Rename weights to model.pt; minimize README

README.md

@@ -7,28 +7,7 @@ tags:
 
 # PartFlow — model weights
 
-Trained stage models for **PartFlow**, a feedforward 3D editing network that
-learns from semantic-part transformations (project: *Pxform — Feedforward 3D
-Editing Learns from Semantic-Part Transformation*).
-
-Inference code: see the PartFlow GitHub repository.
-
-## Contents
-
-| Path | Stage | Description |
-|---|---|---|
-| `stage1_ss/denoiser_ema0.9999_step0080000.pt` | Stage 1 | Sparse-structure flow model (EMA, step 80k) |
-| `stage1_ss/config.json` | — | Stage 1 model config |
-| `stage2_slat/denoiser_ema0.9999_step0080000.pt` | Stage 2 | Structured-latent (SLAT) flow model (EMA, step 80k) |
-| `stage2_slat/config.json` | — | Stage 2 model config |
-
-These are EMA weights (decay 0.9999). The frozen DINOv2 encoder and TRELLIS
-SS/SLAT decoders are not included — PartFlow fetches them from their official
-repos at inference time.
-
-## Usage
-
-```bash
-python download_weights.py        # pulls this repo into ./weights/
-python inference.py --input examples/add_ring_band --output_dir outputs
-```
+| Path | Stage |
+|---|---|
+| `stage1_ss/model.pt` | Stage 1 |
+| `stage2_slat/model.pt` | Stage 2 |

stage1_ss/config.json

@@ -1,87 +0,0 @@
-{
-    "models": {
-        "denoiser": {
-            "name": "ControlledSparseStructureFlowModel",
-            "args": {
-                "resolution": 16,
-                "in_channels": 8,
-                "out_channels": 8,
-                "model_channels": 1024,
-                "cond_channels": 1024,
-                "num_blocks": 24,
-                "num_heads": 16,
-                "mlp_ratio": 4,
-                "patch_size": 1,
-                "pe_mode": "ape",
-                "qk_rms_norm": true,
-                "use_fp16": true
-            }
-        }
-    },
-    "dataset": {
-        "name": "H3DOriImageSparseStructureLatent",
-        "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": "EditingControlNetImageFlowMatchingCFGTrainer",
-        "args": {
-            "max_steps": 120000,
-            "batch_split": 1,
-            "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/ss_flow_img_dit_L_16l8_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_ss_mask_loss": true,
-            "lambda_ss_mask": 6.0,
-            "disable_mask_loss_for_global_style": true
-        }
-    }
-}

stage2_slat/config.json

@@ -1,91 +0,0 @@
-{
-    "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
-        }
-    }
-}