From 00fa3ac012a4dd9b0d4535a9bd0bfd311c1c178c Mon Sep 17 00:00:00 2001
From: WncFht <2130212584@qq.com>
Date: Sun, 13 Jul 2025 11:44:49 +0800
Subject: [PATCH 1/2] =?UTF-8?q?feat:=20=E4=B8=BA=20gradio=5Fapp.py=20?=
 =?UTF-8?q?=E5=8A=A0=E4=B8=8A=20enable=5Fflashvdm?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 gradio_app.py | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/gradio_app.py b/gradio_app.py
index 47ec574..46b219d 100644
--- a/gradio_app.py
+++ b/gradio_app.py
@@ -748,8 +748,7 @@ if __name__ == '__main__':
     parser.add_argument('--compile', action='store_true')
     parser.add_argument('--low_vram_mode', action='store_true')
     args = parser.parse_args()
-    args.enable_flashvdm = False
-
+    
     SAVE_DIR = args.cache_path
     os.makedirs(SAVE_DIR, exist_ok=True)
 

From a6509b95fb690f68a60b70e6d03c623aecdca842 Mon Sep 17 00:00:00 2001
From: WncFht <2130212584@qq.com>
Date: Sun, 13 Jul 2025 11:48:13 +0800
Subject: [PATCH 2/2] =?UTF-8?q?feat:=20=E4=B8=BA=20api=5Fserver=20?=
 =?UTF-8?q?=E5=8A=A0=E4=B8=8A=20enable=5Fflashvdm?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 api_server.py   |  8 +++++++-
 model_worker.py | 15 +++++++++++++--
 2 files changed, 20 insertions(+), 3 deletions(-)

diff --git a/api_server.py b/api_server.py
index 1257b7f..ad1f219 100644
--- a/api_server.py
+++ b/api_server.py
@@ -199,7 +199,10 @@ if __name__ == "__main__":
     parser.add_argument("--model_path", type=str, default='tencent/Hunyuan3D-2.1')
     parser.add_argument("--subfolder", type=str, default='hunyuan3d-dit-v2-1')
     parser.add_argument("--device", type=str, default="cuda")
+    parser.add_argument('--mc_algo', type=str, default='mc')
     parser.add_argument("--limit-model-concurrency", type=int, default=5)
+    parser.add_argument('--enable_flashvdm', action='store_true')
+    parser.add_argument('--compile', action='store_true')
     parser.add_argument('--low_vram_mode', action='store_true')
     parser.add_argument('--cache-path', type=str, default='./gradio_cache')
     args = parser.parse_args()
@@ -219,6 +222,9 @@ if __name__ == "__main__":
         low_vram_mode=args.low_vram_mode,
         worker_id=worker_id,
         model_semaphore=model_semaphore,
-        save_dir=SAVE_DIR
+        save_dir=SAVE_DIR,
+        mc_algo=args.mc_algo,
+        enable_flashvdm=args.enable_flashvdm,
+        compile=args.compile
     )
     uvicorn.run(app, host=args.host, port=args.port, log_level="info")
diff --git a/model_worker.py b/model_worker.py
index 536d448..854f1d0 100644
--- a/model_worker.py
+++ b/model_worker.py
@@ -64,7 +64,10 @@ class ModelWorker:
                  low_vram_mode=False,
                  worker_id=None,
                  model_semaphore=None,
-                 save_dir='gradio_cache'):
+                 save_dir='gradio_cache',
+                 mc_algo='mc',
+                 enable_flashvdm=False,
+                 compile=False):
         """
         Initialize the model worker.
         
@@ -83,6 +86,9 @@ class ModelWorker:
         self.low_vram_mode = low_vram_mode
         self.model_semaphore = model_semaphore
         self.save_dir = save_dir
+        self.mc_algo = mc_algo
+        self.enable_flashvdm = enable_flashvdm
+        self.compile = compile
         
         logger.info(f"Loading the model {model_path} on worker {self.worker_id} ...")
 
@@ -91,7 +97,12 @@ class ModelWorker:
         
         # Initialize shape generation pipeline (matching demo.py)
         self.pipeline = Hunyuan3DDiTFlowMatchingPipeline.from_pretrained(model_path)
-        
+        if self.enable_flashvdm:
+            mc_algo = 'mc' if self.device in ['cpu', 'mps'] else self.mc_algo
+            self.pipeline.enable_flashvdm(mc_algo=mc_algo)
+        if self.compile:
+            self.pipeline.compile()
+            
         # Initialize texture generation pipeline (matching demo.py)
         max_num_view = 6  # can be 6 to 9
         resolution = 512  # can be 768 or 512