Upload training_args.json

training_args.json

@@ -0,0 +1,106 @@
+{
+  "cmdline": "train.py --network flexivit_reg1_s16_rms_ls --tag dino-v2-il-all --opt adamw --lr 0.0004 --lr-scheduler-update iter --lr-scheduler cosine --lr-cosine-min 1e-7 --batch-size 128 --warmup-epochs 10 --epochs 100 --size 240 --wd 0.05 --norm-wd 0 --grad-accum-steps 4 --smoothing-alpha 0.1 --mixup-alpha 0.8 --cutmix --aug-level 8 --model-ema --ra-sampler --ra-reps 2 --clip-grad-norm 1 --model-config min_patch_size=10,max_patch_size=40 --amp --amp-dtype bfloat16 --compile --rgb-mode none --layer-decay 0.7 --resume-epoch 0 --data-path data/training_il-all_packed --val-path data/validation_il-all_packed",
+  "network": "flexivit_reg1_s16_rms_ls",
+  "net_param": null,
+  "model_config": {
+    "min_patch_size": 10,
+    "max_patch_size": 40
+  },
+  "pretrained": false,
+  "reset_head": false,
+  "freeze_body": false,
+  "freeze_stages": null,
+  "unfreeze_features": false,
+  "compile": true,
+  "compile_opt": false,
+  "opt": "adamw",
+  "momentum": 0.9,
+  "nesterov": false,
+  "opt_eps": null,
+  "opt_betas": null,
+  "opt_alpha": null,
+  "lr": 0.0004,
+  "bias_lr": null,
+  "lr_scale": null,
+  "lr_scale_type": "linear",
+  "wd": 0.05,
+  "norm_wd": 0.0,
+  "bias_weight_decay": null,
+  "transformer_embedding_decay": null,
+  "layer_decay": 0.7,
+  "lr_scheduler_update": "iter",
+  "lr_scheduler": "cosine",
+  "lr_step_size": 40,
+  "lr_steps": null,
+  "lr_step_gamma": 0.75,
+  "lr_cosine_min": 1e-07,
+  "lr_power": 1.0,
+  "grad_accum_steps": 4,
+  "channels": 3,
+  "size": [
+    240,
+    240
+  ],
+  "freeze_bn": false,
+  "sync_bn": false,
+  "batch_size": 128,
+  "warmup_epochs": 10,
+  "smoothing_alpha": 0.1,
+  "mixup_alpha": 0.8,
+  "cutmix": true,
+  "aug_type": "birder",
+  "aug_level": 8,
+  "use_grayscale": false,
+  "ra_num_ops": 2,
+  "ra_magnitude": 9,
+  "augmix_severity": 3,
+  "resize_min_scale": null,
+  "re_prob": null,
+  "simple_crop": false,
+  "rgb_mode": "none",
+  "bce_loss": false,
+  "bce_threshold": 0.0,
+  "epochs": 100,
+  "stop_epoch": 101,
+  "save_frequency": 5,
+  "keep_last": null,
+  "resume_epoch": 0,
+  "load_states": false,
+  "load_scheduler": false,
+  "model_ema": true,
+  "model_ema_steps": 32,
+  "model_ema_decay": 0.9999,
+  "ra_sampler": true,
+  "ra_reps": 2,
+  "tag": "dino-v2-il-all",
+  "log_interval": 50,
+  "num_workers": 8,
+  "prefetch_factor": null,
+  "drop_last": false,
+  "model_dtype": "float32",
+  "amp": true,
+  "amp_dtype": "bfloat16",
+  "fast_matmul": false,
+  "grad_anomaly_detection": false,
+  "world_size": 2,
+  "dist_url": "env://",
+  "clip_grad_norm": 1.0,
+  "gpu": 0,
+  "cpu": false,
+  "use_deterministic_algorithms": false,
+  "plot_lr": false,
+  "no_summary": false,
+  "val_path": "data/validation_il-all_packed",
+  "data_path": "data/training_il-all_packed",
+  "wds": false,
+  "wds_info": null,
+  "wds_class_file": null,
+  "wds_cache_dir": null,
+  "wds_train_size": null,
+  "wds_val_size": null,
+  "wds_training_split": "training",
+  "wds_val_split": "validation",
+  "rank": 0,
+  "distributed": true,
+  "dist_backend": "nccl"
+}