Rebased to main after refit acceleration is merged.

Author: cehongwang

examples/dynamo/mutable_torchtrt_module_example.py

@@ -21,8 +21,8 @@
 import torch_tensorrt as torch_trt
 import torchvision.models as models
 
-np.random.seed(0)
-torch.manual_seed(0)
+np.random.seed(5)
+torch.manual_seed(5)
 inputs = [torch.rand((1, 3, 224, 224)).to("cuda")]
 
 # %%
@@ -76,7 +76,7 @@
 from diffusers import DiffusionPipeline
 
 with torch.no_grad():
-    kwargs = {
+    settings = {
         "use_python_runtime": True,
         "enabled_precisions": {torch.float16},
         "debug": True,
@@ -86,7 +86,7 @@
     model_id = "runwayml/stable-diffusion-v1-5"
     device = "cuda:0"
 
-    prompt = "portrait of a woman standing, shuimobysim, wuchangshuo, best quality"
+    prompt = "house in forest, shuimobysim, wuchangshuo, best quality"
     negative = "(worst quality:2), (low quality:2), (normal quality:2), lowres, normal quality, skin spots, acnes, skin blemishes, age spot, glans, (watermark:2),"
 
     pipe = DiffusionPipeline.from_pretrained(
@@ -95,7 +95,7 @@
     pipe.to(device)
 
     # The only extra line you need
-    pipe.unet = torch_trt.MutableTorchTensorRTModule(pipe.unet, **kwargs)
+    pipe.unet = torch_trt.MutableTorchTensorRTModule(pipe.unet, **settings)
 
     image = pipe(prompt, negative_prompt=negative, num_inference_steps=30).images[0]
     image.save("./without_LoRA_mutable.jpg")

py/torch_tensorrt/dynamo/runtime/_MutableTorchTensorRTModule.py

@@ -25,7 +25,7 @@ class RefitFlag(Enum):
 
 
 class RefitState:
-    _state: RefitFlag = RefitFlag.UNKNOWN
+    _state: RefitFlag = RefitFlag.NEEDS_RECOMPILE
 
     def set_state(self, state: RefitFlag) -> None:
         if isinstance(state, RefitFlag):
@@ -267,12 +267,14 @@ def refit_gm(self) -> None:
                     self.original_model.state_dict()
                 )
             )
-        self.gm = refit_module_weights(self.gm, self.exp_program)
+        self.gm = refit_module_weights(
+            self.gm, self.exp_program, use_weight_map_cache=True, in_place=True
+        )
 
         self.original_model.cpu()
         torch.cuda.empty_cache()
 
-    def _compile(self) -> None:
+    def compile(self) -> None:
         """
         (Re)compile the TRT graph module using the PyTorch module.
         This function should be called whenever the weight structure get changed (shape, more layers...)
@@ -349,7 +351,7 @@ def forward(self, *args: Any, **kwargs: Any) -> Any:
         # Step 3: Refit/recompile accordingly
         if self.refit_state.get_state() == RefitFlag.NEEDS_RECOMPILE:
             logger.info("(Re)Compiling the engine...")
-            self._compile()
+            self.compile()
             self.store_state_dict_metadata()
             self.refit_state.set_state(RefitFlag.LIVE)
 
@@ -360,7 +362,7 @@ def forward(self, *args: Any, **kwargs: Any) -> Any:
             except Exception as e:
                 logger.error(e)
                 logger.error("Model refit failed. Recompiling the graph module.")
-                self._compile()
+                self.compile()
                 self.store_state_dict_metadata()
             self.refit_state.set_state(RefitFlag.LIVE)
 
@@ -369,6 +371,10 @@ def forward(self, *args: Any, **kwargs: Any) -> Any:
         self.run_info = (args, kwargs, result)
         return result
 
+    def to(self, device: str):
+        logger.warning("Original PyTorch model is moved. CPU offload may failed.")
+        self.orignial_model.to(device)
+        
     def __deepcopy__(self, memo: Any) -> Any:
         cls = self.__class__
         result = cls.__new__(cls)