Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+future_action_eval.png filter=lfs diff=lfs merge=lfs -text
+main_rollout_action_figure_final.png filter=lfs diff=lfs merge=lfs -text

benchmark.json

@@ -0,0 +1,63 @@
+{
+  "action_dim": 3,
+  "aux_dynamics_hidden_dim": null,
+  "batch_size": 1,
+  "checkpoint": "outputs/simvp_baseline/2026-05-25_08-44-43_full5ep_motionctxV2_grid2x4_fmScale3_dualWM10_delta05_MSE_dLong05_dSteer15_corr01_noDetach_conv1x1_ctrlSteerS030/checkpoints/best.pt",
+  "conditioning_dim": 32,
+  "conditioning_fusion": null,
+  "conditioning_injection": "single",
+  "conditioning_representation": "longitudinal_steer_speed",
+  "control_steer_input_scale": 0.3,
+  "device": "cuda",
+  "dual_conv1x1_fusion_stats": {
+    "bias_norm": 0.050001209756022555,
+    "enabled": true,
+    "param_count": 2080,
+    "simvp_half_weight_norm": 34.854761816360536,
+    "weight_norm": 35.060819933559266,
+    "wm_half_weight_norm": 3.7956123535264505,
+    "wm_to_simvp_weight_norm_ratio": 0.10889795700009121
+  },
+  "dual_fusion": "conv1x1",
+  "dual_wm_conditioning": "film",
+  "dual_wm_gated": true,
+  "dual_wm_hidden_dim": 512,
+  "dual_wm_num_layers": 3,
+  "dual_wm_scale": 1.0,
+  "future_action_detach_latents": false,
+  "future_action_future_motion_scale": 3.0,
+  "future_action_head_variant": "motion_context_v2",
+  "future_action_source": "wm",
+  "future_action_spatial_grid": [
+    2,
+    4
+  ],
+  "future_action_spatial_pooling": "grid",
+  "future_action_token_dim": 256,
+  "future_frames_per_second": 205.4795994525576,
+  "has_future_action_cls_heads": true,
+  "has_future_action_head": true,
+  "height": 320,
+  "iters": 200,
+  "latency_ms_per_batch": 38.93330540508032,
+  "model_variant": "av_wm_dual_bigwm",
+  "normalize_controls": true,
+  "parameters": 11452811,
+  "peak_vram_mb": 1243.41455078125,
+  "pred_future_actions_shape": [
+    1,
+    8,
+    2
+  ],
+  "samples_per_second": 25.6849499315697,
+  "simvp_conditioning": "none",
+  "simvp_conditioning_stage": "input",
+  "speed_dim": 1,
+  "speed_scale": 20.0,
+  "warmup": 20,
+  "width": 512,
+  "wm_latent_residual": false,
+  "wm_residual_gated": true,
+  "wm_residual_hidden_dim": 128,
+  "wm_residual_scale": 0.1
+}

best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b0f9b50f8b219e14d91fbf9756abef8c38baa8f3f97fc007afe99fe420b2b2b4
+size 137531672

