############################# Distributed XGBoost with Dask ############################# `Dask `_ is a parallel computing library built on Python. Dask allows easy management of distributed workers and excels at handling large distributed data science workflows. The implementation in XGBoost originates from `dask-xgboost `_ with some extended functionalities and a different interface. The tutorial here focuses on basic usage of dask with CPU tree algorithms. For an overview of GPU based training and internal workings, see `A New, Official Dask API for XGBoost `_. .. note:: The integration is not tested with Windows. **Contents** .. contents:: :backlinks: none :local: ************ Requirements ************ Dask can be installed using either pip or conda (see the dask `installation documentation `_ for more information). For accelerating XGBoost with GPUs, `dask-cuda `__ is recommended for creating GPU clusters. ******** Overview ******** A dask cluster consists of three different components: a centralized scheduler, one or more workers, and one or more clients which act as the user-facing entry point for submitting tasks to the cluster. When using XGBoost with dask, one needs to call the XGBoost dask interface from the client side. Below is a small example which illustrates basic usage of running XGBoost on a dask cluster: .. code-block:: python from xgboost import dask as dxgb import dask.array as da import dask.distributed if __name__ == "__main__": cluster = dask.distributed.LocalCluster() client = dask.distributed.Client(cluster) # X and y must be Dask dataframes or arrays num_obs = 1e5 num_features = 20 X = da.random.random(size=(num_obs, num_features), chunks=(1000, num_features)) y = da.random.random(size=(num_obs, 1), chunks=(1000, 1)) dtrain = dxgb.DaskDMatrix(client, X, y) # or # dtrain = dxgb.DaskQuantileDMatrix(client, X, y) output = dxgb.train( client, {"verbosity": 2, "tree_method": "hist", "objective": "reg:squarederror"}, dtrain, num_boost_round=4, evals=[(dtrain, "train")], ) Here we first create a cluster in single-node mode with :py:class:`distributed.LocalCluster`, then connect a :py:class:`distributed.Client` to this cluster, setting up an environment for later computation. Notice that the cluster construction is guarded by ``__name__ == "__main__"``, which is necessary otherwise there might be obscure errors. We then create a :py:class:`xgboost.dask.DaskDMatrix` object and pass it to :py:func:`xgboost.dask.train`, along with some other parameters, much like XGBoost's normal, non-dask interface. Unlike that interface, ``data`` and ``label`` must be either :py:class:`Dask DataFrame ` or :py:class:`Dask Array ` instances. The primary difference with XGBoost's dask interface is we pass our dask client as an additional argument for carrying out the computation. Note that if client is set to ``None``, XGBoost will use the default client returned by dask. There are two sets of APIs implemented in XGBoost. The first set is functional API illustrated in above example. Given the data and a set of parameters, the ``train`` function returns a model and the computation history as a Python dictionary: .. code-block:: python { "booster": Booster, "history": dict, } For prediction, pass the ``output`` returned by ``train`` into :py:func:`xgboost.dask.predict`: .. code-block:: python prediction = dxgb.predict(client, output, dtrain) # Or equivalently, pass ``output['booster']``: prediction = dxgb.predict(client, output['booster'], dtrain) Eliminating the construction of DaskDMatrix is also possible, this can make the computation a bit faster when meta information like ``base_margin`` is not needed: .. code-block:: python prediction = dxgb.predict(client, output, X) # Use inplace version. prediction = dxgb.inplace_predict(client, output, X) Here ``prediction`` is a dask ``Array`` object containing predictions from model if input is a ``DaskDMatrix`` or ``da.Array``. When putting dask collection directly into the ``predict`` function or using :py:func:`xgboost.dask.inplace_predict`, the output type depends on input data. See next section for details. Alternatively, XGBoost also implements the Scikit-Learn interface with :py:class:`~xgboost.dask.DaskXGBClassifier`, :py:class:`~xgboost.dask.DaskXGBRegressor`, :py:class:`~xgboost.dask.DaskXGBRanker` and 2 random forest variances. This wrapper is similar to the single node Scikit-Learn interface in xgboost, with dask collection as inputs and has an additional ``client`` attribute. See following sections and :ref:`dask-examples` for more examples. ****************** Running prediction ****************** In