4bit quantized support (wip)

2023-04-17 11:37:39 -04:00
parent 12de7b7cf7
commit 77fca25f1b
11 changed files with 108 additions and 63 deletions
--- a/src/axolotl/datasets.py
+++ b/src/axolotl/datasets.py
@@ -31,6 +31,7 @@ class TokenizedPromptDataset(IterableDataset):
                pass


+# TODO this isn't the best since it can't interleave datasets
 class ConstantLengthDataset(IterableDataset):
    """
    Iterable dataset that returns constant length chunks of tokens from stream of text files.