Remove TensorRT support at the current time until we validate it works, remove time step recorder

invokeai/app/invocations/onnx.py

@@ -264,8 +264,6 @@ class ONNXTextToLatentsInvocation(BaseInvocation):
                     (input.type for input in unet.session.get_inputs() if input.name == "timestep"), "tensor(float16)"
                 )
                 timestep_dtype = ORT_TO_NP_TYPE[timestep_dtype]
-                import time
-                times = []
                 for i in tqdm(range(len(scheduler.timesteps))):
                     t = scheduler.timesteps[i]
                     # expand the latents if we are doing classifier free guidance
@@ -275,9 +273,7 @@ class ONNXTextToLatentsInvocation(BaseInvocation):
 
                     # predict the noise residual
                     timestep = np.array([t], dtype=timestep_dtype)
-                    start_time = time.time()
                     noise_pred = unet(sample=latent_model_input, timestep=timestep, encoder_hidden_states=prompt_embeds)
-                    times.append(time.time() - start_time)
                     noise_pred = noise_pred[0]
 
                     # perform guidance
@@ -307,7 +303,6 @@ class ONNXTextToLatentsInvocation(BaseInvocation):
                     # call the callback, if provided
                     #if callback is not None and i % callback_steps == 0:
                     #    callback(i, t, latents)
-                print(times)
                 unet.release_session()
 
         torch.cuda.empty_cache()

invokeai/backend/model_management/models/base.py

@@ -581,6 +581,8 @@ class IAIOnnxRuntimeModel:
                 providers.append(self.provider)
             else:
                 providers = get_available_providers()
+            if "TensorrtExecutionProvider" in providers:
+                providers.remove("TensorrtExecutionProvider")
             try:
                 self.session = InferenceSession(self.proto.SerializeToString(), providers=providers, sess_options=sess)
             except Exception as e: