| [ |
| "/opt/anaconda3/envs/dlb-candi/bin/python", |
| "-u", |
| "/home/ruizzhao/diffusion_baseline/upstreams/candi/main.py", |
| "mode=train", |
| "data=openwebtext-split", |
| "data.cache_dir=/home/ruizzhao/diffusion_baseline/data/training_cache/candi/owt", |
| "data.tokenizer_name_or_path=/home/ruizzhao/diffusion_baseline/data/raw/huggingface/hub/models--gpt2/snapshots/607a30d783dfa663caf39e06633721c8d4cfcd7e", |
| "model=small", |
| "model.length=1024", |
| "loader.global_batch_size=512", |
| "loader.eval_global_batch_size=512", |
| "loader.batch_size=16", |
| "loader.eval_batch_size=16", |
| "trainer.devices=1", |
| "trainer.num_nodes=1", |
| "trainer.accumulate_grad_batches=32", |
| "trainer.max_steps=1000000", |
| "trainer.precision=bf16", |
| "optim.lr=0.0003", |
| "lr_scheduler=constant_warmup", |
| "lr_scheduler.num_warmup_steps=2500", |
| "callbacks.checkpoint_every_n_steps.every_n_train_steps=20000", |
| "checkpointing.save_dir=/home/ruizzhao/diffusion_baseline/checkpoints/reference_reproduction/candi/owt", |
| "checkpointing.resume_from_ckpt=true", |
| "checkpointing.resume_ckpt_path=/home/ruizzhao/diffusion_baseline/checkpoints/reference_reproduction/candi/owt/checkpoints/last.ckpt", |
| "hydra.run.dir=/home/ruizzhao/diffusion_baseline/checkpoints/reference_reproduction/candi/owt", |
| "hydra.job.chdir=true", |
| "wandb=null", |
| "seed=42", |
| "scratch_dir=/home/ruizzhao/diffusion_baseline/work/candi", |
| "algo=candi", |
| "algo.sampler=cached", |
| "algo.mixed_coeff=0.5", |
| "algo.step_size=1.0", |
| "algo.temp=1.0" |
| ] |
|
|
