models batch 1

lidm_dynamic/config.json

@@ -0,0 +1,52 @@
+{
+  "_class_name": "UNet2DModel",
+  "_diffusers_version": "0.27.2",
+  "act_fn": "silu",
+  "add_attention": true,
+  "attention_head_dim": 8,
+  "attn_norm_num_groups": null,
+  "block_out_channels": [
+    128,
+    256,
+    256,
+    512
+  ],
+  "center_input_sample": false,
+  "class_embed_type": null,
+  "decay": 0.9999,
+  "down_block_types": [
+    "AttnDownBlock2D",
+    "AttnDownBlock2D",
+    "AttnDownBlock2D",
+    "DownBlock2D"
+  ],
+  "downsample_padding": 1,
+  "downsample_type": "resnet",
+  "dropout": 0.0,
+  "flip_sin_to_cos": true,
+  "freq_shift": 0,
+  "in_channels": 4,
+  "inv_gamma": 1.0,
+  "layers_per_block": 2,
+  "mid_block_scale_factor": 1,
+  "min_decay": 0.0,
+  "norm_eps": 1e-05,
+  "norm_num_groups": 32,
+  "num_class_embeds": null,
+  "num_train_timesteps": null,
+  "optimization_step": 100,
+  "out_channels": 4,
+  "power": 0.6666666666666666,
+  "resnet_time_scale_shift": "default",
+  "sample_size": 14,
+  "time_embedding_type": "positional",
+  "up_block_types": [
+    "UpBlock2D",
+    "AttnUpBlock2D",
+    "AttnUpBlock2D",
+    "AttnUpBlock2D"
+  ],
+  "update_after_step": 0,
+  "upsample_type": "resnet",
+  "use_ema_warmup": false
+}

lidm_dynamic/config.yaml

@@ -0,0 +1,105 @@
+wandb_group: lidm
+output_dir: experiments/lidm_dynamic
+pretrained_model_name_or_path: null
+vae_path: models/vae
+globals:
+  target_fps: 32
+  target_nframes: 64
+  outputs:
+  - image
+datasets:
+- name: Latent
+  active: true
+  params:
+    root: data/latents/dynamic
+    target_fps: ${globals.target_fps}
+    target_nframes: ${globals.target_nframes}
+    target_resolution: 14
+    outputs: ${globals.outputs}
+unet:
+  _class_name: UNet2DModel
+  sample_size: 14
+  in_channels: 4
+  out_channels: 4
+  center_input_sample: false
+  time_embedding_type: positional
+  freq_shift: 0
+  flip_sin_to_cos: true
+  down_block_types:
+  - AttnDownBlock2D
+  - AttnDownBlock2D
+  - AttnDownBlock2D
+  - DownBlock2D
+  up_block_types:
+  - UpBlock2D
+  - AttnUpBlock2D
+  - AttnUpBlock2D
+  - AttnUpBlock2D
+  block_out_channels:
+  - 128
+  - 256
+  - 256
+  - 512
+  layers_per_block: 2
+  mid_block_scale_factor: 1
+  downsample_padding: 1
+  downsample_type: resnet
+  upsample_type: resnet
+  dropout: 0.0
+  act_fn: silu
+  attention_head_dim: 8
+  norm_num_groups: 32
+  attn_norm_num_groups: null
+  norm_eps: 1.0e-05
+  resnet_time_scale_shift: default
+  class_embed_type: null
+  num_class_embeds: null
+noise_scheduler:
+  _class_name: DDPMScheduler
+  num_train_timesteps: 1000
+  beta_start: 0.0001
+  beta_end: 0.02
+  beta_schedule: linear
+  variance_type: fixed_small
+  clip_sample: true
+  clip_sample_range: 4.0
+  prediction_type: v_prediction
+  thresholding: false
+  dynamic_thresholding_ratio: 0.995
+  sample_max_value: 1.0
+  timestep_spacing: leading
+  steps_offset: 0
+train_batch_size: 256
+dataloader_num_workers: 16
+max_train_steps: 500000
+learning_rate: 0.0003
+lr_warmup_steps: 500
+scale_lr: false
+lr_scheduler: constant
+use_8bit_adam: false
+gradient_accumulation_steps: 1
+noise_offset: 0.0
+gradient_checkpointing: false
+use_ema: true
+enable_xformers_memory_efficient_attention: false
+allow_tf32: true
+adam_beta1: 0.9
+adam_beta2: 0.999
+adam_weight_decay: 0.01
+adam_epsilon: 1.0e-08
+max_grad_norm: 1.0
+logging_dir: logs
+mixed_precision: fp16
+validation_timesteps: 128
+validation_fps: ${globals.target_fps}
+validation_frames: ${globals.target_nframes}
+validation_count: 4
+validation_guidance: 1.0
+validation_steps: 2500
+report_to: wandb
+checkpointing_steps: 100
+checkpoints_total_limit: 100
+resume_from_checkpoint: null
+tracker_project_name: echosyn
+seed: 42
+num_train_epochs: 250000

