{"run_name": "v1_2_nano", "model": {"encoder_config": {"supported_modality_names": ["sentinel2_l2a", "sentinel1", "landsat", "worldcover", "srtm", "openstreetmap_raster", "wri_canopy_height_map", "cdl", "worldcereal"], "embedding_size": 128, "max_patch_size": 8, "min_patch_size": 1, "num_heads": 8, "mlp_ratio": 4.0, "depth": 4, "drop_path": 0.1, "max_sequence_length": 12, "num_register_tokens": 0, "learnable_channel_embeddings": true, "random_channel_embeddings": false, "num_projection_layers": 1, "aggregate_then_project": true, "use_flash_attn": false, "frozen_patch_embeddings": false, "qk_norm": false, "log_token_norm_stats": false, "tokenization_config": {"overrides": {"sentinel2_l2a": {"band_groups": [["B02", "B03", "B04", "B08", "B05", "B06", "B07", "B8A", "B11", "B12", "B01", "B09"]], "_CLASS_": "olmoearth_pretrain.nn.tokenization.ModalityTokenization"}, "landsat": {"band_groups": [["B8", "B1", "B2", "B3", "B4", "B5", "B6", "B7", "B9", "B10", "B11"]], "_CLASS_": "olmoearth_pretrain.nn.tokenization.ModalityTokenization"}}, "_CLASS_": "olmoearth_pretrain.nn.tokenization.TokenizationConfig"}, "use_linear_patch_embed": true, "band_dropout_rate": 0.2, "random_band_dropout": true, "band_dropout_modalities": ["sentinel2_l2a", "landsat"], "patch_embed_hidden_sizes": [12], "spatial_pos_encoding": "rope_3d_mixed", "rope_base": 10000.0, "rope_coordinate_scale": 1.0, "rope_mixed_base": 10.0, "temporal_rope_dim_frac": 0.25, "rope_temporal_coordinate_scale": 0.03333333333333333, "_CLASS_": "olmoearth_pretrain.nn.flexi_vit.EncoderConfig"}, "decoder_config": {"supported_modality_names": ["sentinel2_l2a", "sentinel1", "landsat", "worldcover", "srtm", "openstreetmap_raster", "wri_canopy_height_map", "cdl", "worldcereal"], "encoder_embedding_size": 128, "decoder_embedding_size": 128, "depth": 4, "mlp_ratio": 4.0, "num_heads": 8, "max_sequence_length": 12, "drop_path": 0.0, "learnable_channel_embeddings": true, "random_channel_embeddings": false, "use_flash_attn": false, "qk_norm": false, "tokenization_config": {"overrides": {"sentinel2_l2a": {"band_groups": [["B02", "B03", "B04", "B08", "B05", "B06", "B07", "B8A", "B11", "B12", "B01", "B09"]], "_CLASS_": "olmoearth_pretrain.nn.tokenization.ModalityTokenization"}, "landsat": {"band_groups": [["B8", "B1", "B2", "B3", "B4", "B5", "B6", "B7", "B9", "B10", "B11"]], "_CLASS_": "olmoearth_pretrain.nn.tokenization.ModalityTokenization"}}, "_CLASS_": "olmoearth_pretrain.nn.tokenization.TokenizationConfig"}, "spatial_pos_encoding": "rope_3d_mixed", "rope_base": 10000.0, "rope_coordinate_scale": 1.0, "rope_mixed_base": 10.0, "temporal_rope_dim_frac": 0.25, "rope_temporal_coordinate_scale": 0.03333333333333333, "_CLASS_": "olmoearth_pretrain.nn.flexi_vit.PredictorConfig"}, "_CLASS_": "olmoearth_pretrain.nn.latent_mim.LatentMIMConfig"}, "dataset": {"h5py_dir": "/weka/dfive-default/helios/dataset/osm_sampling/h5py_data_w_missing_timesteps_zstd_3_128_x_4/cdl_gse_landsat_openstreetmap_raster_sentinel1_sentinel2_l2a_srtm_worldcereal_worldcover_worldpop_wri_canopy_height_map/1138828", "training_modalities": ["sentinel2_l2a", "sentinel1", "landsat", "worldcover", "srtm", "openstreetmap_raster", "wri_canopy_height_map", "cdl", "worldcereal"], "dtype": "float32", "normalize": true, "dataset_percentage": 