previous example we used ``DaskDMatrix`` as input to ``predict`` function. In practice, it's also possible to call ``predict`` function directly on dask collections like ``Array`` and ``DataFrame`` and might have better prediction performance. When ``DataFrame`` is used as prediction input, the result is a dask ``Series`` instead of array. Also, there's in-place predict support on dask interface, which can help reducing both memory usage and prediction time. .. code-block:: python # dtrain is the DaskDMatrix defined above. prediction = dxgb.predict(client, booster, dtrain) or equivalently: .. code-block:: python # where X is a dask DataFrame or dask Array. prediction = dxgb.predict(client, booster, X) Also for inplace prediction: .. code-block:: python # where X is a dask DataFrame or dask Array backed by cupy or cuDF. booster.set_param({"device": "cuda"}) prediction = dxgb.inplace_predict(client, booster, X) When input is ``da.Array`` object, output is always ``da.Array``. However, if the input type is ``dd.DataFrame``, output can be ``dd.Series``, ``dd.DataFrame`` or ``da.Array``, depending on output shape. For example, when SHAP-based prediction is used, the return value can have 3 or 4 dimensions , in such cases an ``Array`` is always returned. The performance of running prediction, either using ``predict`` or ``inplace_predict``, is sensitive to number of blocks. Internally, it's implemented using ``da.map_blocks`` and ``dd.map_partitions``. When number of partitions is large and each of them have only small amount of data, the overhead of calling predict becomes visible. On the other hand, if not using GPU, the number of threads used for prediction on each block matters. Right now, xgboost uses single thread for each partition. If the number of blocks on each workers is smaller than number of cores, then the CPU workers might not be fully utilized. One simple optimization for running consecutive predictions is using :py:class:`distributed.Future`: .. code-block:: python dataset = [X_0, X_1, X_2] booster_f = client.scatter(booster, broadcast=True) futures = [] for X in dataset: # Here we pass in a future instead of concrete booster shap_f = dxgb.predict(client, booster_f, X, pred_contribs=True) futures.append(shap_f) results = client.gather(futures) This is only available on functional interface, as the Scikit-Learn wrapper doesn't know how to maintain a valid future for booster. To obtain the booster object from Scikit-Learn wrapper object: .. code-block:: python cls = dxgb.DaskXGBClassifier() cls.fit(X, y) booster = cls.get_booster() ******************************** Scikit-Learn Estimator Interface ******************************** As mentioned previously, there's another interface that mimics the scikit-learn estimators with higher level of of abstraction. The interface is easier to use compared to the functional interface but with more constraints. It's worth mentioning that, although the interface mimics scikit-learn estimators, it doesn't work with normal scikit-learn utilities like ``GridSearchCV`` as scikit-learn doesn't understand distributed dask data collection. .. code-block:: python from distributed import LocalCluster, Client from xgboost import dask as dxgb def main(client: Client) -> None: X, y = load_data() clf = dxgb.DaskXGBClassifier(n_estimators=100, tree_method="hist") clf.client = client # assign the client clf.fit(X, y, eval_set=[(X, y)]) proba = clf.predict_proba(X) if __name__ == "__main__": with LocalCluster() as cluster: with Client(cluster) as client: main(client) **************** GPU acceleration **************** For most of the use cases with GPUs, the `Dask-CUDA `__ project should be used to create the cluster, which automatically configures the correct device ordinal for worker processes. As a result, users should NOT specify the ordinal (good: ``device=cuda``, bad: ``device=cuda:1``). See :ref:`sphx_glr_python_dask-examples_gpu_training.py` and :ref:`sphx_glr_python_dask-examples_sklearn_gpu_training.py` for worked examples. *************************** Working with other clusters *************************** Using Dask's ``LocalCluster`` is convenient for getting started quickly on a local machine. Once you're ready to scale your work, though, there are a number of ways to deploy Dask on a distributed cluster. You can use `Dask-CUDA `_, for example, for GPUs and you can use Dask Cloud Provider to `deploy Dask clusters in the cloud `_. See the `Dask documentation for a more comprehensive list `__. In the example below, a ``KubeCluster`` is used for `deploying Dask on Kubernetes `_: .. code-block:: python from dask_kubernetes.operator