diff --git a/.placeholder b/.placeholder new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/ablation/baseline_sd21_train_genpercept_exr_1card_ensure_bs16_per_accu/unet/config.json b/ablation/baseline_sd21_train_genpercept_exr_1card_ensure_bs16_per_accu/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/baseline_sd21_train_genpercept_exr_1card_ensure_bs16_per_accu/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/baseline_sd21_train_genpercept_exr_1card_ensure_bs16_per_accu/unet/diffusion_pytorch_model.safetensors b/ablation/baseline_sd21_train_genpercept_exr_1card_ensure_bs16_per_accu/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..9880a680170aebbd6f4ea3b2ca47cb6e79bc20f2 --- /dev/null +++ b/ablation/baseline_sd21_train_genpercept_exr_1card_ensure_bs16_per_accu/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f00642f2a10c6fdfac3f6e7bb1c4bef9cc2b32e0ccb4672f51c9d0a62fa1a058 +size 3463726472 diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point0002125_0point003/unet/config.json b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point0002125_0point003/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point0002125_0point003/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point0002125_0point003/unet/diffusion_pytorch_model.safetensors b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point0002125_0point003/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..c4ffafc019de44c5e3fa965f15c9965091d32770 --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point0002125_0point003/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:031dd03142dac0ecfcac66bd54ef52d502904c9a40c5812dd62c0a082ff0392a +size 3463726472 diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point000425_0point006/unet/config.json b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point000425_0point006/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point000425_0point006/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point000425_0point006/unet/diffusion_pytorch_model.safetensors b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point000425_0point006/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..00e72c4bb14005832f63a96ca256cdf0d082f3f1 --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point000425_0point006/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6c64d2f8e438ac0bc149c964e34c929ea0241c882111b8b123ea59f7d67752bc +size 3463726472 diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point00085_0point012/unet/config.json b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point00085_0point012/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point00085_0point012/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point00085_0point012/unet/diffusion_pytorch_model.safetensors b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point00085_0point012/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..6b7597288002066d3e0bae6d51a67988fba1994e --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point00085_0point012/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2b19facea4bd341f10ee01c1da6a8a44e0b8fd70737f88af8c53e463b7817425 +size 3463726472 diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point00340_0point048/unet/config.json b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point00340_0point048/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point00340_0point048/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point00340_0point048/unet/diffusion_pytorch_model.safetensors b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point00340_0point048/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..f7192a3d8b86813d24970f837403fca016637f8d --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point00340_0point048/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c752e4e64339cf11a1beba59b0c73871a2d1342549ef5551e83c7fb0a2b54e22 +size 3463726472 diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point05440_0point768/unet/config.json b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point05440_0point768/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point05440_0point768/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point05440_0point768/unet/diffusion_pytorch_model.safetensors b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point05440_0point768/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..40d6b40c96fc6554db03bb3d0b78ff2ff33f3393 --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point05440_0point768/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5e6040642c4a224206c9d32d3c0134ecdb17210aae925e53366807533e7c04c3 +size 3463726472 diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point1360_0point192/unet/config.json b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point1360_0point192/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point1360_0point192/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point1360_0point192/unet/diffusion_pytorch_model.safetensors b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point1360_0point192/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..85d1ab52cd1bf75a464dbcdad625a7a64116ad42 --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_rgb_blending_0point1360_0point192/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8df116e6eb331bf9478cf1ad70f49e2f0bc06eadb859a63c5e0958f72a8cbf6f +size 3463726472 diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point0002125_0point003/unet/config.json