1.0, "seed": 0, "apply_cutmix": false, "_CLASS_": "olmoearth_pretrain.data.dataset.OlmoEarthDatasetConfig"}, "data_loader": {"work_dir": "/weka/dfive-default/olmoearth_pretrain/checkpoints/yawenzzzz/v1_2_nano", "global_batch_size": 512, "min_patch_size": 1, "max_patch_size": 8, "sampled_hw_p_list": [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12], "seed": 3622, "token_budget": 2250, "shuffle": true, "num_workers": 16, "prefetch_factor": 4, "drop_last": true, "num_dataset_repeats_per_epoch": 1, "masking_config": {"strategy_config": {"type": "random_time_with_decode", "encode_ratio": 0.5, "decode_ratio": 0.5, "random_ratio": 0.5, "only_decode_modalities": ["worldcover", "srtm", "openstreetmap_raster", "wri_canopy_height_map", "cdl", "worldcereal"]}, "tokenization_config": {"overrides": {"sentinel2_l2a": {"band_groups": [["B02", "B03", "B04", "B08", "B05", "B06", "B07", "B8A", "B11", "B12", "B01", "B09"]], "_CLASS_": "olmoearth_pretrain.nn.tokenization.ModalityTokenization"}, "landsat": {"band_groups": [["B8", "B1", "B2", "B3", "B4", "B5", "B6", "B7", "B9", "B10", "B11"]], "_CLASS_": "olmoearth_pretrain.nn.tokenization.ModalityTokenization"}}, "_CLASS_": "olmoearth_pretrain.nn.tokenization.TokenizationConfig"}, "_CLASS_": "olmoearth_pretrain.train.masking.MaskingConfig"}, "num_masked_views": 2, "_CLASS_": "olmoearth_pretrain.data.dataloader.OlmoEarthDataLoaderConfig"}, "train_module": {"optim_config": {"compile": false, "fixed_fields": ["initial_lr"], "lr": 0.0002, "betas": [0.9, 0.999], "eps": 1e-08, "weight_decay": 0.02, "fused": false, "_CLASS_": "olmo_core.optim.adamw.AdamWConfig"}, "rank_microbatch_size": 64, "transform_config": {"transform_type": "flip_and_rotate", "transform_kwargs": {}, "_CLASS_": "olmoearth_pretrain.data.transform.TransformConfig"}, "compile_model": false, "dp_config": {"name": "fsdp", "param_dtype": "bfloat16", "reduce_dtype": "float32", "_CLASS_": "olmo_core.distributed.parallel.data_parallel.DataParallelConfig"}, "compile_loss": false, "max_grad_norm": 1.0, "scheduler": {"lr_field": "lr", "initial_lr_field": "initial_lr", "units": "steps", "warmup": 8000, "alpha_f": 0.1, "warmup_min_lr": 0.0, "_CLASS_": "olmo_core.optim.scheduler.CosWithWarmup"}, "find_unused_parameters": true, "loss_config": {"loss_config": {"tau": 0.1, "same_target_threshold": 0.999, "mask_negatives_for_modalities": ["worldcover", "srtm", "openstreetmap_raster", "wri_canopy_height_map", "cdl", "worldcereal"]}, "_CLASS_": "olmoearth_pretrain.train.loss.LossConfig"}, "masking_config": {"strategy_config": {"type": "random_time_with_decode", "encode_ratio": 0.5, "decode_ratio": 0.5, "random_ratio": 0.5, "only_decode_modalities": ["worldcover", "srtm", "openstreetmap_raster", "wri_canopy_height_map", "cdl", "worldcereal"]}, "tokenization_config": {"overrides": {"sentinel2_l2a": {"band_groups": [["B02", "B03", "B04", "B08", "B05", "B06", "B07", "B8A", "B11", "B12", "B01", "B09"]], "_CLASS_": "olmoearth_pretrain.nn.tokenization.ModalityTokenization"}, "landsat": {"band_groups": [["B8", "B1", "B2", "B3", "B4", "B5", "B6", "B7", "B9", "B10", "B11"]], "_CLASS_": "olmoearth_pretrain.nn.tokenization.ModalityTokenization"}}, "_CLASS_": "olmoearth_pretrain.nn.tokenization.TokenizationConfig"}, "_CLASS_": "olmoearth_pretrain.train.masking.MaskingConfig"}, "token_exit_cfg": {"sentinel2_l2a": 