fix all bugs

examples/test_onnx_imagenet.py

@@ -70,7 +70,7 @@ if __name__ == "__main__":
     GlobalCounters.reset()
     p = run_onnx_jit(**{t_name:img})
     assert p.shape == (1,1000)
-    t = p.argmax().item()
+    t = p.to('cpu').argmax().item()
     hit += y==t
     print(f"target: {y:3d}  pred: {t:3d}  acc: {hit/(i+1)*100:.2f}%")
 

tinygrad/runtime/ops_dsp.py

@@ -318,6 +318,7 @@ def vectorize_shuffle(vec:UOp):
   return None
 
 def multicore_range(r:UOp):
+  # NOTE: THIS IS BROKEN if this is a reduce range. TODO: check for that
   if getenv("MULTICORE", 0) != 1: return None
   if any(x.op is Ops.SPECIAL for x in r.toposort): return None
   core = UOp(Ops.SPECIAL, dtypes.int, arg=("g0", 2))