b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point0002125_0point003/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point0002125_0point003/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point0002125_0point003/unet/diffusion_pytorch_model.safetensors b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point0002125_0point003/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..f4f70ed01294ae828149b4ffb4f7588691126c0f --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point0002125_0point003/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bb001e3e19839e96a671b2e2f57439cacb66265b7432250ea7a16c30d552a0b3 +size 3463772560 diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point0002125_0point003_wo_multi_res_noise/unet/config.json b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point0002125_0point003_wo_multi_res_noise/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point0002125_0point003_wo_multi_res_noise/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point0002125_0point003_wo_multi_res_noise/unet/diffusion_pytorch_model.safetensors b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point0002125_0point003_wo_multi_res_noise/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..38c9800c7c68b89e69832b73025c621aedfc4d19 --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point0002125_0point003_wo_multi_res_noise/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:40860cbb541a74b24e3266caa8c1bccb4f1327897b344557a01c7b951d417c38 +size 3463772560 diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point000425_0point006/unet/config.json b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point000425_0point006/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point000425_0point006/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point000425_0point006/unet/diffusion_pytorch_model.safetensors b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point000425_0point006/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..91d70332cc1d5a22419f536ddd60cf65e0c349c7 --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point000425_0point006/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:181ea48d7991b7c43b1142f04013f3d69a62fb367b9feec5a665ff58c1a03a9e +size 3463772560 diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point000425_0point006_wo_multi_res_noise/unet/config.json b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point000425_0point006_wo_multi_res_noise/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point000425_0point006_wo_multi_res_noise/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point000425_0point006_wo_multi_res_noise/unet/diffusion_pytorch_model.safetensors b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point000425_0point006_wo_multi_res_noise/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..92420baf4d1f9e8564166fde033575945a4c3a56 --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point000425_0point006_wo_multi_res_noise/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6a39c6f202aed986d09bab8ccdd1b873a6a301c8670503df14307b88269e287e +size 3463772560 diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point00085_0point012/unet/config.json b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point00085_0point012/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point00085_0point012/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point00085_0point012/unet/diffusion_pytorch_model.safetensors b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point00085_0point012/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..4a05f7b78c7757b6504f68882e23e2c7ae22045b --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point00085_0point012/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7f1e90d2c2b3ca7c1a331ff0bcfb4b538f7a0033057a06e50965914d81b85b57 +size 3463772560 diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point00085_0point012_wo_multi_res_noise/unet/config.json b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point00085_0point012_wo_multi_res_noise/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point00085_0point012_wo_multi_res_noise/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point00085_0point012_wo_multi_res_noise/unet/diffusion_pytorch_model.safetensors b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point00085_0point012_wo_multi_res_noise/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..2696004f1e455799ef5927f3b6cca998e3c136eb --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point00085_0point012_wo_multi_res_noise/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:307fcfd5fbd1d60fa5a898f4a333e9b89311137f995527b40ebb10be971e617d +size 3463772560 diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point00340_0point048/unet/config.json b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point00340_0point048/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point00340_0point048/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point00340_0point048/unet/diffusion_pytorch_model.safetensors b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point00340_0point048/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..6d8a244d25baa978aead1dff73b1862bcfe46025 --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point00340_0point048/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c71a32058aeb40f22271e83e3e175d6ef1563ed937b4039e622acfc0e5e4e0fd +size 3463772560 diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point00340_0point048_wo_multi_res_noise/unet/config.json b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point00340_0point048_wo_multi_res_noise/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point00340_0point048_wo_multi_res_noise/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point00340_0point048_wo_multi_res_noise/unet/diffusion_pytorch_model.safetensors b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point00340_0point048_wo_multi_res_noise/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..4ae7328f3aa716578a959155399ffcd71c2c5d1a --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point00340_0point048_wo_multi_res_noise/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dc93bd37c3f2f740ad5fda47f4631788cb5df036b51e5f638fd2b8bbb095b22a +size 3463772560 diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point05440_0point768/unet/config.json b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point05440_0point768/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point05440_0point768/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point05440_0point768/unet/diffusion_pytorch_model.safetensors b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point05440_0point768/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..c255489dedbb26c6dad83b546ec81657bb4b4df0 --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point05440_0point768/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1ea295f516bf34ac568555bfb08c198e2c846831016ea732b07892730f95c92c +size 3463772560 diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point05440_0point768_wo_multi_res_noise/unet/config.json b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point05440_0point768_wo_multi_res_noise/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point05440_0point768_wo_multi_res_noise/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point05440_0point768_wo_multi_res_noise/unet/diffusion_pytorch_model.safetensors b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point05440_0point768_wo_multi_res_noise/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..36ec963d60273b1c4eabe1410042a1f79d9a492c --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point05440_0point768_wo_multi_res_noise/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ac081ec32237ba0c7e05fa38be7454fdae14648759395cfea4cd19d8ea358d07 +size 3463772560 diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point1360_0point192/unet/config.json b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point1360_0point192/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point1360_0point192/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point1360_0point192/unet/diffusion_pytorch_model.safetensors b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point1360_0point192/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..13a0da5bd57445c0f2fb53b95495a848b7725815 --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point1360_0point192/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8cb125af4301bd563579c71d9f162d27d57cece33e996c4ab69feacdc4e874dc +size 3463772560 diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point1360_0point192_wo_multi_res_noise/unet/config.json b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point1360_0point192_wo_multi_res_noise/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point1360_0point192_wo_multi_res_noise/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point1360_0point192_wo_multi_res_noise/unet/diffusion_pytorch_model.safetensors b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point1360_0point192_wo_multi_res_noise/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..02cb0aaaecd5dd068d8c87a7edf5283f72caafa4 --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_0point1360_0point192_wo_multi_res_noise/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b29bc021f3d25edc35592f5126cfa1f808da2af031dba58cf70a20ee0d9fd655 +size 3463772560 diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_1point0_1point0/unet/config.json b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_1point0_1point0/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_1point0_1point0/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_1point0_1point0/unet/diffusion_pytorch_model.safetensors b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_1point0_1point0/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..b0f9c9f00555b2d4a7c4bba7dff1f43737b7ec6a --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_1point0_1point0/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d9eaff485302db77a4aa74cf9b078bbe912eb8bc6fd7553dd583d8cee184bdbf +size 3463772560 diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_1point0_1point0_wo_multi_res_noise/unet/config.json b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_1point0_1point0_wo_multi_res_noise/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_1point0_1point0_wo_multi_res_noise/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_1point0_1point0_wo_multi_res_noise/unet/diffusion_pytorch_model.safetensors b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_1point0_1point0_wo_multi_res_noise/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..d0735a967a811192ca568668b696abac0b4af1fd --- /dev/null +++ b/ablation/beta_values/sd21_train_marigold_exr_1card_ensure_wo_rgb_blending_1point0_1point0_wo_multi_res_noise/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:37c5a1e3ce53c94dcfe342fc156893d17b4cca801ac6b4fd6bd3f8381f8add96 +size 3463772560 diff --git a/ablation/components/sd21_train_genpercept_exr_1card_ensure_dpt_head_identity_mse_loss_decoder_3e-5_bs16_per_accu/dpt_head_identity/model.safetensors b/ablation/components/sd21_train_genpercept_exr_1card_ensure_dpt_head_identity_mse_loss_decoder_3e-5_bs16_per_accu/dpt_head_identity/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..eae023054945d0c815c4972bdd3baa47c499c6d7 --- /dev/null +++ b/ablation/components/sd21_train_genpercept_exr_1card_ensure_dpt_head_identity_mse_loss_decoder_3e-5_bs16_per_accu/dpt_head_identity/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c1ae3f131a505d331f493d9270f523f55601bf15d6a002c01aff38dd600f73ef +size 73903140 diff --git a/ablation/components/sd21_train_genpercept_exr_1card_ensure_dpt_head_identity_mse_loss_decoder_3e-5_bs16_per_accu/unet/config.json b/ablation/components/sd21_train_genpercept_exr_1card_ensure_dpt_head_identity_mse_loss_decoder_3e-5_bs16_per_accu/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/components/sd21_train_genpercept_exr_1card_ensure_dpt_head_identity_mse_loss_decoder_3e-5_bs16_per_accu/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/components/sd21_train_genpercept_exr_1card_ensure_dpt_head_identity_mse_loss_decoder_3e-5_bs16_per_accu/unet/diffusion_pytorch_model.safetensors b/ablation/components/sd21_train_genpercept_exr_1card_ensure_dpt_head_identity_mse_loss_decoder_3e-5_bs16_per_accu/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..0a1448a8520a4f269843bb8bde69d8abc8535567 --- /dev/null +++ b/ablation/components/sd21_train_genpercept_exr_1card_ensure_dpt_head_identity_mse_loss_decoder_3e-5_bs16_per_accu/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0b95ecee234fc2e11a733fd56c7607eb141196fe217d7582577a28159c5e55b7 +size 3463677496 diff --git a/ablation/components/sd21_train_genpercept_exr_1card_ensure_pixel_mse_loss_bs8_per_accu/unet/config.json b/ablation/components/sd21_train_genpercept_exr_1card_ensure_pixel_mse_loss_bs8_per_accu/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/components/sd21_train_genpercept_exr_1card_ensure_pixel_mse_loss_bs8_per_accu/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/components/sd21_train_genpercept_exr_1card_ensure_pixel_mse_loss_bs8_per_accu/unet/diffusion_pytorch_model.safetensors b/ablation/components/sd21_train_genpercept_exr_1card_ensure_pixel_mse_loss_bs8_per_accu/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..052c27fac2d3b1908a842929d0e2feed940e277a --- /dev/null +++ b/ablation/components/sd21_train_genpercept_exr_1card_ensure_pixel_mse_loss_bs8_per_accu/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a47bea69c66317e874279e7b201576ab0b127dc7464eba9a275c52c225790973 +size 3463726472 diff --git a/ablation/components/sd21_train_genpercept_exr_1card_ensure_pixel_mse_ssi_grad_loss_bs8_per_accu/unet/config.json b/ablation/components/sd21_train_genpercept_exr_1card_ensure_pixel_mse_ssi_grad_loss_bs8_per_accu/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/components/sd21_train_genpercept_exr_1card_ensure_pixel_mse_ssi_grad_loss_bs8_per_accu/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/components/sd21_train_genpercept_exr_1card_ensure_pixel_mse_ssi_grad_loss_bs8_per_accu/unet/diffusion_pytorch_model.safetensors b/ablation/components/sd21_train_genpercept_exr_1card_ensure_pixel_mse_ssi_grad_loss_bs8_per_accu/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..aacc132faa1abad9af5ad305c583dbac28565044 --- /dev/null +++ b/ablation/components/sd21_train_genpercept_exr_1card_ensure_pixel_mse_ssi_grad_loss_bs8_per_accu/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3d20b29cd7008bab0a5df67554871e949a9cf86e1b5fab84da6dce5e525a3112 +size 3463726472 diff --git a/ablation/components/sd21_train_genpercept_exr_1card_ensure_unet_from_scratch/unet/config.json b/ablation/components/sd21_train_genpercept_exr_1card_ensure_unet_from_scratch/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/components/sd21_train_genpercept_exr_1card_ensure_unet_from_scratch/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/components/sd21_train_genpercept_exr_1card_ensure_unet_from_scratch/unet/diffusion_pytorch_model.safetensors b/ablation/components/sd21_train_genpercept_exr_1card_ensure_unet_from_scratch/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..59eca306b8108d5630351ab292a43c8e4781a0ce --- /dev/null +++ b/ablation/components/sd21_train_genpercept_exr_1card_ensure_unet_from_scratch/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e5b12a17d09934c0cbfbf78c9e610839edb7d99271cc12bb7a72e32626e22b8d +size 3463726472 diff --git a/ablation/components/sd21_train_genpercept_exr_1card_ensure_vae_decoder_from_scratch_bs4_per_accu/unet/config.json b/ablation/components/sd21_train_genpercept_exr_1card_ensure_vae_decoder_from_scratch_bs4_per_accu/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/components/sd21_train_genpercept_exr_1card_ensure_vae_decoder_from_scratch_bs4_per_accu/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/components/sd21_train_genpercept_exr_1card_ensure_vae_decoder_from_scratch_bs4_per_accu/unet/diffusion_pytorch_model.safetensors