import KubeCluster # Need to install the ``dask-kubernetes`` package from dask_kubernetes.operator.kubecluster.kubecluster import CreateMode from dask.distributed import Client from xgboost import dask as dxgb import dask.array as da def main(): '''Connect to a remote kube cluster with GPU nodes and run training on it.''' m = 1000 n = 10 kWorkers = 2 # assuming you have 2 GPU nodes on that cluster. # You need to work out the worker-spec yourself. See document in dask_kubernetes for # its usage. Here we just want to show that XGBoost works on various clusters. # See notes below for why we use pre-allocated cluster. with KubeCluster( name="xgboost-test", image="my-image-name:latest", n_workers=kWorkers, create_mode=CreateMode.CONNECT_ONLY, shutdown_on_close=False, ) as cluster: with Client(cluster) as client: X = da.random.random(size=(m, n), chunks=100) y = X.sum(axis=1) regressor = dxgb.DaskXGBRegressor(n_estimators=10, missing=0.0) regressor.client = client regressor.set_params(tree_method='hist', device="cuda") regressor.fit(X, y, eval_set=[(X, y)]) if __name__ == '__main__': # Launch the kube cluster on somewhere like GKE, then run this as client process. # main function will connect to that cluster and start training xgboost model. main() Different cluster classes might have subtle differences like network configuration, or specific cluster implementation might contains bugs that we are not aware of. Open an issue if such case is found and there's no documentation on how to resolve it in that cluster implementation. An interesting aspect of the Kubernetes cluster is that the pods may become available after the Dask workflow has begun, which can cause issues with distributed XGBoost since XGBoost expects the nodes used by input data to remain unchanged during training. To use Kubernetes clusters, it is necessary to wait for all the pods to be online before submitting XGBoost tasks. One can either create a wait function in Python or simply pre-allocate a cluster with k8s tools (like ``kubectl``) before running dask workflows. To pre-allocate a cluster, we can first generate the cluster spec using dask kubernetes: .. code-block:: python import json from dask_kubernetes.operator import make_cluster_spec spec = make_cluster_spec(name="xgboost-test", image="my-image-name:latest", n_workers=16) with open("cluster-spec.json", "w") as fd: json.dump(spec, fd, indent=2) .. code-block:: sh kubectl apply -f ./cluster-spec.json Check whether the pods are available: .. code-block:: sh kubectl get pods Once all pods have been initialized, the Dask XGBoost workflow can be run, as in the previous example. It is important to ensure that the cluster sets the parameter ``create_mode=CreateMode.CONNECT_ONLY`` and optionally ``shutdown_on_close=False`` if you do not want to shut down the cluster after a single job. ******* Threads ******* XGBoost has built in support for parallel computation through threads by the setting ``nthread`` parameter (``n_jobs`` for scikit-learn). If these parameters are set, they will override the configuration in Dask. For example: .. code-block:: python with dask.distributed.LocalCluster(n_workers=7, threads_per_worker=4) as cluster: There are 4 threads allocated for each dask worker. Then by default XGBoost will use 4 threads in each process for training. But if ``nthread`` parameter is set: .. code-block:: python output = dxgb.train( client, {"verbosity": 1, "nthread": 8, "tree_method": "hist"}, dtrain, num_boost_round=4, evals=[(dtrain, "train")], ) XGBoost will use 8 threads in each training process. ******************** Working with asyncio ******************** .. versionadded:: 1.2.0 XGBoost's dask interface supports the new :py:mod:`asyncio` in Python and can be integrated into asynchronous workflows. For using dask with asynchronous operations, please refer to `this dask example `_ and document in `distributed `_. To use XGBoost's Dask interface asynchronously, the ``client`` which is passed as an argument for training and prediction must be operating in asynchronous mode by specifying ``asynchronous=True`` when the ``client`` is created (example below). All functions (including ``DaskDMatrix``) provided by the functional interface will then return coroutines which can then be awaited to retrieve their result. Please note that XGBoost is a compute-bounded application, where parallelism is more important than concurrency. The support for `asyncio` is more about compatibility instead of performance gain. Functional interface: .. code-block:: python async with dask.distributed.Client(scheduler_address, asynchronous=True) as client: X, y = generate_array() m = await dxgb.DaskDMatrix(client, X, y) output = await dxgb.train(client, {}, dtrain=m) with_m = await dxgb.predict(client, output, m) with_X = await dxgb.predict(client, output, X) inplace = await dxgb.inplace_predict(client, output, X) # Use ``client.compute`` instead of the ``compute`` method from dask collection print(await client.compute(with_m)) While for the Scikit-Learn interface, trivial methods like ``set_params`` and accessing class attributes like ``evals_result()`` do not require ``await``. Other methods involving actual computation will return a coroutine and hence require awaiting: .. code-block:: python async with dask.distributed.Client(scheduler_address, asynchronous=True) as client: X, y = generate_array() regressor = await dxgb.DaskXGBRegressor(verbosity=1, n_estimators=2) regressor.set_params(tree_method='hist') # trivial method, synchronous operation regressor.client = client # accessing attribute, synchronous operation regressor = await regressor.fit(X, y, eval_set=[(X, y)]) prediction = await regressor.predict(X) # Use `client.compute` instead of the `compute` method from dask collection print(await client.compute(prediction)) ***************************** Evaluation and Early Stopping ***************************** .. versionadded:: 1.3.0 The Dask interface allows the use of validation sets that are stored in distributed collections (Dask DataFrame or Dask Array). These can be used for evaluation and early stopping. To enable early stopping, pass one or more validation sets containing ``DaskDMatrix`` objects. .. code-block:: python import dask.array as da from xgboost import dask as dxgb num_rows = 1e6 num_features = 100 num_partitions = 10 rows_per_chunk = num_rows / num_partitions data = da.random.random( size=(num_rows, num_features), chunks=(rows_per_chunk, num_features) ) labels = da.random.random( size=(num_rows, 1), chunks=(rows_per_chunk, 1) ) X_eval = da.random.random( size=(num_rows, num_features), chunks=(rows_per_chunk, num_features) ) y_eval = da.random.random( size=(num_rows, 1), chunks=(rows_per_chunk, 1) ) dtrain = dxgb.DaskDMatrix( client=client, data=data, label=labels ) dvalid = dxgb.DaskDMatrix( client=client, data=X_eval, label=y_eval ) result = dxgb.train( client=client, params={ "objective": "reg:squarederror", }, dtrain=dtrain, num_boost_round=10, evals=[(dvalid, "valid1")], early_stopping_rounds=3 ) When validation sets are provided to :py:func:`xgboost.dask.train` in this way, the model object returned by :py:func:`xgboost.dask.train` contains a history of evaluation metrics for each validation set, across all boosting rounds. .. code-block:: python print(result["history"]) # {'valid1': OrderedDict([('rmse', [0.28857, 0.28858, 0.288592, 0.288598])])} If early stopping is enabled by also passing ``early_stopping_rounds``, you can check the best iteration in the returned booster. .. code-block:: python booster = result["booster"] print(booster.best_iteration) best_model = booster[: booster.best_iteration] ******************* Other customization ******************* XGBoost dask interface accepts other advanced features found in single node Python interface, including callback functions, custom evaluation metric and objective: .. code-block:: python def eval_error_metric(predt, dtrain: xgb.DMatrix): label = dtrain.get_label() r = np.zeros(predt.shape) gt = predt > 0.5 r[gt] = 1 - label[gt] le = predt <= 0.5 r[le] = label[le] return 'CustomErr', np.sum(r) # custom callback early_stop = xgb.callback.EarlyStopping( rounds=early_stopping_rounds, metric_name="CustomErr", data_name="Train", save_best=True, ) booster = dxgb.train( client, params={ "objective": "binary:logistic", "eval_metric": ["error", "rmse"], "tree_method": "hist", }, dtrain=D_train, evals=[(D_train, "Train"), (D_valid, "Valid")], feval=eval_error_metric, # custom evaluation metric num_boost_round=100, callbacks=[early_stop], ) ********************** Hyper-parameter tuning ********************** See https://github.com/coiled/dask-xgboost-nyctaxi for a set of examples of using XGBoost with dask and optuna. .. _ltr-dask: **************** Learning to Rank **************** .. versionadded:: 3.0.0 .. note:: Position debiasing is not yet supported. There are two operation modes in the Dask learning to rank for performance reasons. The difference is whether a distributed global sort is needed. Please see :ref:`ltr-dist` for how ranking works with distributed training in general. Below we will discuss some of the Dask-specific features. First, if you use the :py:class:`~xgboost.dask.DaskQuantileDMatrix` interface or the :py:class:`~xgboost.dask.DaskXGBRanker` with ``allow_group_split`` set to ``True``, XGBoost will try to sort and group the samples for each worker based on the query ID. This mode tries to skip the global sort and sort only worker-local data, and hence no inter-worker data shuffle. Please note that even worker-local sort is costly, particularly in terms of memory usage as there's no spilling when :py:meth:`~pandas.DataFrame.sort_values` is used, and we need to concatenate the data. XGBoost first checks whether the QID is already sorted before actually performing the sorting operation. One can choose this if the query groups are relatively consecutive, meaning most of the samples within a query group are close to each other and are likely to be resided to the same worker. Don't use this if you have performed a random shuffle on your data. If the input data is random, then there's no way we can guarantee most of data within the same group being in the same worker. For large query groups, this might not be an issue. But for small query groups, it's possible that each worker gets only one or two samples from their group for all groups, which can lead to disastrous performance. In that case, we can partition the data according to query group, which is the default behavior of the :py:class:`~xgboost.dask.DaskXGBRanker` unless the ``allow_group_split`` is set to ``True``. This mode performs a sort and a groupby on the entire dataset in addition to an encoding operation for the query group IDs. Along with partition fragmentation, this option can lead to slow performance. See :ref:`sphx_glr_python_dask-examples_dask_learning_to_rank.py` for a worked example. .. _tracker-ip: *************** Troubleshooting *************** - In some environments XGBoost might fail to resolve the IP address of the scheduler, a symptom is user receiving ``OSError: [Errno 99] Cannot assign requested address`` error during training. A quick workaround is to specify the address explicitly. To do that the collective :py:class:`~xgboost.collective.Config` is used: .. versionadded:: 3.0.0 .. code-block:: python import dask from distributed import Client from xgboost import dask as dxgb from xgboost.collective import Config # let xgboost know the scheduler address coll_cfg = Config(retry=1, timeout=20, tracker_host_ip="10.23.170.98", tracker_port=0) with Client(scheduler_file="sched.json") as client: reg = dxgb.DaskXGBRegressor(coll_cfg=coll_cfg) - Please note that XGBoost requires a different port than dask. By default, on a unix-like system XGBoost uses the port 0 to find available ports, which may fail if a user is running in a restricted docker environment. In this case, please open additional ports in the container and specify it as in the above snippet. - If you encounter a NCCL system error while training with GPU enabled, which usually includes the error message `NCCL failure: unhandled system error`, you can specify its network configuration using one of the environment variables listed in the `NCCL document `__ such as the ``NCCL_SOCKET_IFNAME``. In addition, you can use ``NCCL_DEBUG`` to obtain debug logs. - If NCCL fails to initialize in a container environment, it might be caused by limited system shared memory. With docker, one can try the flag: `--shm-size=4g`. - MIG (Multi-Instance GPU) is not yet supported by NCCL. You will receive an error message that includes `Multiple processes within a communication group ...` upon initialization. .. _nccl-load: - Starting from version 2.1.0, to reduce the size of the binary wheel, the XGBoost package (installed using pip) loads NCCL from the environment instead of bundling it directly. This means that if you encounter an error message like "Failed to load nccl ...", it indicates that NCCL is not installed or properly configured in your environment. To resolve this issue, you can install NCCL using pip: .. code-block:: sh pip install nvidia-nccl-cu12 # (or with any compatible CUDA version) The default conda installation of XGBoost should not encounter this error. If you are using a customized XGBoost, please make sure one of the followings is true: + XGBoost is NOT compiled with the `USE_DLOPEN_NCCL` flag. + The `dmlc_nccl_path` parameter is set to full NCCL path when initializing the collective. Here are some additional tips for troubleshooting NCCL dependency issues: + Check the NCCL installation path and verify that it's installed correctly. We try to find NCCL by using ``from nvidia.nccl import lib`` in Python when XGBoost is installed using pip. + Ensure that you have the correct CUDA version installed. NCCL requires a compatible CUDA version to function