|
{ |
|
"data_root_dir": "/scr/user/data", |
|
"hf_token": ".hf_token", |
|
"pretrained_checkpoint": "", |
|
"resume_epoch": null, |
|
"resume_step": null, |
|
"run_id": "prism-dinosiglip-224px+mx-oxe-magic-soup-plus+n8+b32+x7", |
|
"run_id_note": null, |
|
"run_root_dir": "./runs", |
|
"save_interval": 2500, |
|
"seed": 7, |
|
"stage": "vla-full-train", |
|
"trackers": [ |
|
"jsonl", |
|
"wandb" |
|
], |
|
"vla": { |
|
"base_vlm": "prism-dinosiglip-224px+7b", |
|
"data_mix": "oxe_magic_soup_plus_minus", |
|
"enable_gradient_checkpointing": true, |
|
"enable_mixed_precision_training": true, |
|
"epochs": 1000, |
|
"expected_world_size": 64, |
|
"freeze_vision_backbone": false, |
|
"global_batch_size": 2048, |
|
"learning_rate": 2e-05, |
|
"lr_scheduler_type": "constant", |
|
"max_grad_norm": 1.0, |
|
"max_steps": null, |
|
"per_device_batch_size": 32, |
|
"reduce_in_full_precision": true, |
|
"shuffle_buffer_size": 256000, |
|
"train_strategy": "fsdp-full-shard", |
|
"type": "prism-dinosiglip-224px+mx-oxe-magic-soup-plus", |
|
"vla_id": "prism-dinosiglip-224px+mx-oxe-magic-soup-plus", |
|
"warmup_ratio": 0.0, |
|
"weight_decay": 0.0 |
|
}, |
|
"wandb_entity": "", |
|
"wandb_project": "" |
|
} |
|
|