Upload stage2_slat/config.json with huggingface_hub

stage2_slat/config.json

@@ -0,0 +1,91 @@
+{
+    "models": {
+        "denoiser": {
+            "name": "ControlledSLatFlowModel",
+            "args": {
+                "resolution": 64,
+                "in_channels": 8,
+                "out_channels": 8,
+                "model_channels": 1024,
+                "cond_channels": 1024,
+                "num_blocks": 24,
+                "num_heads": 16,
+                "mlp_ratio": 4,
+                "patch_size": 2,
+                "num_io_res_blocks": 2,
+                "io_block_channels": [
+                    128
+                ],
+                "pe_mode": "ape",
+                "qk_rms_norm": true,
+                "use_fp16": true
+            }
+        }
+    },
+    "dataset": {
+        "name": "H3DOriSLatImage",
+        "args": {
+            "h3d_root_path": "/mnt/zsn/zsn_workspace/PartCraft3D/data/H3D_v1",
+            "split": "train",
+            "edit_types": [
+                "scale",
+                "modification",
+                "addition",
+                "deletion"
+            ],
+            "final_pass_only": false,
+            "image_size": 518,
+            "mask_sidecar_root": "/mnt/zsn/zsn_workspace/Ori3DEdit_wjw/outputs/mask_sidecars"
+        }
+    },
+    "trainer": {
+        "name": "EditingControlNetImageSparseFlowMatchingCFGTrainer",
+        "args": {
+            "max_steps": 100000,
+            "batch_split": 2,
+            "optimizer": {
+                "name": "AdamW",
+                "args": {
+                    "lr": 0.00012,
+                    "weight_decay": 0.0
+                }
+            },
+            "ema_rate": [
+                0.9999
+            ],
+            "fp16_mode": "inflat_all",
+            "fp16_scale_growth": 0.001,
+            "grad_clip": {
+                "name": "AdaptiveGradClipper",
+                "args": {
+                    "max_norm": 1.0,
+                    "clip_percentile": 95
+                }
+            },
+            "i_print": 50,
+            "i_log": 50,
+            "i_sample": 99999,
+            "i_save": 5000,
+            "p_uncond": 0.1,
+            "t_schedule": {
+                "name": "logitNormal",
+                "args": {
+                    "mean": 1.0,
+                    "std": 1.0
+                }
+            },
+            "sigma_min": 1e-05,
+            "image_cond_model": "dinov2_vitl14_reg",
+            "controlnet_pretrain": "./pretrain/slat_flow_img_dit_L_64l8p2_fp16.safetensors",
+            "batch_size": 16,
+            "batch_size_per_gpu": 8,
+            "dataloader_num_workers": 2,
+            "dataloader_pin_memory": true,
+            "dataloader_persistent_workers": true,
+            "use_mask_loss": true,
+            "use_slat_mask_loss": true,
+            "lambda_slat_mask": 0.3,
+            "disable_mask_loss_for_global_style": true
+        }
+    }
+}