train_sft.py

scripts/train_sft.py:None–None  ·  view source on GitHub ↗

Source from the content-addressed store, hash-verified

1"""
2Supervised Fine-Tuning of the pretrained base on packed instruction data.
3
4Loads the base checkpoint, trains with the prompt-masked SFT loss, periodically reports

Callers

nothing calls this directly

Calls 1

mainFunction · 0.70

Tested by

no test coverage detected