b/ablation/components/sd21_train_genpercept_exr_1card_ensure_vae_decoder_from_scratch_bs4_per_accu/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..37607e9fbbe512167490e0c2a0a04b94b491b376 --- /dev/null +++ b/ablation/components/sd21_train_genpercept_exr_1card_ensure_vae_decoder_from_scratch_bs4_per_accu/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:07ff3000eee9a8b72e8509ab59430fc97355332ea51eb1469abd51f45e4ca05c +size 3463726472 diff --git a/ablation/data_quality/sd21_train_genpercept_exr_1card_ensure_taskonomy_cityscapes/unet/config.json b/ablation/data_quality/sd21_train_genpercept_exr_1card_ensure_taskonomy_cityscapes/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/data_quality/sd21_train_genpercept_exr_1card_ensure_taskonomy_cityscapes/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/data_quality/sd21_train_genpercept_exr_1card_ensure_taskonomy_cityscapes/unet/diffusion_pytorch_model.safetensors b/ablation/data_quality/sd21_train_genpercept_exr_1card_ensure_taskonomy_cityscapes/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..4bd47b1c4b36140a65eb1bd70d82eaab49e9e499 --- /dev/null +++ b/ablation/data_quality/sd21_train_genpercept_exr_1card_ensure_taskonomy_cityscapes/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fa583dae96e0103886fe0385e01738eacef7260b1c7c93b42663dc253fb8dff4 +size 3463726472 diff --git a/ablation/lora/sd21_train_genpercept_exr_1card_ensure_dpt_head_identity_mse_loss_decoder_3e-5_freeze_unet/dpt_head_identity/model.safetensors b/ablation/lora/sd21_train_genpercept_exr_1card_ensure_dpt_head_identity_mse_loss_decoder_3e-5_freeze_unet/dpt_head_identity/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..3e4c12776932c0ee09e14378af8f0b4b86bd3c6f --- /dev/null +++ b/ablation/lora/sd21_train_genpercept_exr_1card_ensure_dpt_head_identity_mse_loss_decoder_3e-5_freeze_unet/dpt_head_identity/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:99dbe45f95387c9d4efd593914ebd9815dbc9b7fb3f3cf63539e24e09d549010 +size 73903140 diff --git a/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_1024/unet/config.json b/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_1024/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_1024/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_1024/unet/diffusion_pytorch_model.safetensors b/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_1024/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..0a6b4ad7ca27eb99e0bb9bbe0ea60859f28e19cf --- /dev/null +++ b/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_1024/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:63a10dc18059144a4a7677e3a99acd6b3dd3f248e8d0a355cd13ea772a44d726 +size 4313639160 diff --git a/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_16/unet/config.json b/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_16/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_16/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_16/unet/diffusion_pytorch_model.safetensors b/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_16/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..2a5ae5385fe6fb9698cd177a19486133d80fc388 --- /dev/null +++ b/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_16/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:047d546bd1146633c79a92d90775258cf047d5c789ee1f20fb2dc53ffe57b202 +size 3477046816 diff --git a/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_256/unet/config.json b/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_256/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_256/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_256/unet/diffusion_pytorch_model.safetensors b/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_256/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..9548c41cc398bb506961890b2bf8d7b61a4ba3ef --- /dev/null +++ b/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_256/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e63429baf110b6297b59ed3654f75f8398f0338b333516da3b4967f50daf14f9 +size 3676235640 diff --git a/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_4/unet/config.json b/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_4/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_4/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_4/unet/diffusion_pytorch_model.safetensors b/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_4/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..fd76a1f491dcd1831c3764c715d662845741095c --- /dev/null +++ b/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_4/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:210ccb2c40384e060044f555e1b5f39f407839525103215aad86f50ed1d82cce +size 3467087136 diff --git a/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_64/unet/config.json b/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_64/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_64/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_64/unet/diffusion_pytorch_model.safetensors b/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_64/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..a6d836d4d8678b637250e337675912145ee2165b --- /dev/null +++ b/ablation/lora/sd21_train_genpercept_exr_1card_ensure_lora_rank_64/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:51e4438515a5068260328676ae111850c4058fa4b5812dce8bc961b7a3f300fd +size 3516884536 diff --git a/ablation/text_prompt/sd21_train_genpercept_exr_1card_ensure_text_input_a_high_quality_rgb_image/unet/config.json