Nov 3, 2025

1. 19:45

   Persist phase completion state to resume reliably

   Vaishnav14220
   bef2610
2. 17:54

   Tune training hyperparameters for L4 GPU

   Vaishnav14220
   bd72f86
3. 13:57

   Push datasets via push_to_hub and load from Hub on resume

   Vaishnav14220
   fdbfba8
4. 06:04

   Use sacrebleu directly for training metrics

   Vaishnav14220
   55ff09b
5. 03:26

   Upload cached datasets and allow resumable downloads

   Vaishnav14220
   9b99b56
6. 02:56

   Align training arguments with latest Transformers API

   Vaishnav14220
   a3db032
7. 00:35

   Silence repeated RDKit warnings during dataset prep

   Vaishnav14220
   2b02a39
8. 00:27

   Orchestrate full ORD training pipeline in Space

   Vaishnav14220
   29a351f