nvidia/RADIO
45283
1{2 "adaptor_names": null,3 "architectures": [4 "RADIOModel"5 ],6 "args": {7 "aa": null,8 "amp": false,9 "amp_dtype": "float16",10 "amp_impl": "native",11 "aug_repeats": 0,12 "aug_splits": 0,13 "auto_loss_balance_mode": "manual",14 "batch_size": 32,15 "bn_eps": null,16 "bn_momentum": null,17 "cache_dir": null,18 "channels_last": false,19 "checkpoint_hist": 10,20 "chk_keep_forever": 10,21 "class_map": "",22 "clip_grad": null,23 "clip_mode": "norm",24 "cls_token_per_teacher": true,25 "coco_annotations_file": "/datasets/coco2017-adlsa/annotations/captions_val2017.json",26 "coco_image_dir": "/datasets/coco2017-adlsa/val2017",27 "color_jitter": 0.4,28 "cooldown_epochs": 0,29 "cpe_max_size": 2048,30 "crd_loss": false,31 "crd_loss_weight": 0.8,32 "crop_pct": null,33 "cutmix": 0.0,34 "cutmix_minmax": null,35 "data_dir": [36 [37 "/lustre/fsw/portfolios/llmservice/projects/llmservice_nlp_fm/datasets/captioning/datacomp/dc1b/stage2",38 0.9539 ],40 [41 "/lustre/fsw/portfolios/llmservice/projects/llmservice_nlp_fm/datasets/segmentation/sam/stage1",42 0.0543 ]44 ],45 "dataset": "nvgpt4",46 "dataset_download": false,47 "debug_full_knn": false,48 "decay_epochs": 90,49 "decay_milestones": [50 90,51 180,52 27053 ],54 "decay_rate": 0.1,55 "device": "cuda:0",56 "dist_bn": "reduce",57 "distributed": true,58 "drop": 0.0,59 "drop_block": null,60 "drop_connect": null,61 "drop_path": null,62 "dtype": "bfloat16",63 "epoch_repeats": 0.0,64 "epochs": 50,65 "eval": false,66 "eval_metric": "knn_top1",67 "eval_teacher": false,68 "eval_teacher_only": false,69 "eval_throughput": false,70 "experiment": "checkpoints",71 "fast_norm": false,72 "feature_summarizer": "cls_token",73 "feature_upscale_factor": null,74 "force_new_wandb_id": false,75 "force_spectral_reparam": false,76 "fuser": "",77 "gp": null,78 "grad_accum_steps": 1,79 "grad_checkpointing": false,80 "head_init_bias": null,81 "head_init_scale": null,82 "hflip": 0.5,83 "img_size": null,84 "in_chans": null,85 "initial_checkpoint": "",86 "input_size": null,87 "interpolation": "",88 "layer_decay": null,89 "local_rank": 0,90 "log_interval": 50,91 "log_mlflow": false,92 "log_wandb": true,93 "loss": "cosine",94 "loss_auto_balance": false,95 "lr": 0.001,96 "lr_base": 0.1,97 "lr_base_scale": "",98 "lr_base_size": 256,99 "lr_cycle_decay": 0.5,100 "lr_cycle_limit": 1,101 "lr_cycle_mul": 1.0,102 "lr_k_decay": 1.0,103 "lr_noise": null,104 "lr_noise_pct": 0.67,105 "lr_noise_std": 1.0,106 "mean": null,107 "mesa": {108 "gaussian_kl": false,109 "start_epoch": 100110 },111 "min_lr": 0,112 "mixup": 0.0,113 "mixup_mode": "batch",114 "mixup_off_epoch": 0,115 "mixup_prob": 1.0,116 "mixup_switch_prob": 0.5,117 "mlp_hidden_size": 1520,118 "mlp_num_inner": 3,119 "mlp_version": "v2",120 "model": "vit_huge_patch16_224_mlpnorm",121 "model_ema": {122 "decay": 0.9998,123 "force_cpu": false,124 "power": false,125 "power_stds": [126 0.05,127 0.1128 ],129 "start_epoch": 0130 },131 "model_kwargs": {},132 "model_norm": true,133 "momentum": 0.9,134 "no_aug": false,135 "no_ddp_bb": false,136 "no_prefetcher": false,137 "no_resume_opt": false,138 "num_classes": null,139 "opt": "lamb",140 "opt_betas": null,141 "opt_eps": null,142 "opt_kwargs": {143 "filter_bias_and_bn": false144 },145 "output": "/lustre/fs6/portfolios/llmservice/users/mranzinger/output/evfm/ohem/3-13-24_vit-h-16_bf16_ep50",146 "patience_epochs": 10,147 "pin_mem": false,148 "prefetcher": true,149 "pretrained": false,150 "rank": 0,151 "ratio": [152 0.75,153 1.3333333333333333154 ],155 "recount": 1,156 "recovery_interval": 0,157 "register_multiple": 16,158 "remode": "pixel",159 "reprob": 0.0,160 "resplit": false,161 "resume": "/lustre/fs6/portfolios/llmservice/users/mranzinger/output/evfm/ohem/3-13-24_vit-h-16_bf16_ep50/checkpoints/checkpoint-48.pth.tar",162 "save_images": false,163 "scale": [164 0.5,165 1.0166 ],167 "sched": "cosine",168 "sched_on_updates": true,169 "seed": 42,170 "smoothing": 0.1,171 "spectral_reparam": false,172 "split_bn": false,173 "start_epoch": null,174 "std": null,175 "steps_per_epoch": 2000,176 "sync_bn": false,177 "synchronize_step": true,178 "teachers": [179 {180 "amp": true,181 "amp_dtype": "bfloat16",182 "batch_size": 16,183 "data_dir": [184 [185 "/lustre/fsw/portfolios/llmservice/projects/llmservice_nlp_fm/datasets/captioning/datacomp/dc1b/stage2",186 0.95187 ],188 [189 "/lustre/fsw/portfolios/llmservice/projects/llmservice_nlp_fm/datasets/segmentation/sam/stage1",190 0.05191 ]192 ],193 "fd_loss_weight": 1.0,194 "fd_normalize": false,195 "feature_distillation": true,196 "input_size": 378,197 "match_pre_proj": false,198 "model": "ViT-H-14-378-quickgelu",199 "name": "clip",200 "pretrained": "dfn5b",201 "sample_rate": 16,202 "student_resolution": 432,203 "summary_loss_weight": 1.0,204 "torchcompile": true,205 "type": "open_clip",206 "vitdet_prob": 0.05,207 "vitdet_window_sizes": [208 3,209 9,210 9,211 9212 ]213 },214 {215 "amp": false,216 "amp_dtype": "bfloat16",217 "batch_size": 16,218 "fd_loss_weight": 0.8,219 "fd_normalize": false,220 "feature_distillation": true,221 "input_size": 336,222 "match_pre_proj": false,223 "model": "ViT-L/14@336px",224 "name": "openai_clip",225 "pretrained": "openai",226 "sample_rate": 16,227 "summary_loss_weight": 0.8,228 "torchcompile": true,229 "type": "openai_clip",230 "use_summary": false231 },232 {233 "amp": true,234 "amp_dtype": "bfloat16",235 "batch_size": 16,236 "fd_loss_weight": 2.0,237 "fd_normalize": false,238 "feature_distillation": true,239 "input_size": 378,240 "model": "dinov2_vitg14_reg",241 "name": "dino_v2",242 "sample_rate": 16,243 "summary_loss_weight": 1.0,244 "torchcompile": true,245 "type": "dino_v2"246 },247 {248 "amp": false,249 "batch_size": 2,250 "data_dir": [251 [252 "/lustre/fsw/portfolios/llmservice/projects/llmservice_nlp_fm/datasets/segmentation/sam/stage1",253 0.4254 ]255 ],256 "fd_loss_fn": "MSE",257 "fd_loss_weight": 0.25,258 "fd_normalize": false,259 "fd_ohem": true,260 "feature_distillation": true,261 "input_size": 1024,262 "model": "vit-h",263 "name": "sam",264 "sample_rate": 2,265 "student_resolution": 1024,266 "summary_loss_weight": 1e-05,267 "type": "sam",268 "use_summary": false,269 "vitdet_prob": 0.99,270 "vitdet_window_sizes": [271 8,272 16,273 16274 ]275 },276 {277 "amp": true,278 "batch_size": 2,279 "data_dir": [280 [281 "/lustre/fsw/portfolios/llmservice/projects/llmservice_nlp_fm/datasets/ocr/publaynet/webdataset",282 0.4283 ],284 [285 "/lustre/fsw/portfolios/llmservice/projects/llmservice_nlp_fm/datasets/ocr/staging/arxiv/hocr",286 0.4287 ],288 [289 "/lustre/fsw/portfolios/llmservice/projects/llmservice_nlp_fm/datasets/ocr/scene-text/scene-text/text_ocr/webdataset",290 0.15291 ],292 [293 "/lustre/fsw/portfolios/llmservice/projects/llmservice_nlp_fm/datasets/ocr/scene-text/scene-text/hiertext/webdataset",294 0.05295 ]296 ],297 "fd_loss_fn": "MSE",298 "fd_loss_weight": 0.13,299 "fd_normalize": false,300 "fd_ohem": true,301 "fd_upsample_factor": 4,302 "feature_distillation": true,303 "input_size": 1024,304 "model": "quality",305 "name": "rtx-translate",306 "sample_rate": 2,307 "student_resolution": 1024,308 "summary_loss_weight": 1e-05,309 "type": "rtx_translate",310 "use_summary": false,311 "vitdet_prob": 0.99,312 "vitdet_window_sizes": [313 8,314 16,315 16316 ]317 }318 ],319 "torchcompile": null,320 "torchscript": false,321 "train_interpolation": "random",322 "train_split": "train",323 "tta": 0,324 "use_coco": false,325 "use_multi_epochs_loader": false,326 "val_data_dir": "/lustre/fsw/portfolios/llmservice/projects/llmservice_nlp_fm/datasets/classification/imagenet-1k/webdataset",327 "val_ema_only": false,328 "val_img_size": 432,329 "val_jobs_script": "run_validation_jobs_vit-h-16.sh",330 "val_split": "val",331 "validation_batch_size": 64,332 "vflip": 0.0,333 "wandb_entity": "",334 "wandb_group": "ohem",335 "wandb_job_type": "",336 "wandb_name": "",337 "wandb_project": "",338 "warmup_epochs": 0.5,339 "warmup_lr": 1e-05,340 "warmup_prefix": false,341 "weight_decay": 0.02,342 "worker_seeding": "all",343 "workers": 8,344 "world_size": 128345 },346 "auto_map": {347 "AutoConfig": "hf_model.RADIOConfig",348 "AutoModel": "hf_model.RADIOModel"349 },350 "max_resolution": 2048,351 "patch_size": 16,352 "preferred_resolution": [353 432,354 432355 ],356 "torch_dtype": "bfloat16",357 "transformers_version": "4.40.1",358 "version": "radio_v2.1",359 "vitdet_window_size": null360}361