Skip to content

Latest commit

 

History

History
22 lines (15 loc) · 655 Bytes

README.md

File metadata and controls

22 lines (15 loc) · 655 Bytes

RWKV-Notebooks

Uses jsonl and binidx instead of the garbage RAM-intensive scripts.

Example.jsonl file contents:

{"text": "This is a sentence"}
{"text": "Instruction: a\n\nInput: b\n\nResponse: c"}
{"text": "Question: a\n\nAnswer: b"}

The tokenizer will combine all jsonl files inside your dataset folder into two files the train script will read. Read the outputs of the cells to know what to do.


RWKV-infctx:

Only tested 0.4B-World

Open In Colab

RWKV-LorA:

Soon™