0, "sentinel1": 0, "landsat": 0, "worldcover": 0, "srtm": 0, "openstreetmap_raster": 0, "wri_canopy_height_map": 0, "cdl": 0, "worldcereal": 0}, "ema_decay": [1.0, 1.0], "contrastive_config": {"loss_config": {"weight": 0.05}, "_CLASS_": "olmoearth_pretrain.train.loss.LossConfig"}, "reinit_targets": false, "_CLASS_": "olmoearth_pretrain.train.train_module.contrastive_latentmim.ContrastiveLatentMIMTrainModuleConfig"}, "trainer": {"save_folder": "/weka/dfive-default/olmoearth_pretrain/checkpoints/yawenzzzz/v1_2_nano", "work_dir": "/weka/dfive-default/olmoearth_pretrain/checkpoints/yawenzzzz/v1_2_nano", "load_strategy": "if_available", "checkpointer": {"work_dir": "/weka/dfive-default/olmoearth_pretrain/checkpoints/yawenzzzz/v1_2_nano", "pre_download": false, "throttle_uploads": false, "_CLASS_": "olmo_core.train.checkpoint.CheckpointerConfig"}, "save_overwrite": false, "max_duration": {"value": 300, "unit": "epochs", "_CLASS_": "olmo_core.train.common.Duration"}, "cancel_check_interval": 25, "metrics_collect_interval": 10, "callbacks": {"wandb": {"enabled": true, "name": "v1_2_nano", "project": "2026_06_12_v1_2_size_sweep", "entity": "eai-ai2", "cancel_tags": ["cancel", "canceled", "cancelled"], "upload_dataset_distribution_pre_train": true, "upload_modality_data_band_distribution_pre_train": false, "restart_on_same_run": true, "_CLASS_": "olmoearth_pretrain.train.callbacks.wandb.OlmoEarthWandBCallback"}, "speed_monitor": {"_CLASS_": "olmoearth_pretrain.train.callbacks.speed_monitor.OlmoEarthSpeedMonitorCallback"}, "gpu_memory_monitor": {"_CLASS_": "olmo_core.train.callbacks.gpu_memory_monitor.GPUMemoryMonitorCallback"}, "config_saver": {"fname": "config.json", "_CLASS_": "olmo_core.train.callbacks.config_saver.ConfigSaverCallback"}, "downstream_evaluator": {"tasks": {"m-eurosat": {"dataset": "m-eurosat", "num_workers": 0, "pooling_type": "mean", "norm_stats_from_pretrained": true, "input_modalities": ["sentinel2_l2a"], "input_layers": [], "embedding_batch_size": 128, "probe_batch_size": 32, "linear_probe_eval_interval": 50, "ft_batch_size": 32, "finetune_seed": 42, "epochs": 50, "patch_size": 4, "eval_interval": {"value": 4000, "unit": "steps", "_CLASS_": "olmo_core.train.common.Duration"}, "eval_mode": "knn", "probe_type": "linear", "use_pooled_tokens": false, "label_fraction": 1.0, "norm_method": "norm_no_clip_2_std", "select_best_by_primary_metric": false, "max_train_samples_seed": 42, "quantize_embeddings": false, "use_dice_loss": false, "primary_metric": "accuracy", "pretrain_max_samples": 512, "pretrain_label_seed": 42, "pretrain_train_samples": 512, "pretrain_valid_samples": 512, "pretrain_test_samples": 512, "pretrain_split_strategy": "random", "pretrain_geographic_bin_size_deg": 5.0, "_CLASS_": "olmoearth_pretrain.train.callbacks.evaluator_callback.DownstreamTaskConfig"}, "m_so2sat": {"dataset": "m-so2sat", "num_workers": 4, "pooling_type": "mean", "norm_stats_from_pretrained": true, "input_modalities": ["sentinel2_l2a"], "input_layers": [], "embedding_batch_size": 128, "probe_batch_size": 32, "linear_probe_eval_interval": 50, "ft_batch_size": 32, "finetune_seed": 42, "epochs": 50, "patch_size": 4, "eval_interval": {"value": 20000, "unit": "steps", "_CLASS_": "olmo_core.train.common.Duration"}, "eval_mode": "knn", "probe_type": "linear", "use_pooled_tokens": false, "label_fraction": 1.0, "norm_method": "norm_no_clip_2_std", "select_best_by_primary_metric": false, "max_train_samples_seed": 42, "quantize_embeddings": false, "use_dice_loss": false, "primary_metric": "accuracy", "pretrain_max_samples": 512, "pretrain_label_seed": 42, "pretrain_train_samples": 512, "pretrain_valid_samples": 512, "pretrain_test_samples": 512, "pretrain_split_strategy": "random", "pretrain_geographic_bin_size_deg": 5.0, "_CLASS_": "olmoearth_pretrain.train.callbacks.evaluator_callback.DownstreamTaskConfig"}, "mados": {"dataset": "mados", "num_workers": 8, "pooling_type": "mean", "norm_stats_from_pretrained": false, "input_modalities": ["sentinel2_l2a"], "input_layers": [], "embedding_batch_size": 128, "probe_lr": 0.01, "probe_batch_size": 128, "linear_probe_eval_interval": 50, "ft_batch_size": 32, "finetune_seed": 42, "epochs": 50, "patch_size": 4, "eval_interval": {"value": 4000, "unit": "steps", "_CLASS_": "olmo_core.train.common.Duration"}, "eval_mode": "linear_probe", "probe_type": "linear", "use_pooled_tokens": false, "label_fraction": 1.0, "norm_method": "norm_no_clip_2_std", "select_best_by_primary_metric": false, "max_train_samples_seed": 42, "quantize_embeddings": false, "use_dice_loss": false, "primary_metric": "micro_f1", "pretrain_max_samples": 512, "pretrain_label_seed": 42, "pretrain_train_samples": 512, "pretrain_valid_samples": 512, "pretrain_test_samples": 512, "pretrain_split_strategy": "random", "pretrain_geographic_bin_size_deg": 5.0, "_CLASS_": "olmoearth_pretrain.train.callbacks.evaluator_callback.DownstreamTaskConfig"}, "pastis": {"dataset": "pastis", "num_workers": 2, "pooling_type": "mean", "norm_stats_from_pretrained": true, "input_modalities": ["sentinel2_l2a"], "input_layers": [], "embedding_batch_size": 32, "probe_lr": 0.1, "probe_batch_size": 8, "linear_probe_eval_interval": 50, "ft_batch_size": 32, "finetune_seed": 42, "epochs": 50, "patch_size": 4, "eval_interval": {"value": 20000, "unit": "steps", "_CLASS_": "olmo_core.train.common.Duration"}, "eval_mode": "linear_probe", "probe_type": "linear", "use_pooled_tokens": false, "label_fraction": 1.0, "norm_method": "norm_no_clip_2_std", "select_best_by_primary_metric": false, "max_train_samples_seed": 42, "quantize_embeddings": false, "use_dice_loss": false, "primary_metric": "miou", "pretrain_max_samples": 512, "pretrain_label_seed": 42, "pretrain_train_samples": 512, "pretrain_valid_samples": 512, "pretrain_test_samples": 512, "pretrain_split_strategy": "random", "pretrain_geographic_bin_size_deg": 5.0, "_CLASS_": "olmoearth_pretrain.train.callbacks.evaluator_callback.DownstreamTaskConfig"}, "yemen_crop": {"dataset": "yemen_crop", "num_workers": 2, "pooling_type": "mean", "norm_stats_from_pretrained": true, "input_modalities": ["sentinel2_l2a"], "input_layers": [], "embedding_batch_size": 32, "probe_lr": 0.001, "probe_batch_size": 8, "linear_probe_eval_interval": 50, "ft_batch_size": 32, "finetune_seed": 42, "epochs": 50, "patch_size": 4, "eval_interval": {"value": 20000, "unit": "steps", "_CLASS_": "olmo_core.train.common.Duration"}, "eval_mode": "linear_probe", "probe_type": "linear", "use_pooled_tokens": false, "label_fraction": 