Train/satbird pred

configs/data/satbird-usa.yaml

@@ -0,0 +1,20 @@
+_target_: src.data.base_datamodule.BaseDataModule
+
+dataset:
+  _target_: src.data.satbird_dataset.SatBirdDataset
+
+  data_dir: ${paths.data_dir}
+  modalities:
+  use_target_data: true
+  use_aux_data: none
+  seed: ${seed}
+  cache_dir: ${paths.cache_dir}
+  study_site: USA-summer
+
+batch_size: 16
+num_workers: 1
+pin_memory: false
+split_mode: "from_file"
+save_split: false
+saved_split_file_name: "split_indices_satbird-USA-summer.pth"
+seed: ${seed}

configs/experiment/satbird_sum_prediction/avr_aef_128_256_deeper_mlp.yaml

@@ -0,0 +1,58 @@
+# @package _global_
+
+# to execute this experiment run:
+# python train.py experiment=example
+
+# AEF with average encoder and MLP prediction head, for tile sizes 128 and 256, each with 3 random seeds
+
+defaults:
+  - override /model: s2bms_prediction
+  - override /data: satbird-usa
+  - override /metrics: satbird_predictive
+
+tags:
+  - prediction
+  - aef
+  - aef_${str:${data.dataset.modalities.aef_avr.size}}
+  - ${str:${seed}}
+
+experiment_name: avr_aef_${str:${data.dataset.modalities.aef_avr.size}}_deeper_mlp
+
+model:
+  geo_encoder:
+    _target_: src.models.components.geo_encoders.identity_encoder.IdentityEncoder
+    geo_data_name: aef_avr
+  trainable_modules: [prediction_head]
+  prediction_head:
+    _target_: src.models.components.pred_heads.mlp_pred_head.MLPPredictionHead
+    nn_layers: 3
+
+data:
+  dataset:
+    modalities:
+      aef_avr:
+        path: ${paths.data_dir}/satbird-USA-summer/eo/aef-satbird-usa-summer_average-${data.dataset.modalities.aef_avr.size}.csv
+        enable_nans: true
+    use_target_data: true
+    use_aux_data: false
+  caption_builder:
+  pin_memory: true
+  num_workers: 14
+  batch_size: 128
+  persistent_workers: true
+  saved_split_file_name: "satbird-USA-summer_aef_union_val_test.pth"
+
+logger:
+  wandb:
+    tags: ${tags}
+    group: "predictive"
+    project: "satbird-usa-summer_prediction"
+  aim:
+    experiment: "predictive"
+
+hydra:
+  mode: MULTIRUN
+  sweeper:
+    params:
+      seed: 12345, 404, 654
+      ++data.dataset.modalities.aef_avr.size: 128, 256

configs/experiment/satbird_sum_prediction/avr_aef_128_256_lin.yaml

@@ -0,0 +1,57 @@
+# @package _global_
+
+# to execute this experiment run:
+# python train.py experiment=example
+
+# AEF with average encoder and linear prediction head, for tile sizes 128 and 256, each with 3 random seeds
+
+defaults:
+  - override /model: s2bms_prediction
+  - override /data: satbird-usa
+  - override /metrics: satbird_predictive
+
+tags:
+  - prediction
+  - aef
+  - aef_${str:${data.dataset.modalities.aef_avr.size}}
+  - ${str:${seed}}
+
+experiment_name: avr_aef_${str:${data.dataset.modalities.aef_avr.size}}_linear
+
+model:
+  geo_encoder:
+    _target_: src.models.components.geo_encoders.identity_encoder.IdentityEncoder
+    geo_data_name: aef_avr
+  trainable_modules: [prediction_head]
+  prediction_head:
+    _target_: src.models.components.pred_heads.linear_pred_head.LinearPredictionHead
+
+data:
+  dataset:
+    modalities:
+      aef_avr:
+        path: ${paths.data_dir}/satbird-USA-summer/eo/aef-satbird-usa-summer_average-${data.dataset.modalities.aef_avr.size}.csv
+        enable_nans: true
+    use_target_data: true
+    use_aux_data: false
+  caption_builder:
+  pin_memory: true
+  num_workers: 14
+  batch_size: 128
+  persistent_workers: true
+  saved_split_file_name: "satbird-USA-summer_aef_union_val_test.pth"
+
+logger:
+  wandb:
+    tags: ${tags}
+    group: "predictive"
+    project: "satbird-usa-summer_prediction"
+  aim:
+    experiment: "predictive"
+
+hydra:
+  mode: MULTIRUN
+  sweeper:
+    params:
+      seed: 12345, 404, 654
+      ++data.dataset.modalities.aef_avr.size: 128, 256

