oops, stay in float32

examples/efficientnet.py

@@ -108,7 +108,8 @@ class EfficientNet:
           mv = eval(mk.replace(".weight", ""))
         except AttributeError:
           mv = eval(mk.replace(".bias", "_bias"))
-      mv.data[:] = v.numpy() if k != '_fc.weight' else v.numpy().T
+      vnp = v.numpy().astype(np.float32)
+      mv.data[:] = vnp if k != '_fc.weight' else vnp.T
 
 if __name__ == "__main__":
   # instantiate my net
@@ -121,7 +122,7 @@ if __name__ == "__main__":
   img = img.resize((224, 224))
   img = np.moveaxis(np.array(img), [2,0,1], [0,1,2])
   img = img.astype(np.float32).reshape(1,3,224,224)
-  print(img.shape)
+  print(img.shape, img.dtype)
 
   # run the net
   out = model.forward(Tensor(img))

tinygrad/tensor.py

@@ -15,7 +15,7 @@ class Tensor:
       assert(False)
     if data.dtype != np.float32:
       # warning? float64 is actually needed for numerical jacobian
-      pass
+      print("warning, %r isn't float32" % (data.shape,))
 
     self.data = data
     self.grad = None
@@ -75,11 +75,11 @@ class Tensor:
     return self.sum().mul(div)
 
   def sqrt(self):
-    root = Tensor(np.zeros(self.shape)+0.5)
+    root = Tensor(np.zeros(self.shape, dtype=self.data.dtype)+0.5)
     return self.pow(root)
 
   def div(self, y):
-    root = Tensor(np.zeros(self.shape)-1)
+    root = Tensor(np.zeros(self.shape, dtype=self.data.dtype)-1)
     return self.mul(y.pow(root))
 
 # An instantiation of the Function is the Context