1.0, "norm_method": "norm_no_clip_2_std", "select_best_by_primary_metric": false, "max_train_samples_seed": 42, "quantize_embeddings": false, "use_dice_loss": false, "pretrain_max_samples": 512, "pretrain_label_seed": 42, "pretrain_train_samples": 512, "pretrain_valid_samples": 512, "pretrain_test_samples": 512, "pretrain_split_strategy": "random", "pretrain_geographic_bin_size_deg": 5.0, "_CLASS_": "olmoearth_pretrain.train.callbacks.evaluator_callback.DownstreamTaskConfig"}, "geo_ecosystem_annual_test": {"dataset": "geo_ecosystem_annual_test", "num_workers": 8, "pooling_type": "mean", "norm_stats_from_pretrained": true, "input_modalities": ["sentinel2_l2a"], "input_layers": [], "embedding_batch_size": 32, "probe_lr": 0.01, "probe_batch_size": 8, "linear_probe_eval_interval": 50, "ft_batch_size": 32, "finetune_seed": 42, "epochs": 50, "patch_size": 4, "eval_interval": {"value": 20000, "unit": "steps", "_CLASS_": "olmo_core.train.common.Duration"}, "eval_mode": "linear_probe", "probe_type": "linear", "use_pooled_tokens": false, "label_fraction": 1.0, "norm_method": "norm_no_clip_2_std", "select_best_by_primary_metric": false, "max_train_samples_seed": 42, "quantize_embeddings": false, "use_dice_loss": false, "pretrain_max_samples": 512, "pretrain_label_seed": 42, "pretrain_train_samples": 512, "pretrain_valid_samples": 512, "pretrain_test_samples": 512, "pretrain_split_strategy": "random", "pretrain_geographic_bin_size_deg": 5.0, "_CLASS_": "olmoearth_pretrain.train.callbacks.evaluator_callback.DownstreamTaskConfig"}, "canada_wildfire_sat_eval_split": {"dataset": "canada_wildfire_sat_eval_split", "num_workers": 2, "pooling_type": "mean", "norm_stats_from_pretrained": true, "input_modalities": ["sentinel2_l2a"], "input_layers": [], "embedding_batch_size": 32, "probe_lr": 0.1, "probe_batch_size": 16, "linear_probe_eval_interval": 50, "ft_batch_size": 32, "finetune_seed": 42, "epochs": 50, "patch_size": 5, "eval_interval": {"value": 20000, "unit": "steps", "_CLASS_": "olmo_core.train.common.Duration"}, "eval_mode": "linear_probe", "probe_type": "linear", "use_pooled_tokens": false, "label_fraction": 1.0, "norm_method": "norm_no_clip_2_std", "select_best_by_primary_metric": false, "max_train_samples_seed": 42, "quantize_embeddings": false, "use_dice_loss": true, "primary_metric": "class_f1", "primary_metric_class": 1, "pretrain_max_samples": 512, "pretrain_label_seed": 42, "pretrain_train_samples": 512, "pretrain_valid_samples": 512, "pretrain_test_samples": 512, "pretrain_split_strategy": "random", "pretrain_geographic_bin_size_deg": 5.0, "_CLASS_": "olmoearth_pretrain.train.callbacks.evaluator_callback.DownstreamTaskConfig"}}, "enabled": true, "eval_on_startup": false, "cancel_after_first_eval": false, "run_on_test": false, "n_bootstrap": 0, "bootstrap_seed": 42, "_CLASS_": "olmoearth_pretrain.train.callbacks.evaluator_callback.DownstreamEvaluatorCallbackConfig"}, "garbage_collector": {"gc_interval": 1, "enabled": true, "_CLASS_": "olmo_core.train.callbacks.garbage_collector.GarbageCollectorCallback"}, "beaker": {"enabled": true, "result_dir": "/results", "_CLASS_": "olmo_core.train.callbacks.beaker.BeakerCallback"}, "checkpointer": {"save_interval": 