time ilab data generate
shuf -n 15000 .local/share/instructlab/datasets/`ls -1 .local/share/instructlab/datasets/ | head -n1`/skills_train_msgs_*.jsonl > .local/share/instructlab/datasets/`ls -1 .local/share/instructlab/datasets/ | head -n1`/skills_train_msgs_reduced.jsonl

time ilab model train -y --force-clear-phased-cache --enable-serving-output --strategy lab-multiphase --phased-phase1-data ~/.local/share/instructlab/datasets/`ls -1 ~/.local/share/instructlab/datasets/ | head -n1`/knowledge_train_msgs_*.jsonl --phased-phase2-data ~/.local/share/instructlab/datasets/`ls -1 .local/share/instructlab/datasets/ | head -n1`/skills_train_msgs_reduced.jsonl --phased-phase1-num-epochs 2 --phased-phase2-num-epochs 2

 torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1.33 GiB. GPU 6 has a total capacity of 39.38 GiB of which 571.38 MiB is free. Including non-PyTorch memory, this process has 0 bytes memory in use. Of the allocated memory 34.54 GiB is allocated by PyTorch, and 2.65 GiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation.  See documentation for Memory Management  (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables)