Upload run_txt2img_axe_infer.py with huggingface_hub

run_txt2img_axe_infer.py

@@ -134,6 +134,7 @@ if __name__ == '__main__':
         # print(i, timestep)
         
         unet_start = time.time()
+        latent = latent.astype(np.float32)
         noise_pred = unet_session_main.run(None, {"sample": latent, \
                                             "/down_blocks.0/resnets.0/act_1/Mul_output_0": np.expand_dims(time_input[i], axis=0), \
                                             "encoder_hidden_states": prompt_embeds_npy})[0]
@@ -175,7 +176,7 @@ if __name__ == '__main__':
     # vae inference
     vae_start = time.time()    
     latent = latent / 0.18215
-    image = vae_decoder.run(None, {"x": latent})[0]
+    image = vae_decoder.run(None, {"x": latent.astype(np.float32)})[0]
     print(f"vae inference take {(1000 * (time.time() - vae_start)):.1f}ms")
     
     # save result