config.json

@@ -0,0 +1,201 @@
+{
+  "benchmark": "outputs/simvp_baseline/2026-05-25_08-44-43_full5ep_motionctxV2_grid2x4_fmScale3_dualWM10_delta05_MSE_dLong05_dSteer15_corr01_noDetach_conv1x1_ctrlSteerS030/benchmark",
+  "checkpoints": "outputs/simvp_baseline/2026-05-25_08-44-43_full5ep_motionctxV2_grid2x4_fmScale3_dualWM10_delta05_MSE_dLong05_dSteer15_corr01_noDetach_conv1x1_ctrlSteerS030/checkpoints",
+  "rollout_grids": "outputs/simvp_baseline/2026-05-25_08-44-43_full5ep_motionctxV2_grid2x4_fmScale3_dualWM10_delta05_MSE_dLong05_dSteer15_corr01_noDetach_conv1x1_ctrlSteerS030/rollout_grids",
+  "run": {
+    "action_dim": 3,
+    "aux_dynamics_hidden_dim": 128,
+    "aux_dynamics_loss": false,
+    "aux_dynamics_loss_type": "smooth_l1",
+    "aux_dynamics_target": "steer_longitudinal_speed",
+    "aux_dynamics_weight": 0.05,
+    "balanced_sampling": false,
+    "batch_size": 4,
+    "conditioning_dim": 32,
+    "conditioning_fusion": null,
+    "conditioning_injection": "single",
+    "conditioning_representation": "longitudinal_steer_speed",
+    "control_steer_input_scale": 0.3,
+    "data_root": "/home/aws/work/data/mile_action_diverse/train",
+    "debug_activation_stats": false,
+    "debug_dual_gate": false,
+    "device": "cuda",
+    "drop_path": 0.0,
+    "dual_align_direction": "simvp_to_wm",
+    "dual_align_loss_type": "smooth_l1",
+    "dual_align_loss_weight": 0.0,
+    "dual_detach_simvp_target": true,
+    "dual_fusion": "conv1x1",
+    "dual_simvp_image_loss_weight": 0.0,
+    "dual_wm_conditioning": "film",
+    "dual_wm_gated": true,
+    "dual_wm_hidden_dim": 512,
+    "dual_wm_image_loss_weight": 0.0,
+    "dual_wm_num_layers": 3,
+    "dual_wm_scale": 1.0,
+    "epochs": 5,
+    "eval_every_steps": 5000,
+    "full_train_samples": 93427,
+    "full_val_samples": 13452,
+    "future_action_cls_loss": true,
+    "future_action_cls_weight": 0.1,
+    "future_action_corr_loss_weight": 0.1,
+    "future_action_delta_longitudinal_weight": 0.5,
+    "future_action_delta_loss": true,
+    "future_action_delta_loss_type": "mse",
+    "future_action_delta_loss_weight": 0.5,
+    "future_action_delta_steer_weight": 1.5,
+    "future_action_detach_latents": false,
+    "future_action_dropout": 0.0,
+    "future_action_future_motion_scale": 3.0,
+    "future_action_head_variant": "motion_context_v2",
+    "future_action_hidden_dim": 256,
+    "future_action_longitudinal_cls_weight": 1.0,
+    "future_action_loss": true,
+    "future_action_loss_type": "mse",
+    "future_action_loss_weight": 1.0,
+    "future_action_num_layers": 1,
+    "future_action_source": "wm",
+    "future_action_spatial_grid": "2x4",
+    "future_action_spatial_pooling": "grid",
+    "future_action_steer_cls_weight": 1.0,
+    "future_action_target": "longitudinal_steer",
+    "future_steer_target_scale": 0.3,
+    "grad_clip_norm": 1.0,
+    "height": 320,
+    "hid_s": 32,
+    "hid_t": 256,
+    "internal_split_overlap": null,
+    "log_grad_norms": true,
+    "log_interval": 20,
+    "longitudinal_coast_threshold": 0.05,
+    "lr": 0.001,
+    "maneuver_metadata": null,
+    "maneuver_speed_metadata": "outputs/maneuver_metadata/maneuver_metadata.csv",
+    "maneuver_speed_weights": "straight_accel=0.5,straight_const=0.15,straight_decel=0.8,mild_turn_accel=2,mild_turn_const=1.0,mild_turn_decel=2,sharp_turn_accel=4,sharp_turn_const=3,sharp_turn_decel=4",
+    "maneuver_weights": "straight=1,mild_turn=1,sharp_turn=1",
+    "max_interim_val_batches": 50,
+    "max_train_samples": null,
+    "max_train_steps": null,
+    "max_val_batches": null,
+    "max_val_samples": null,
+    "model_type": "gSTA",
+    "model_variant": "av_wm_dual_bigwm",
+    "n_s": 4,
+    "n_t": 4,
+    "no_amp": true,
+    "no_progress_bar": false,
+    "normalize_controls": true,
+    "num_workers": 12,
+    "output_dir": "outputs/simvp_baseline",
