Question About Training Dataset
#3
by
Clausss
- opened
Hi, I have a question. Did you use 100% of the rows from the following datasets for training this model?
THUDM/webglm-qa
databricks/databricks-dolly-15k
cognitivecomputations/wizard_vicuna_70k_unfiltered
totally-not-an-llm/EverythingLM-data-V3
Amod/mental_health_counseling_conversations
sablo/oasst2_curated
starfishmedical/webGPT_x_dolly
Open-Orca/OpenOrca
mlabonne/chatml_dpo_pairs
Hi, @Clausss !
100% of all, except the Open-Orca/OpenOrca, which was too big, so I used just random rows of that dataset.
About the training parameters, unfortunately, I haven't recorded which parameters I used for training this one, but it was similar to the ones I used to train Felladrin/Pythia-31M-Chat-v1 (the training parameters can be found on the Readme).
Thanks
Clausss
changed discussion status to
closed