lidm_dynamic/diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c337bfad7278d2507f10c9bb3d61055ed15032f87d4cb9951abfae8ed3fa87e
+size 294245640

lvdm/config.json

@@ -0,0 +1,44 @@
+{
+  "_class_name": "UNetSpatioTemporalConditionModel",
+  "_diffusers_version": "0.27.2",
+  "addition_time_embed_dim": 1,
+  "block_out_channels": [
+    128,
+    256,
+    256,
+    512
+  ],
+  "cross_attention_dim": 1,
+  "decay": 0.9999,
+  "down_block_types": [
+    "CrossAttnDownBlockSpatioTemporal",
+    "CrossAttnDownBlockSpatioTemporal",
+    "CrossAttnDownBlockSpatioTemporal",
+    "DownBlockSpatioTemporal"
+  ],
+  "in_channels": 8,
+  "inv_gamma": 1.0,
+  "layers_per_block": 2,
+  "min_decay": 0.0,
+  "num_attention_heads": [
+    8,
+    16,
+    16,
+    32
+  ],
+  "num_frames": 64,
+  "optimization_step": 60000,
+  "out_channels": 4,
+  "power": 0.6666666666666666,
+  "projection_class_embeddings_input_dim": 1,
+  "sample_size": 14,
+  "transformer_layers_per_block": 1,
+  "up_block_types": [
+    "UpBlockSpatioTemporal",
+    "CrossAttnUpBlockSpatioTemporal",
+    "CrossAttnUpBlockSpatioTemporal",
+    "CrossAttnUpBlockSpatioTemporal"
+  ],
+  "update_after_step": 0,
+  "use_ema_warmup": false
+}

lvdm/config.yaml

@@ -0,0 +1,121 @@
+wandb_group: lvdm
+output_dir: experiments/lvdm
+pretrained_model_name_or_path: null
+vae_path: models/vae
+globals:
+  target_fps: 32
+  target_nframes: 64
+  outputs:
+  - video
+  - lvef
+  - image
+datasets:
+- name: Latent
+  active: true
+  params:
+    root: data/latents/dynamic
+    target_fps: ${globals.target_fps}
+    target_nframes: ${globals.target_nframes}
+    target_resolution: 14
+    outputs: ${globals.outputs}
+- name: Latent
+  active: true
+  params:
+    root: data/latents/ped_a4c
+    target_fps: ${globals.target_fps}
+    target_nframes: ${globals.target_nframes}
+    target_resolution: 14
+    outputs: ${globals.outputs}
+- name: Latent
+  active: true
+  params:
+    root: data/latents/ped_psax
+    target_fps: ${globals.target_fps}
+    target_nframes: ${globals.target_nframes}
+    target_resolution: 14
+    outputs: ${globals.outputs}
+unet:
+  _class_name: UNetSpatioTemporalConditionModel
+  addition_time_embed_dim: 1
+  block_out_channels:
+  - 128
+  - 256
+  - 256
+  - 512
+  cross_attention_dim: 1
+  down_block_types:
+  - CrossAttnDownBlockSpatioTemporal
+  - CrossAttnDownBlockSpatioTemporal
+  - CrossAttnDownBlockSpatioTemporal
+  - DownBlockSpatioTemporal
+  in_channels: 8
+  layers_per_block: 2
+  num_attention_heads:
+  - 8
+  - 16
+  - 16
+  - 32
+  num_frames: ${globals.target_nframes}
+  out_channels: 4
+  projection_class_embeddings_input_dim: 1
+  sample_size: 14
+  transformer_layers_per_block: 1
+  up_block_types:
+  - UpBlockSpatioTemporal
+  - CrossAttnUpBlockSpatioTemporal
+  - CrossAttnUpBlockSpatioTemporal
+  - CrossAttnUpBlockSpatioTemporal
+noise_scheduler:
+  _class_name: DDPMScheduler
+  num_train_timesteps: 1000
+  beta_start: 0.0001
+  beta_end: 0.02
+  beta_schedule: linear
+  variance_type: fixed_small
+  clip_sample: true
+  clip_sample_range: 4.0
+  prediction_type: v_prediction
+  thresholding: false
+  dynamic_thresholding_ratio: 0.995
+  sample_max_value: 1.0
+  timestep_spacing: leading
+  steps_offset: 0
+train_batch_size: 16
+dataloader_num_workers: 16
+max_train_steps: 500000
+learning_rate: 0.0001
+lr_warmup_steps: 500
+scale_lr: false
+lr_scheduler: constant
+use_8bit_adam: false
+gradient_accumulation_steps: 1
+noise_offset: 0.1
+drop_conditionning: 0.1
+gradient_checkpointing: false
+use_ema: true
+enable_xformers_memory_efficient_attention: false
+allow_tf32: true
+adam_beta1: 0.9
+adam_beta2: 0.999
+adam_weight_decay: 0.01
+adam_epsilon: 1.0e-08
+max_grad_norm: 1.0
+logging_dir: logs
+mixed_precision: fp16
+validation_timesteps: 128
+validation_fps: ${globals.target_fps}
+validation_frames: ${globals.target_nframes}
+validation_lvefs:
+- 0.0
+- 0.4
+- 0.7
+- 1.0
+validation_guidance: 1.0
+validation_steps: 1500
+report_to: wandb
+checkpointing_steps: 10000
+checkpoints_total_limit: 100
+resume_from_checkpoint: null
+tracker_project_name: echosyn
+seed: 42
+num_train_epochs: 893

