to device support (#10)

* to device support

.gitignore

@@ -166,3 +166,4 @@ runs/
 jupyters/custom_*
 
 *logs/
+.DS_store

README.md

@@ -2,7 +2,7 @@
 ### Generate images from texts
 
 ```
-pip install rudalle==0.0.1rc2
+pip install rudalle==0.0.1rc3
 ```
 ### 🤗 HF Models:
 [ruDALL-E Malevich (XL)](https://huggingface.co/sberbank-ai/rudalle-Malevich)

rudalle/__init__.py

@@ -22,4 +22,4 @@
     'image_prompts',
 ]
 
-__version__ = '0.0.1-rc2'
+__version__ = '0.0.1-rc3'

rudalle/dalle/__init__.py

@@ -59,6 +59,9 @@ def get_rudalle_model(name, pretrained=True, fp16=False, device='cpu', cache_dir
     # TODO docstring
     assert name in MODELS
 
+    if fp16 and device == 'cpu':
+        print('Warning! Using both fp16 and cpu doesnt support. You can use cuda device or turn off fp16.')
+
     config = MODELS[name]
     model = DalleModel(device=device, fp16=fp16, **config['model_params'])
     if pretrained:

rudalle/dalle/fp16.py

@@ -58,3 +58,7 @@ def load_state_dict(self, state_dict, strict=True):
 
     def get_param(self, item):
         return self.module.get_param(item)
+
+    def to(self, device, *args, **kwargs):
+        self.module.to(device)
+        return super().to(device, *args, **kwargs)

rudalle/dalle/model.py

@@ -169,3 +169,9 @@ def forward(
 
         loss = (loss_text + self.loss_img_weight * loss_img) / (self.loss_img_weight + 1)
         return loss, {'text': loss_text.data.detach().float(), 'image': loss_img.data.detach().float()}
+
+    def to(self, device, *args, **kwargs):
+        self.device = device
+        self._mask_map = [mask.to(device) for mask in self._mask_map]
+        self.transformer._mask_map = [mask.to(device) for mask in self.transformer._mask_map]
+        return super().to(device, *args, **kwargs)