Error trying to deserialize an object

```python
from distributed.protocol import deserialize_bytes
```

**What happened**:
Error trying to deserialize an object `dask_ml.decomposition.PCA` already fitted.

```python
OSError: Timed out trying to connect to tcp://10.xxx.120.x31:35487 after 30 s
```
I do a fit with my data and export a serialized file.
When trying to instantiate on a _**different cluster**_, it appears that the file tries to connect with the scheduler of the cluster it was trained on.

**What you expected to happen**:
I expected that, regardless of the context (cluster ip, number of workers...), I would be able to import my PCA object.
I've already tried to dump using the `pickle` and `dill` libraries
```python
deserialize_bytes('gs://file.pkl')
```