Iteration 0

checkpoints/checkpoint.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2082569776ff661092a8f569ac547bda561a052c9f09b7b885d52a9a85d75deb
+size 127488502

checkpoints/model_index.json

@@ -0,0 +1,12 @@
+{
+  "_class_name": "DDPMPipeline",
+  "_diffusers_version": "0.27.2",
+  "scheduler": [
+    "diffusers",
+    "DDPMScheduler"
+  ],
+  "unet": [
+    "diffusers",
+    "UNet2DModel"
+  ]
+}

checkpoints/scheduler/scheduler_config.json

@@ -0,0 +1,19 @@
+{
+  "_class_name": "DDPMScheduler",
+  "_diffusers_version": "0.27.2",
+  "beta_end": 0.02,
+  "beta_schedule": "linear",
+  "beta_start": 0.0001,
+  "clip_sample": true,
+  "clip_sample_range": 1.0,
+  "dynamic_thresholding_ratio": 0.995,
+  "num_train_timesteps": 200,
+  "prediction_type": "epsilon",
+  "rescale_betas_zero_snr": false,
+  "sample_max_value": 1.0,
+  "steps_offset": 0,
+  "thresholding": false,
+  "timestep_spacing": "leading",
+  "trained_betas": null,
+  "variance_type": "fixed_large"
+}

checkpoints/unet/config.json

@@ -0,0 +1,45 @@
+{
+  "_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": [
+    32,
+    64,
+    128,
+    256
+  ],
+  "center_input_sample": false,
+  "class_embed_type": null,
+  "down_block_types": [
+    "DownBlock2D",
+    "AttnDownBlock2D",
+    "AttnDownBlock2D",
+    "AttnDownBlock2D"
+  ],
+  "downsample_padding": 1,
+  "downsample_type": "conv",
+  "dropout": 0.0,
+  "flip_sin_to_cos": true,
+  "freq_shift": 0,
+  "in_channels": 1,
+  "layers_per_block": 2,
+  "mid_block_scale_factor": 1,
+  "norm_eps": 1e-05,
+  "norm_num_groups": 32,
+  "num_class_embeds": null,
+  "num_train_timesteps": null,
+  "out_channels": 1,
+  "resnet_time_scale_shift": "default",
+  "sample_size": 32,
+  "time_embedding_type": "positional",
+  "up_block_types": [
+    "AttnUpBlock2D",
+    "AttnUpBlock2D",
+    "AttnUpBlock2D",
+    "UpBlock2D"
+  ],
+  "upsample_type": "conv"
+}

checkpoints/unet/diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b22a513831ebbf2eabf97c8d7a60a24d8d7564eb257e97a748d2ef0a1d099860
+size 63608284

run_args.json

@@ -0,0 +1 @@
+{"model": "UNet", "dataset": "mnist", "save_folder": "/network/scratch/l/luca.scimeca/gfn_generated_results/train_prior_ddpm-mnist-32_0/", "load_path": "/home/mila/l/luca.scimeca/projects/InvDiff/src/models/pretrained", "data_path": "/network/scratch/l/luca.scimeca/data", "prior_training_mode": "dpm", "pretrain_models": false, "replace": true, "baseline_reinforce": "whitening", "show_figures": false, "save_figures": true, "save_model_weights": false, "plot_batch_size": 30, "batched_rtb_size": 64, "rtb_batched_train": false, "exp_name": "train_prior_ddpm-mnist-32", "method": "rtb", "inv_task": "phase_retrieval", "reward_weight": 10, "kl_weight": 0.01, "particles": 10, "scale": 1.0, "channels": 1, "workers": 8, "epochs": 250000, "algo": "mle", "batch_size": 32, "test_batch_size": 50, "lr": 0.0001, "lr_logZ": 0.1, "z_weight_decay": 0, "energy_temperature": 1.0, "traj_length": 200, "sampling_length": 50, "learning_cutoff": 0.1, "learn_var": false, "loss_type": "l2", "accumulate_gradient_every": 1, "detach_freq": 0.0, "detach_cut_off": 1.0, "back_and_forth": false, "bf_length": 50, "drift_only": false, "drift_burn_in": 1000, "mixed_precision": false, "checkpointing": true, "multi_class_index": 0, "finetune_class": [0], "class_name": "0", "classifier_model": "cnn", "classifier_depth": 10, "classifier_pretrained": false, "inverse_sample_number": 0, "langevin": false, "lgv_clip": 100.0, "lgv_clipping": false, "lgv_t_dim": 64, "lgv_hidden_dim": 256, "lgv_num_layers": 3, "lgv_zero_init": false, "compute_fid": false, "num_fid_samples": 50000, "lora": true, "rank": 32, "push_to_hf": 1, "push_to_wandb": true, "notes": "", "seed": 912, "model_config": null, "diffusion_config": null, "task_config": null, "gpu": 0, "c_rate": 0.95, "particle_size": 1, "use_cuda": true, "device": "cuda", "system": "linux", "supported_datasets": ["mnist", "cifar10", "CelebA", "dsprites", "UTKFace", "domainnet-quickdraw", "ffhq"], "splits": ["train", "test"], "image_size": 32, "noise_size": 32, "x_tensor": "images", "y_tensor": "labels", "num_classes": 10, "block_out_channels": [32, 64, 128, 256], "fid_inception_block": 768}