lvdm/diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db1526dbff74ea98cdac06ba63c1d18294d34775d87511d5323ce6e4cc7c1bf4
+size 575506960

regression_dynamic/best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5cb136834dde9cac1bdbc7904c49572c0d2df356e20750205bb7225b6fcfb271
+size 250601638

reidentification_dynamic/config.json

@@ -0,0 +1,24 @@
+{
+	"experiment_description": "reidentification_dynamic",
+	"resumption": false,
+	"resumption_count": null,
+	"previous_experiment": null,
+	"image_path": "data/latents/dynamic/FileList.csv",
+	"siamese_architecture": "ResNet-50",
+	"data_handling": "balanced",
+	"num_workers": 16,
+	"pin_memory": true,
+	"n_channels": 4,
+	"n_features": 128,
+	"image_size": 14,
+	"loss": "BCEWithLogitsLoss",
+	"optimizer": "Adam",
+	"learning_rate": 0.0001,
+	"batch_size": 128,
+	"max_epochs": 1000,
+	"early_stopping": 50,
+	"transform": "pmone",
+	"n_samples_train": 7465,
+	"n_samples_val": 1288,
+	"n_samples_test": 1277
+}

reidentification_dynamic/reidentification_dynamic_best_network.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f462dc5b237d1b2f5a42c35297c6d72b77d07ac3a12b44b1058a204cf95cacb
+size 95425186

vae/config.json

@@ -0,0 +1,31 @@
+{
+  "_class_name": "AutoencoderKL",
+  "_diffusers_version": "0.23.1",
+  "act_fn": "silu",
+  "block_out_channels": [
+    128,
+    256,
+    256,
+    512
+  ],
+  "down_block_types": [
+    "DownEncoderBlock2D",
+    "DownEncoderBlock2D",
+    "DownEncoderBlock2D",
+    "DownEncoderBlock2D"
+  ],
+  "force_upcast": true,
+  "in_channels": 3,
+  "latent_channels": 4,
+  "layers_per_block": 2,
+  "norm_num_groups": 32,
+  "out_channels": 3,
+  "sample_size": 512,
+  "scaling_factor": 0.18215,
+  "up_block_types": [
+    "UpDecoderBlock2D",
+    "UpDecoderBlock2D",
+    "UpDecoderBlock2D",
+    "UpDecoderBlock2D"
+  ]
+}

vae/diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba0be1555511d1e145bfda156062aab744c6f7fc12e930c78c3640baf8183d5b
+size 249675844