refactor(gradio): replace CPU offload with direct GPU unload/lazy-load

Instead of .to('cpu') / .to('cuda'), models are now fully del'd from
GPU (no CPU intermediate) and reloaded on demand:

- _unload_i23d_worker(): del + gc.collect() + empty_cache()
- _ensure_i23d_worker(): lazy reload from pretrained if None
- _unload_tex_pipeline(): del + gc.collect() + empty_cache()
- _ensure_tex_pipeline(): lazy load from tex_conf if None

generation_all() flow in low_vram_mode:
  shape gen → _unload_i23d_worker → _ensure_tex_pipeline →
  texture gen → _unload_tex_pipeline
  (shape model reloads on next _gen_shape call via _ensure_i23d_worker)

Startup: tex_pipeline NOT loaded in low_vram_mode (only tex_conf stored),
reducing startup VRAM from ~13.5GB to ~7.25GB.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

gradio_app.py

@@ -28,6 +28,7 @@ except Exception as e:
     print(f"Warning: Failed to apply torchvision fix: {e}")
 
 
+import gc
 import os
 import random
 import shutil
@@ -213,6 +214,53 @@ height="{height}" width="100%" frameborder="0"></iframe>'
         </div>
     """
 
+# ---------------------------------------------------------------------------
+# VRAM management helpers (used when --low_vram_mode is set)
+# Models are unloaded (del'd) before the other model runs, then reloaded
+# on next request — no CPU intermediate, VRAM freed immediately.
+# ---------------------------------------------------------------------------
+
+def _unload_i23d_worker():
+    """Delete shape model from GPU and free VRAM."""
+    global i23d_worker
+    del i23d_worker
+    i23d_worker = None
+    gc.collect()
+    torch.cuda.empty_cache()
+
+
+def _ensure_i23d_worker():
+    """Reload shape model to GPU if it was previously unloaded."""
+    global i23d_worker
+    if i23d_worker is None:
+        from hy3dshape import Hunyuan3DDiTFlowMatchingPipeline
+        logger.info("Reloading shape model to GPU...")
+        i23d_worker = Hunyuan3DDiTFlowMatchingPipeline.from_pretrained(
+            args.model_path,
+            subfolder=args.subfolder,
+            use_safetensors=False,
+            device=args.device,
+        )
+
+
+def _unload_tex_pipeline():
+    """Delete texture pipeline from GPU and free VRAM."""
+    global tex_pipeline
+    del tex_pipeline
+    tex_pipeline = None
+    gc.collect()
+    torch.cuda.empty_cache()
+
+
+def _ensure_tex_pipeline():
+    """Load texture pipeline to GPU if not already loaded."""
+    global tex_pipeline
+    if tex_pipeline is None and tex_conf is not None:
+        from hy3dpaint.textureGenPipeline import Hunyuan3DPaintPipeline
+        logger.info("Loading texture pipeline to GPU...")
+        tex_pipeline = Hunyuan3DPaintPipeline(tex_conf)
+
+
 @spaces.GPU(duration=60)
 def _gen_shape(
     caption=None,
@@ -297,6 +345,9 @@ def _gen_shape(
     # image to white model
     start_time = time.time()
 
+    if args.low_vram_mode:
+        _ensure_i23d_worker()
+
     generator = torch.Generator()
     generator = generator.manual_seed(int(seed))
     outputs = i23d_worker(
@@ -379,18 +430,17 @@ def generation_all(
 
     text_path = os.path.join(save_folder, f'textured_mesh.obj')
 
-    # In low_vram_mode: offload shape model to CPU before texture gen to free VRAM,
+    # In low_vram_mode: unload shape model entirely (del, no CPU copy) to free VRAM,
-    # mirroring the sequential-load strategy in batch_generate.py.
+    # then load texture pipeline on demand. Shape model reloads lazily on next request.
     if args.low_vram_mode:
-        i23d_worker.to('cpu')
+        _unload_i23d_worker()
-        torch.cuda.empty_cache()
+        _ensure_tex_pipeline()
 
     path_textured = tex_pipeline(mesh_path=path, image_path=image, output_mesh_path=text_path, save_glb=False)
 
-    # Restore shape model to GPU so subsequent requests don't need to reload from disk.
+    # Unload texture pipeline after use so VRAM is free for the next shape request.
     if args.low_vram_mode:
-        i23d_worker.to('cuda')
+        _unload_tex_pipeline()
-        torch.cuda.empty_cache()
         
     logger.info("---Texture Generation takes %s seconds ---" % (time.time() - tmp_time))
     stats['time']['texture generation'] = time.time() - tmp_time
@@ -808,11 +858,13 @@ if __name__ == '__main__':
             #     texgen_worker.enable_model_cpu_offload()
 
             from hy3dpaint.textureGenPipeline import Hunyuan3DPaintPipeline, Hunyuan3DPaintConfig
-            conf = Hunyuan3DPaintConfig(max_num_view=9, resolution=512)
+            tex_conf = Hunyuan3DPaintConfig(max_num_view=9, resolution=512)
-            conf.realesrgan_ckpt_path = "hy3dpaint/ckpt/RealESRGAN_x4plus.pth"
+            tex_conf.realesrgan_ckpt_path = "hy3dpaint/ckpt/RealESRGAN_x4plus.pth"
-            conf.multiview_cfg_path = "hy3dpaint/cfgs/hunyuan-paint-pbr.yaml"
+            tex_conf.multiview_cfg_path = "hy3dpaint/cfgs/hunyuan-paint-pbr.yaml"
-            conf.custom_pipeline = "hy3dpaint/hunyuanpaintpbr"
+            tex_conf.custom_pipeline = "hy3dpaint/hunyuanpaintpbr"
-            tex_pipeline = Hunyuan3DPaintPipeline(conf)
+            if not args.low_vram_mode:
+                # Load immediately; in low_vram_mode we load on-demand per request.
+                tex_pipeline = Hunyuan3DPaintPipeline(tex_conf)
         
             # Not help much, ignore for now.
             # if args.compile: