[docs] Second push of changes #5391
Conversation
|
Test PASSed. |
richardliaw
changed the title
|
Test PASSed. |
|
Test PASSed. |
|
Test PASSed. |
|
Test FAILed. |
|
Test PASSed. |
richardliaw
changed the title
|
Test FAILed. |
| - `Tune`_: Hyperparameter Optimization Framework | ||
| - `RLlib`_: Scalable Reinforcement Learning | ||
| - `Distributed Training <http://ray.readthedocs.io/en/latest/distributed_training.html>`__ | ||
| .. code-block:: python |
There was a problem hiding this comment.
Maybe add some additional line above this one like "Execute Python functions in parallel."
|
|
||
| .. _`installation documentation`: http://ray.readthedocs.io/en/latest/installation.html | ||
|
|
There was a problem hiding this comment.
Do we want these links to the papers? If we keep them, should we add the other relevant papers, e.g., RLlib?
There was a problem hiding this comment.
We should yeah (currently, they can be more informative than the rest of the documentation!)
doc/source/installation.rst
Outdated
|
|
||
| .. code-block:: bash | ||
|
|
||
| pip install git+https://github.com/ray-project/ray.git#subdirectory=python |
There was a problem hiding this comment.
This line doesn't quite work because we still need bazel to be installed first. @mitar has pointed this out.
doc/source/object-store.rst
Outdated
| # Note that inside the remote function, the actual argument is provided. | ||
| return len(y) | ||
|
|
||
| argument = [1, 2, 3, 4] |
There was a problem hiding this comment.
No longer accurate. We put them in the object store if ray._raylet. returns False and otherwise we pickle them and inline them in the task spec. In this case, the list will be inlined in the task spec. This happens whenever the object consists of primitive types and/or numpy arrays and in aggregate is smaller than about 10KB ish.
Of course check_simple_value is not part of the Ray API.
doc/source/object-store.rst
Outdated
| # This implicitly places `argument` into the object store. | ||
| remote_function.remote(argument) | ||
|
|
||
| In the single node setting, all Ray workers can read the same object in the object store without copying (zero-copy reads). Once an object is placed in the object store, it is immutable. Any writes to the read-only object will result in a copy into the local process memory. |
There was a problem hiding this comment.
Not quite true, writes are more likely to just raise an exception saying that the object is read only (for numpy arrays). For non numpy arrays, we actually do create a copy.
cc @pcmoritz
README.rst
Outdated
|
|
||
| `Tune`_ is a library for hyperparameter tuning at any scale. | ||
|
|
||
| - Launch a multi-node distributed asynchronous hyperparameter sweep in less than 10 lines of code. |
There was a problem hiding this comment.
| - Launch a multi-node distributed asynchronous hyperparameter sweep in less than 10 lines of code. | |
| - Launch a multi-node distributed hyperparameter sweep in less than 10 lines of code. |
doc/source/object-store.rst
Outdated
| - Whenever possible, use numpy arrays for maximum performance. | ||
|
|
||
|
|
||
| Object Store (Plasma) |
There was a problem hiding this comment.
This should be moved to the top of the doc, and the serialization stuff to the bottom (serialization is just not that important for most users.)
|
Test PASSed. |
|
Test PASSed. |
|
Test PASSed. |
|
Test PASSed. |
|
Test FAILed. |
|
Test PASSed. |
|
Test PASSed. |
|
Test PASSed. |
What do these changes do?
Related issue number
Linter
scripts/format.shto lint the changes in this PR.