b/ablation/text_prompt/sd21_train_genpercept_exr_1card_ensure_text_input_a_high_quality_rgb_image/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/text_prompt/sd21_train_genpercept_exr_1card_ensure_text_input_a_high_quality_rgb_image/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/text_prompt/sd21_train_genpercept_exr_1card_ensure_text_input_a_high_quality_rgb_image/unet/diffusion_pytorch_model.safetensors b/ablation/text_prompt/sd21_train_genpercept_exr_1card_ensure_text_input_a_high_quality_rgb_image/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..8390a43f0ee397e81643805d46b9898d2c24dfc1 --- /dev/null +++ b/ablation/text_prompt/sd21_train_genpercept_exr_1card_ensure_text_input_a_high_quality_rgb_image/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a52cd4c6579823b8debfe6de569c3bec191aadf1af92324d29cbc619e3971cd5 +size 3463726472 diff --git a/ablation/text_prompt/sd21_train_genpercept_exr_1card_ensure_text_input_random/unet/config.json b/ablation/text_prompt/sd21_train_genpercept_exr_1card_ensure_text_input_random/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/text_prompt/sd21_train_genpercept_exr_1card_ensure_text_input_random/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/text_prompt/sd21_train_genpercept_exr_1card_ensure_text_input_random/unet/diffusion_pytorch_model.safetensors b/ablation/text_prompt/sd21_train_genpercept_exr_1card_ensure_text_input_random/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..79b7d622874fe70e08eb02951fec622d36b9c585 --- /dev/null +++ b/ablation/text_prompt/sd21_train_genpercept_exr_1card_ensure_text_input_random/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:46b540e0b343326d29ecf7bfe6260c793ba12651c58ce72db1f211fe6b210066 +size 3463726472 diff --git a/ablation/timesteps/sd21_train_genpercept_exr_1card_ensure_fix_timesteps_1/unet/config.json b/ablation/timesteps/sd21_train_genpercept_exr_1card_ensure_fix_timesteps_1/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/timesteps/sd21_train_genpercept_exr_1card_ensure_fix_timesteps_1/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/timesteps/sd21_train_genpercept_exr_1card_ensure_fix_timesteps_1/unet/diffusion_pytorch_model.safetensors b/ablation/timesteps/sd21_train_genpercept_exr_1card_ensure_fix_timesteps_1/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..92f8524a6ee7ddcfdb59df9dce640a8d45654a6a --- /dev/null +++ b/ablation/timesteps/sd21_train_genpercept_exr_1card_ensure_fix_timesteps_1/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e27f91e690b8445cae2d56853b576de002c9a7a684a57c56e90cfe3ebb9c4bca +size 3463726472 diff --git a/ablation/timesteps/sd21_train_genpercept_exr_1card_ensure_fix_timesteps_500/unet/config.json b/ablation/timesteps/sd21_train_genpercept_exr_1card_ensure_fix_timesteps_500/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/timesteps/sd21_train_genpercept_exr_1card_ensure_fix_timesteps_500/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/timesteps/sd21_train_genpercept_exr_1card_ensure_fix_timesteps_500/unet/diffusion_pytorch_model.safetensors b/ablation/timesteps/sd21_train_genpercept_exr_1card_ensure_fix_timesteps_500/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..921298651442643e0e57eb2209f42073a9ea96c0 --- /dev/null +++ b/ablation/timesteps/sd21_train_genpercept_exr_1card_ensure_fix_timesteps_500/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a4757579d674717db109c1c144345913b028d68611650a7581706f4650a3d354 +size 3463726472 diff --git a/ablation/timesteps/sd21_train_genpercept_exr_1card_ensure_fix_timesteps_900/unet/config.json b/ablation/timesteps/sd21_train_genpercept_exr_1card_ensure_fix_timesteps_900/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/ablation/timesteps/sd21_train_genpercept_exr_1card_ensure_fix_timesteps_900/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/ablation/timesteps/sd21_train_genpercept_exr_1card_ensure_fix_timesteps_900/unet/diffusion_pytorch_model.safetensors b/ablation/timesteps/sd21_train_genpercept_exr_1card_ensure_fix_timesteps_900/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..c3e5222aeb5bdfc73aab4915b602f6edeec1e37b --- /dev/null +++ b/ablation/timesteps/sd21_train_genpercept_exr_1card_ensure_fix_timesteps_900/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ec9f7650b79ef077ad3e97df96e94847903bb50b25179bb0ebfb11f5f37a756d +size 3463726472 diff --git a/main_paper/sd21_train_genpercept_exr_1card_ensure_depth_bs8_per_accu_pixel_mse_ssi_grad_loss/unet/config.json b/main_paper/sd21_train_genpercept_exr_1card_ensure_depth_bs8_per_accu_pixel_mse_ssi_grad_loss/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/main_paper/sd21_train_genpercept_exr_1card_ensure_depth_bs8_per_accu_pixel_mse_ssi_grad_loss/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/main_paper/sd21_train_genpercept_exr_1card_ensure_depth_bs8_per_accu_pixel_mse_ssi_grad_loss/unet/diffusion_pytorch_model.safetensors b/main_paper/sd21_train_genpercept_exr_1card_ensure_depth_bs8_per_accu_pixel_mse_ssi_grad_loss/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..aacc132faa1abad9af5ad305c583dbac28565044 --- /dev/null +++ b/main_paper/sd21_train_genpercept_exr_1card_ensure_depth_bs8_per_accu_pixel_mse_ssi_grad_loss/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3d20b29cd7008bab0a5df67554871e949a9cf86e1b5fab84da6dce5e525a3112 +size 3463726472 diff --git a/main_paper/sd21_train_genpercept_exr_1card_ensure_dis_bs4_per_accu_768_pixel_mse_loss/unet/config.json