5000, "ephemeral_save_interval": 250, "remove": "ephemeral_only", "enabled": true, "_CLASS_": "olmo_core.train.callbacks.checkpointer.CheckpointerCallback"}}, "bookkeeping_soft_timeout": 30, "no_checkpoints": false, "no_evals": false, "_CLASS_": "olmo_core.train.config.TrainerConfig"}, "launch": {"name": "v1_2_nano-train-3b65ef41", "cmd": ["scripts/official/v1_2/nano.py", "train", "v1_2_nano", "ai2/jupiter", "--launch.num_gpus=8", "--launch.priority=urgent", "--launch.clusters=[ai2/jupiter]", "--trainer.callbacks.wandb.project=2026_06_12_v1_2_size_sweep"], "budget": "ai2/atec-olmoearth", "task_name": "train", "workspace": "ai2/earth-systems", "setup_steps": ["echo \"$GCP_CREDENTIALS\" > $GOOGLE_APPLICATION_CREDENTIALS", "conda install gh --channel conda-forge", "gh auth status", "gh repo clone $REPO_URL .", "git checkout \"$GIT_REF\"", "git submodule update --init --recursive", "pip install uv", "export PATH=\"/root/.local/bin:$PATH\" ", "uv sync --locked --all-extras", "venv_path=$(uv run python -c 'import sys; print(sys.executable)')", "source \"$(dirname \"$venv_path\")/activate\"", "uv pip install breizhcrops==0.0.4.1 ", "uv pip show torch", "uv run python -c 'import torch; print(torch.__version__); print(torch.version.cuda); print(torch.cuda.get_arch_list())'"], "beaker_image": "petew/olmo-core-tch271cu128-2025-09-15", "num_nodes": 1, "num_gpus": 8, "shared_memory": "256GiB", "clusters": ["ai2/jupiter"], "shared_filesystem": true, "priority": "urgent", "preemptible": true, "env_vars": [{"name": "GOOGLE_APPLICATION_CREDENTIALS", "value": "/etc/gcp_credentials.json", "_CLASS_": "olmo_core.launch.beaker.BeakerEnvVar"}, {"name": "NCCL_DEBUG", "value": "WARN", "_CLASS_": "olmo_core.launch.beaker.BeakerEnvVar"}, {"name": "TORCH_NCCL_TRACE_BUFFER_SIZE", "value": "0", "_CLASS_": "olmo_core.launch.beaker.BeakerEnvVar"}, {"name": "TORCH_NCCL_BLOCKING_WAIT", "value": "0", "_CLASS_": "olmo_core.launch.beaker.BeakerEnvVar"}], "env_secrets": [{"name": "BEAKER_TOKEN", "secret": "yawenzzzz_BEAKER_TOKEN", "_CLASS_": "olmo_core.launch.beaker.BeakerEnvSecret"}, {"name": "WANDB_API_KEY", "secret": "yawenzzzz_WANDB_API_KEY", "_CLASS_": "olmo_core.launch.beaker.BeakerEnvSecret"}, {"name": "GITHUB_TOKEN", "secret": "yawenzzzz_GITHUB_TOKEN", "_CLASS_": "olmo_core.launch.beaker.BeakerEnvSecret"}, {"name": "GCP_CREDENTIALS", "secret": "HELIOS_GCP_CREDENTIALS", "_CLASS_": "olmo_core.launch.beaker.BeakerEnvSecret"}], "nfs": false, "weka_buckets": [{"bucket": "dfive-default", "mount": "/weka/dfive-default", "_CLASS_": "olmo_core.launch.beaker.BeakerWekaBucket"}], "allow_dirty": false, "git": {"repo_url": "https://github.com/allenai/olmoearth_pretrain", "ref": "f9dd30abc09b0b8c3bfddc44005a6f79df100804", "branch": "henryh/rope-temporal-core", "_CLASS_": "olmo_core.launch.utils.GitConfig"}, "result_dir": "/results", "use_hostname_constraints": false, "_CLASS_": "olmoearth_pretrain.internal.experiment.OlmoEarthBeakerLaunchConfig"}, "visualize": {"output_dir": "/weka/dfive-default/olmoearth_pretrain/checkpoints/yawenzzzz/v1_2_nano/visualizations", "std_multiplier": 2.0, "_CLASS_": "olmoearth_pretrain.internal.experiment.OlmoEarthVisualizeConfig"}, "init_seed": 12536, "_CLASS_": "olmoearth_pretrain.internal.experiment.OlmoEarthExperimentConfig"}