configs/experiment/satbird_sum_prediction/avr_aef_128_256_mlp.yaml

@@ -0,0 +1,57 @@
+# @package _global_
+
+# to execute this experiment run:
+# python train.py experiment=example
+
+# AEF with average encoder and MLP prediction head, for tile sizes 128 and 256, each with 3 random seeds
+
+defaults:
+  - override /model: s2bms_prediction
+  - override /data: satbird-usa
+  - override /metrics: satbird_predictive
+
+tags:
+  - prediction
+  - aef
+  - aef_${str:${data.dataset.modalities.aef_avr.size}}
+  - ${str:${seed}}
+
+experiment_name: avr_aef_${str:${data.dataset.modalities.aef_avr.size}}_mlp
+
+model:
+  geo_encoder:
+    _target_: src.models.components.geo_encoders.identity_encoder.IdentityEncoder
+    geo_data_name: aef_avr
+  trainable_modules: [prediction_head]
+  prediction_head:
+    _target_: src.models.components.pred_heads.mlp_pred_head.MLPPredictionHead
+
+data:
+  dataset:
+    modalities:
+      aef_avr:
+        path: ${paths.data_dir}/satbird-USA-summer/eo/aef-satbird-usa-summer_average-${data.dataset.modalities.aef_avr.size}.csv
+        enable_nans: true
+    use_target_data: true
+    use_aux_data: false
+  caption_builder:
+  pin_memory: true
+  num_workers: 14
+  batch_size: 128
+  persistent_workers: true
+  saved_split_file_name: "satbird-USA-summer_aef_union_val_test.pth"
+
+logger:
+  wandb:
+    tags: ${tags}
+    group: "predictive"
+    project: "satbird-usa-summer_prediction"
+  aim:
+    experiment: "predictive"
+
+hydra:
+  mode: MULTIRUN
+  sweeper:
+    params:
+      seed: 12345, 404, 654
+      ++data.dataset.modalities.aef_avr.size: 128, 256

configs/experiment/satbird_sum_prediction/geoclip_deeper_mlp.yaml

@@ -0,0 +1,54 @@
+# @package _global_
+
+# to execute this experiment run:
+# python train.py experiment=example
+
+# Geoclip with MLP prediction head, with 3 random seeds
+
+defaults:
+  - override /model: s2bms_prediction
+  - override /data: satbird-usa
+  - override /metrics: satbird_predictive
+
+tags:
+  - prediction
+  - geoclip
+  - ${str:${seed}}
+
+experiment_name: geoclip_deeper_mlp
+
+model:
+  geo_encoder:
+    _target_: src.models.components.geo_encoders.geoclip.GeoClipCoordinateEncoder
+    geo_data_name: coords
+  trainable_modules: [prediction_head]
+  prediction_head:
+    _target_: src.models.components.pred_heads.mlp_pred_head.MLPPredictionHead
+    nn_layers: 3
+
+data:
+  dataset:
+    modalities:
+      coords:
+    use_target_data: true
+    use_aux_data: false
+  caption_builder:
+  pin_memory: true
+  num_workers: 14
+  batch_size: 128
+  persistent_workers: true
+  saved_split_file_name: "satbird-USA-summer_aef_union_val_test.pth"
+
+logger:
+  wandb:
+    tags: ${tags}
+    group: "predictive"
+    project: "satbird-usa-summer_prediction"
+  aim:
+    experiment: "predictive"
+
+hydra:
+  mode: MULTIRUN
+  sweeper:
+    params:
+      seed: 12345, 404, 654

configs/experiment/satbird_sum_prediction/geoclip_lin.yaml

@@ -0,0 +1,53 @@
+# @package _global_
+
+# to execute this experiment run:
+# python train.py experiment=example
+
+# Geoclip with linear prediction head, with 3 random seeds
+
+defaults:
+  - override /model: s2bms_prediction
+  - override /data: satbird-usa
+  - override /metrics: satbird_predictive
+
+tags:
+  - prediction
+  - geoclip
+  - ${str:${seed}}
+
+experiment_name: geoclip_linear
+
+model:
+  geo_encoder:
+    _target_: src.models.components.geo_encoders.geoclip.GeoClipCoordinateEncoder
+    geo_data_name: coords
+  trainable_modules: [prediction_head]
+  prediction_head:
+    _target_: src.models.components.pred_heads.linear_pred_head.LinearPredictionHead
+
+data:
+  dataset:
+    modalities:
+      coords:
+    use_target_data: true
+    use_aux_data: false
+  caption_builder:
+  pin_memory: true
+  num_workers: 14
+  batch_size: 128
+  persistent_workers: true
+  saved_split_file_name: "satbird-USA-summer_aef_union_val_test.pth"
+
+logger:
+  wandb:
+    tags: ${tags}
+    group: "predictive"
+    project: "satbird-usa-summer_prediction"
+  aim:
+    experiment: "predictive"
+
+hydra:
+  mode: MULTIRUN
+  sweeper:
+    params:
+      seed: 12345, 404, 654

configs/experiment/satbird_sum_prediction/geoclip_mlp.yaml

@@ -0,0 +1,53 @@
+# @package _global_
+
+# to execute this experiment run:
+# python train.py experiment=example
+
+# Geoclip with MLP prediction head, with 3 random seeds
+
+defaults:
+  - override /model: s2bms_prediction
+  - override /data: satbird-usa
+  - override /metrics: satbird_predictive
+
+tags:
+  - prediction
+  - geoclip
+  - ${str:${seed}}
+
+experiment_name: geoclip_mlp
+
+model:
+  geo_encoder:
+    _target_: src.models.components.geo_encoders.geoclip.GeoClipCoordinateEncoder
+    geo_data_name: coords
+  trainable_modules: [prediction_head]
+  prediction_head:
+    _target_: src.models.components.pred_heads.mlp_pred_head.MLPPredictionHead
+
+data:
+  dataset:
+    modalities:
+      coords:
+    use_target_data: true
+    use_aux_data: false
+  caption_builder:
+  pin_memory: true
+  num_workers: 14
+  batch_size: 128
+  persistent_workers: true
+  saved_split_file_name: "satbird-USA-summer_aef_union_val_test.pth"
+
+logger:
+  wandb:
+    tags: ${tags}
+    group: "predictive"
+    project: "satbird-usa-summer_prediction"
+  aim:
+    experiment: "predictive"
+
+hydra:
+  mode: MULTIRUN
+  sweeper:
+    params:
+      seed: 12345, 404, 654