b/main_paper/sd21_train_genpercept_exr_1card_ensure_dis_bs4_per_accu_768_pixel_mse_loss/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/main_paper/sd21_train_genpercept_exr_1card_ensure_dis_bs4_per_accu_768_pixel_mse_loss/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/main_paper/sd21_train_genpercept_exr_1card_ensure_dis_bs4_per_accu_768_pixel_mse_loss/unet/diffusion_pytorch_model.safetensors b/main_paper/sd21_train_genpercept_exr_1card_ensure_dis_bs4_per_accu_768_pixel_mse_loss/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..60ccee7eb423cba902931f7ca441353cefd93602 --- /dev/null +++ b/main_paper/sd21_train_genpercept_exr_1card_ensure_dis_bs4_per_accu_768_pixel_mse_loss/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e829f5bcd4479f27d2606448ca378c27ca662ed9ebc0a1333814a009a9cc46d2 +size 3463726472 diff --git a/main_paper/sd21_train_genpercept_exr_1card_ensure_disparity_bs4_per_accu_768_pixel_mse_ssi_grad_loss/unet/config.json b/main_paper/sd21_train_genpercept_exr_1card_ensure_disparity_bs4_per_accu_768_pixel_mse_ssi_grad_loss/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/main_paper/sd21_train_genpercept_exr_1card_ensure_disparity_bs4_per_accu_768_pixel_mse_ssi_grad_loss/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/main_paper/sd21_train_genpercept_exr_1card_ensure_disparity_bs4_per_accu_768_pixel_mse_ssi_grad_loss/unet/diffusion_pytorch_model.safetensors b/main_paper/sd21_train_genpercept_exr_1card_ensure_disparity_bs4_per_accu_768_pixel_mse_ssi_grad_loss/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..53b9eb725f5bd485f19679fa819ca343fe4e0f04 --- /dev/null +++ b/main_paper/sd21_train_genpercept_exr_1card_ensure_disparity_bs4_per_accu_768_pixel_mse_ssi_grad_loss/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e35c90a751eeee4f3ef4b3da78a23a9519517a71f4b03be6d28aa457df13436a +size 3463726472 diff --git a/main_paper/sd21_train_genpercept_exr_1card_ensure_disparity_bs8_per_accu_768_pixel_mse_ssi_grad_loss_dpt_head_identity_decoder_3e-5/dpt_head_identity/model.safetensors b/main_paper/sd21_train_genpercept_exr_1card_ensure_disparity_bs8_per_accu_768_pixel_mse_ssi_grad_loss_dpt_head_identity_decoder_3e-5/dpt_head_identity/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..5f626fe720bf12f0a6b95642a7c46fe64a30ff43 --- /dev/null +++ b/main_paper/sd21_train_genpercept_exr_1card_ensure_disparity_bs8_per_accu_768_pixel_mse_ssi_grad_loss_dpt_head_identity_decoder_3e-5/dpt_head_identity/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c5c53a100bfe92eb126a8ae765735c1b198e17177abd3ae53505a4a84e5b4e90 +size 73903140 diff --git a/main_paper/sd21_train_genpercept_exr_1card_ensure_disparity_bs8_per_accu_768_pixel_mse_ssi_grad_loss_dpt_head_identity_decoder_3e-5/unet/config.json b/main_paper/sd21_train_genpercept_exr_1card_ensure_disparity_bs8_per_accu_768_pixel_mse_ssi_grad_loss_dpt_head_identity_decoder_3e-5/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/main_paper/sd21_train_genpercept_exr_1card_ensure_disparity_bs8_per_accu_768_pixel_mse_ssi_grad_loss_dpt_head_identity_decoder_3e-5/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/main_paper/sd21_train_genpercept_exr_1card_ensure_disparity_bs8_per_accu_768_pixel_mse_ssi_grad_loss_dpt_head_identity_decoder_3e-5/unet/diffusion_pytorch_model.safetensors b/main_paper/sd21_train_genpercept_exr_1card_ensure_disparity_bs8_per_accu_768_pixel_mse_ssi_grad_loss_dpt_head_identity_decoder_3e-5/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..0af835dac5c64fc3ee772139f30981ee343554d8 --- /dev/null +++ b/main_paper/sd21_train_genpercept_exr_1card_ensure_disparity_bs8_per_accu_768_pixel_mse_ssi_grad_loss_dpt_head_identity_decoder_3e-5/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e8b6d8c10cd1fca5aa3f5e83dcdaa3a3cfdb99290a70b0b483ba3c4259518467 +size 3463677496 diff --git a/main_paper/sd21_train_genpercept_exr_1card_ensure_matting_bs4_per_accu_768_pixel_mse_loss/unet/config.json b/main_paper/sd21_train_genpercept_exr_1card_ensure_matting_bs4_per_accu_768_pixel_mse_loss/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/main_paper/sd21_train_genpercept_exr_1card_ensure_matting_bs4_per_accu_768_pixel_mse_loss/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/main_paper/sd21_train_genpercept_exr_1card_ensure_matting_bs4_per_accu_768_pixel_mse_loss/unet/diffusion_pytorch_model.safetensors b/main_paper/sd21_train_genpercept_exr_1card_ensure_matting_bs4_per_accu_768_pixel_mse_loss/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..167788d98c71dfc07d4d7ebd6102254b7c9aceda --- /dev/null +++ b/main_paper/sd21_train_genpercept_exr_1card_ensure_matting_bs4_per_accu_768_pixel_mse_loss/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:45e3ce9e39b3274ea4a58999f172052fa3b787a6bb6693aefb322c36758e22ff +size 3463726472 diff --git a/main_paper/sd21_train_genpercept_exr_1card_ensure_normal_bs4_per_accu_768_angular_loss/unet/config.json b/main_paper/sd21_train_genpercept_exr_1card_ensure_normal_bs4_per_accu_768_angular_loss/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/main_paper/sd21_train_genpercept_exr_1card_ensure_normal_bs4_per_accu_768_angular_loss/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/main_paper/sd21_train_genpercept_exr_1card_ensure_normal_bs4_per_accu_768_angular_loss/unet/diffusion_pytorch_model.safetensors