+    "resume": null,
+    "run_tag": "full5ep_motionctxV2_grid2x4_fmScale3_dualWM10_delta05_MSE_dLong05_dSteer15_corr01_noDetach_conv1x1_ctrlSteerS030",
+    "sampling_strategy": "maneuver_speed_balanced",
+    "seed": 42,
+    "selected_train_towns": [
+      "Town01",
+      "Town03",
+      "Town04"
+    ],
+    "selected_val_towns": [
+      "Town02"
+    ],
+    "simvp_conditioning": "none",
+    "simvp_conditioning_stage": "input",
+    "speed_delta_threshold": 0.3,
+    "speed_dim": 1,
+    "speed_scale": 20.0,
+    "split_seed": 42,
+    "ssim_loss_weight": 0.0,
+    "steer_straight_threshold": 0.03,
+    "stop_on_nan": true,
+    "subset_strategy": "first",
+    "test_data_root": null,
+    "train_data_root": "/home/aws/work/data/mile_action_diverse/train",
+    "train_samples": 93427,
+    "train_towns": [
+      "Town01",
+      "Town03",
+      "Town04"
+    ],
+    "val_data_root": "/home/aws/work/data/mile_action_diverse/val",
+    "val_fraction": 0.1,
+    "val_maneuver_metadata": null,
+    "val_maneuver_speed_metadata": null,
+    "val_samples": 13452,
+    "val_subset_strategy": "first",
+    "val_towns": [
+      "Town02"
+    ],
+    "validation_mode": "explicit_val_root",
+    "weight_decay": 0.0,
+    "weights_path": null,
+    "width": 512,
+    "wm_latent_residual": false,
+    "wm_residual_gated": true,
+    "wm_residual_hidden_dim": 128,
+    "wm_residual_loss": false,
+    "wm_residual_loss_type": "smooth_l1",
+    "wm_residual_loss_weight": 0.05,
+    "wm_residual_scale": 0.1
+  },
+  "run_dir": "outputs/simvp_baseline/2026-05-25_08-44-43_full5ep_motionctxV2_grid2x4_fmScale3_dualWM10_delta05_MSE_dLong05_dSteer15_corr01_noDetach_conv1x1_ctrlSteerS030",
+  "trainer": {
+    "amp": false,
+    "aux_dynamics_loss": false,
+    "aux_dynamics_loss_type": "smooth_l1",
+    "aux_dynamics_weight": 0.05,
+    "batch_size": 4,
+    "control_steer_input_scale": 0.3,
+    "debug_activation_stats": false,
+    "debug_dual_gate": false,
+    "device": "cuda",
+    "dual_align_direction": "simvp_to_wm",
+    "dual_align_loss_type": "smooth_l1",
+    "dual_align_loss_weight": 0.0,
+    "dual_detach_simvp_target": true,
+    "dual_simvp_image_loss_weight": 0.0,
+    "dual_wm_image_loss_weight": 0.0,
+    "epochs": 5,
+    "eval_every_steps": 5000,
+    "future_action_cls_loss": true,
+    "future_action_cls_weight": 0.1,
+    "future_action_corr_loss_weight": 0.1,
+    "future_action_delta_longitudinal_weight": 0.5,
+    "future_action_delta_loss": true,
+    "future_action_delta_loss_type": "mse",
+    "future_action_delta_loss_weight": 0.5,
+    "future_action_delta_steer_weight": 1.5,
+    "future_action_detach_latents": false,
+    "future_action_future_motion_scale": 3.0,
+    "future_action_head_variant": "motion_context_v2",
+    "future_action_longitudinal_cls_weight": 1.0,
+    "future_action_loss": true,
+    "future_action_loss_type": "mse",
+    "future_action_loss_weight": 1.0,
+    "future_action_source": "wm",
+    "future_action_steer_cls_weight": 1.0,
+    "future_steer_target_scale": 0.3,
+    "grad_clip_norm": 1.0,
+    "learning_rate": 0.001,
+    "log_grad_norms": true,
+    "log_interval": 20,
+    "longitudinal_coast_threshold": 0.05,
+    "max_interim_val_batches": 50,
+    "max_train_steps": null,
+    "max_val_batches": null,
+    "num_workers": 12,
+    "output_dir": "outputs/simvp_baseline",
+    "progress_bar": true,
+    "resume_checkpoint": null,
+    "run_tag": "full5ep_motionctxV2_grid2x4_fmScale3_dualWM15_delta05_MSE_dLong05_dSteer15_corr01_noDetach_conv1x1_ctrlSteerS030",
+    "save_every": 1,
+    "ssim_loss_weight": 0.0,
+    "steer_straight_threshold": 0.03,
+    "stop_on_nan": true,
+    "weight_decay": 0.0,
+    "wm_residual_loss": false,
+    "wm_residual_loss_type": "smooth_l1",
+    "wm_residual_loss_weight": 0.05
+  }
+}