Upload folder using huggingface_hub

brca/config.yaml

@@ -0,0 +1,99 @@
+model:
+  base_learning_rate: 2.0e-05
+  target: ldm.models.diffusion.ddpm.LatentDiffusion
+  params:
+    linear_start: 0.0015
+    linear_end: 0.0195
+    num_timesteps_cond: 1
+    log_every_t: 200
+    timesteps: 1000
+    first_stage_key: image
+    cond_stage_key: hybrid
+    image_size: 64
+    channels: 3
+    cond_stage_trainable: true
+    conditioning_key: crossattn
+    use_ema: true
+    x_feat_extracted: true
+    scheduler_config:
+      target: ldm.lr_scheduler.LambdaLinearScheduler
+      params:
+        warm_up_steps:
+        - 1000
+        cycle_lengths:
+        - 10000000000000
+        f_start:
+        - 1.0e-06
+        f_max:
+        - 1.0
+        f_min:
+        - 1.0
+    unet_config:
+      target: ldm.modules.diffusionmodules.openaimodel.UNetModel
+      params:
+        use_checkpoint: true
+        use_fp16: true
+        image_size: 64
+        in_channels: 3
+        out_channels: 3
+        model_channels: 192
+        attention_resolutions:
+        - 8
+        - 4
+        - 2
+        num_res_blocks: 2
+        channel_mult:
+        - 1
+        - 2
+        - 3
+        - 5
+        num_heads: 1
+        use_spatial_transformer: true
+        transformer_depth: 1
+        context_dim: 512
+        ckpt_path: models/ldm_imgnet_unet.ckpt
+    first_stage_config:
+      target: ldm.models.autoencoder.VQModelInterface
+      params:
+        # ckpt_path: models/vq_f4.ckpt
+        embed_dim: 3
+        n_embed: 8192
+        ddconfig:
+          double_z: false
+          z_channels: 3
+          resolution: 256
+          in_channels: 3
+          out_ch: 3
+          ch: 128
+          ch_mult:
+          - 1
+          - 2
+          - 4
+          num_res_blocks: 2
+          attn_resolutions: []
+          dropout: 0.0
+        lossconfig:
+          target: torch.nn.Identity
+    cond_stage_config:
+      target: ldm.modules.encoders.modules.EmbeddingViT2_5
+      params:
+        feat_key: ssl_feat
+        mag_key: mag
+        num_layers: 12
+        input_channels: 1024
+        hidden_channels: 512
+        vit_mlp_dim: 2048
+        p_uncond: 0.1
+# data:
+#   target: main.DataModuleFromConfig
+#   params:
+#     batch_size: 420
+#     num_workers: 8
+#     wrap: false
+#     train:
+#       target: ldm.data.brca.TCGADataset
+#       params:
+#         config:
+#           root: ""
+#           normalize_ssl: true
+#           feat_target_size: 8

brca/weights.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4c5f1d5daa0016d86f7a2d2d6640b637bdb808ee61da201cc19d9d00bef72869
+size 3584016692

naip/config.yaml

@@ -0,0 +1,119 @@
+
+# same architecture as pathology ldm, but for NAIP
+
+model:
+  base_learning_rate: 2e-5
+  target: ldm.models.diffusion.ddpm.LatentDiffusion
+  params:
+    linear_start: 0.0015
+    linear_end: 0.0195
+    num_timesteps_cond: 1
+    log_every_t: 200
+    timesteps: 1000
+    first_stage_key: image
+    cond_stage_key: hybrid
+    image_size: 64
+    channels: 3
+    cond_stage_trainable: true 
+    conditioning_key: crossattn
+    use_ema: True
+    x_feat_extracted: True
+
+    scheduler_config: # 1000 warmup steps
+      target: ldm.lr_scheduler.LambdaLinearScheduler
+      params:
+        warm_up_steps: [ 1000 ]
+        cycle_lengths: [ 10000000000000 ] # incredibly large number to prevent corner cases
+        f_start: [ 1.e-6 ]
+        f_max: [ 1. ]
+        f_min: [ 1. ]
+
+    unet_config:
+      target: ldm.modules.diffusionmodules.openaimodel.UNetModel
+      params:
+        use_checkpoint: True
+        use_fp16: True
+        image_size: 64
+        in_channels: 3
+        out_channels: 3
+        model_channels: 192
+        attention_resolutions: [8, 4, 2]
+        num_res_blocks: 2
+        channel_mult: [ 1,2,3,5 ]
+        num_heads: 1
+        use_spatial_transformer: True
+        transformer_depth: 1
+        context_dim: 512
+        ckpt_path: "models/ldm_imgnet_unet.ckpt"
+
+    first_stage_config:
+      target: ldm.models.autoencoder.VQModelInterface
+      params:
+        # ckpt_path: "models/vq_f4.ckpt"
+        embed_dim: 3
+        n_embed: 8192
+        ddconfig:
+          double_z: false
+          z_channels: 3
+          resolution: 256
+          in_channels: 3
+          out_ch: 3
+          ch: 128
+          ch_mult:
+          - 1
+          - 2
+          - 4
+          num_res_blocks: 2
+          attn_resolutions: []
+          dropout: 0.0
+        lossconfig:
+          target: torch.nn.Identity
+
+
+    cond_stage_config:
+      target: ldm.modules.encoders.modules.EmbeddingViT2_5
+      params:
+        feat_key: "ssl_feat"
+        mag_key: "mag"
+        num_layers: 12
+        input_channels: 1024
+        hidden_channels: 512
+        vit_mlp_dim: 2048
+        p_uncond: 0.1
+        mag_levels: 4
+        
+
+data:
+  target: main.DataModuleFromConfig
+  params:
+    batch_size: 400
+    num_workers: 16
+    wrap: false
+    train:
+      target: ldm.data.naip.NAIPDataset
+      params:
+        config:
+          root: ""
+          normalize_ssl: True
+          feat_target_size: 8
+
+
+lightning:
+  callbacks:
+    metrics_over_trainsteps_checkpoint:
+      params:
+        every_n_train_steps: 20000
+    image_logger:
+      target: main.ImageLogger
+      params:
+        disabled: True
+        batch_frequency: 500000
+        max_images: 8
+        increase_log_steps: False
+        log_images_kwargs:
+          quantize_denoised: False
+          inpaint: False
+          
+  trainer:
+    benchmark: True
+    precision: 16

naip/weights.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4c5f1d5daa0016d86f7a2d2d6640b637bdb808ee61da201cc19d9d00bef72869
+size 3584016692