b/main_paper/sd21_train_genpercept_exr_1card_ensure_normal_bs4_per_accu_768_angular_loss/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..8d18b9cef9600572e2fdba95015def9efa7e5b7a --- /dev/null +++ b/main_paper/sd21_train_genpercept_exr_1card_ensure_normal_bs4_per_accu_768_angular_loss/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6109f943130f72c5d87df3aa5cb774492fa017b38ae72f4b894244479417c985 +size 3463726472 diff --git a/main_paper/sd21_train_genpercept_exr_1card_ensure_seg_bs4_per_accu_768_pixel_mse_loss/unet/config.json b/main_paper/sd21_train_genpercept_exr_1card_ensure_seg_bs4_per_accu_768_pixel_mse_loss/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/main_paper/sd21_train_genpercept_exr_1card_ensure_seg_bs4_per_accu_768_pixel_mse_loss/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/main_paper/sd21_train_genpercept_exr_1card_ensure_seg_bs4_per_accu_768_pixel_mse_loss/unet/diffusion_pytorch_model.safetensors b/main_paper/sd21_train_genpercept_exr_1card_ensure_seg_bs4_per_accu_768_pixel_mse_loss/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..be7126a32679d2e1384d6904d5db3a83391c0d60 --- /dev/null +++ b/main_paper/sd21_train_genpercept_exr_1card_ensure_seg_bs4_per_accu_768_pixel_mse_loss/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:911b5909e84b3fce4988e55e9c38017f349d440e09470307b5082e635a9a0e98 +size 3463726472 diff --git a/others/sd21_train_genpercept_exr_1card_ensure_dis_bs4_per_accu_768_pixel_mse_loss_with_latent_loss/unet/config.json b/others/sd21_train_genpercept_exr_1card_ensure_dis_bs4_per_accu_768_pixel_mse_loss_with_latent_loss/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/others/sd21_train_genpercept_exr_1card_ensure_dis_bs4_per_accu_768_pixel_mse_loss_with_latent_loss/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/others/sd21_train_genpercept_exr_1card_ensure_dis_bs4_per_accu_768_pixel_mse_loss_with_latent_loss/unet/diffusion_pytorch_model.safetensors b/others/sd21_train_genpercept_exr_1card_ensure_dis_bs4_per_accu_768_pixel_mse_loss_with_latent_loss/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..d4ad99236aa9e95bc77ecd91e6a257aa90f376bf --- /dev/null +++ b/others/sd21_train_genpercept_exr_1card_ensure_dis_bs4_per_accu_768_pixel_mse_loss_with_latent_loss/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b21e24d5c269fd7ee09ac770b6b76949d879965a510c088211b83af18f27e7cc +size 3463726472 diff --git a/others/sd21_train_genpercept_exr_1card_ensure_matting_bs4_per_accu_768_pixel_mse_loss_with_latent_loss/unet/config.json b/others/sd21_train_genpercept_exr_1card_ensure_matting_bs4_per_accu_768_pixel_mse_loss_with_latent_loss/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/others/sd21_train_genpercept_exr_1card_ensure_matting_bs4_per_accu_768_pixel_mse_loss_with_latent_loss/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/others/sd21_train_genpercept_exr_1card_ensure_matting_bs4_per_accu_768_pixel_mse_loss_with_latent_loss/unet/diffusion_pytorch_model.safetensors b/others/sd21_train_genpercept_exr_1card_ensure_matting_bs4_per_accu_768_pixel_mse_loss_with_latent_loss/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..812062ce2870fddab7ae8bb42b01ea13e77526c3 --- /dev/null +++ b/others/sd21_train_genpercept_exr_1card_ensure_matting_bs4_per_accu_768_pixel_mse_loss_with_latent_loss/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f76b90c019772d84636f4f415203c2ac9974775dd681e50b58d1b7c11cc8cbd8 +size 3463726472 diff --git a/others/sd21_train_genpercept_exr_1card_ensure_normal_bs4_per_accu_768_angular_loss_with_latent_loss/unet/config.json b/others/sd21_train_genpercept_exr_1card_ensure_normal_bs4_per_accu_768_angular_loss_with_latent_loss/unet/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b1458658e8651398962171a8c5c56c5c0bd5aea --- /dev/null +++ b/others/sd21_train_genpercept_exr_1card_ensure_normal_bs4_per_accu_768_angular_loss_with_latent_loss/unet/config.json @@ -0,0 +1,46 @@ +{ + "_class_name": "UNet2DConditionModel", + "_diffusers_version": "0.10.0.dev0", + "act_fn": "silu", + "attention_head_dim": [ + 5, + 10, + 20, + 20 + ], + "block_out_channels": [ + 320, + 640, + 1280, + 1280 + ], + "center_input_sample": false, + "cross_attention_dim": 1024, + "down_block_types": [ + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "CrossAttnDownBlock2D", + "DownBlock2D" + ], + "downsample_padding": 1, + "dual_cross_attention": false, + "flip_sin_to_cos": true, + "freq_shift": 0, + "in_channels": 4, + "layers_per_block": 2, + "mid_block_scale_factor": 1, + "norm_eps": 1e-05, + "norm_num_groups": 32, + "num_class_embeds": null, + "only_cross_attention": false, + "out_channels": 4, + "sample_size": 96, + "up_block_types": [ + "UpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D", + "CrossAttnUpBlock2D" + ], + "use_linear_projection": true, + "upcast_attention": true +} diff --git a/others/sd21_train_genpercept_exr_1card_ensure_normal_bs4_per_accu_768_angular_loss_with_latent_loss/unet/diffusion_pytorch_model.safetensors b/others/sd21_train_genpercept_exr_1card_ensure_normal_bs4_per_accu_768_angular_loss_with_latent_loss/unet/diffusion_pytorch_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..5348e06ff93e23ad49f403930ec78f4042609481 --- /dev/null +++ b/others/sd21_train_genpercept_exr_1card_ensure_normal_bs4_per_accu_768_angular_loss_with_latent_loss/unet/diffusion_pytorch_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5fd50eb82d514881f807f79175b176ec309786be622cf629719b8090731ff637 +size 3463726472