conversations_V3.jsonl upload

used the dataset-cleaners.py script from AiAF/4chan-boards-sft-datasets scripts directory (https://huggingface.co/datasets/AiAF/4chan-boards-sft-datasets/blob/main/scripts/dataset-cleaners.py) to remove trailing user turns from the end of all samples (this breaks Axolotl Trainer dataset preprocessing).

.gitattributes

@@ -58,3 +58,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
 conversations_V2.jsonl filter=lfs diff=lfs merge=lfs -text
+conversations_V3.jsonl filter=lfs diff=lfs merge=lfs -text

conversations_V3.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:799b0258cb0d18a32ca7306e071e16df9899772eff3fa9c557c333b739515704
+size 2028164283