properly. + If you are not using distributed training with XGBoost and yet see this error, please open an issue on GitHub. + If you continue to encounter NCCL dependency issues, please open an issue on GitHub. ************ IPv6 Support ************ .. versionadded:: 1.7.0 XGBoost has initial IPv6 support for the dask interface on Linux. Due to most of the cluster support for IPv6 is partial (dual stack instead of IPv6 only), we require additional user configuration similar to :ref:`tracker-ip` to help XGBoost obtain the correct address information: .. code-block:: python import dask from distributed import Client from xgboost import dask as dxgb # let xgboost know the scheduler address, use the same bracket format as dask. with dask.config.set({"xgboost.scheduler_address": "[fd20:b6f:f759:9800::]"}): with Client("[fd20:b6f:f759:9800::]") as client: reg = dxgb.DaskXGBRegressor(tree_method="hist") When GPU is used, XGBoost employs `NCCL `_ as the underlying communication framework, which may require some additional configuration via environment variable depending on the setting of the cluster. Please note that IPv6 support is Unix only. ****************************** Logging the evaluation results ****************************** By default, the Dask interface prints evaluation results in the scheduler process. This makes it difficult for a user to monitor training progress. We can define custom evaluation monitors using callback functions. See :ref:`sphx_glr_python_dask-examples_forward_logging.py` for a worked example on how to forward the logs to the client process. In the example, there are two potential solutions using Dask builtin methods, including :py:meth:`distributed.Client.forward_logging` and :py:func:`distributed.print`. Both of them have some caveats but can be a good starting point for developing more sophisticated methods like writing to files. ***************************************************************************** Why is the initialization of ``DaskDMatrix`` so slow and throws weird errors ***************************************************************************** The dask API in XGBoost requires construction of ``DaskDMatrix``. With the Scikit-Learn interface, ``DaskDMatrix`` is implicitly constructed for all input data during the ``fit`` or ``predict`` steps. You might have observed that ``DaskDMatrix`` construction can take large amounts of time, and sometimes throws errors that don't seem to be relevant to ``DaskDMatrix``. Here is a brief explanation for why. By default most dask computations are `lazily evaluated `_, which means that computation is not carried out until you explicitly ask for a result by, for example, calling ``compute()``. See the previous link for details in dask, and `this wiki `_ for information on the general concept of lazy evaluation. The ``DaskDMatrix`` constructor forces lazy computations to be evaluated, which means it's where all your earlier computation actually being carried out, including operations like ``dd.read_csv()``. To isolate the computation in ``DaskDMatrix`` from other lazy computations, one can explicitly wait for results of input data before constructing a ``DaskDMatrix``. Also dask's `diagnostics dashboard `_ can be used to monitor what operations are currently being performed. ******************* Reproducible Result ******************* In a single node mode, we can always expect the same training result between runs as along as the underlying platforms are the same. However, it's difficult to obtain reproducible result in a distributed environment, since the tasks might get different machine allocation or have different amount of available resources during different sessions. There are heuristics and guidelines on how to achieve it but no proven method for guaranteeing such deterministic behavior. The Dask interface in XGBoost tries to provide reproducible result with best effort. This section highlights some known criteria and try to share some insights into the issue. There are primarily two different tasks for XGBoost the carry out, training and inference. Inference is reproducible given the same software and hardware along with the same run-time configurations. The remaining of this section will focus on training. Many of the challenges come from the fact that we are using approximation algorithms, The sketching algorithm used to find histogram bins is an approximation to the exact quantile algorithm, the `AUC` metric in a distributed environment is an approximation to the exact `AUC` score, and floating-point number is an approximation to real number. Floating-point is an issue as its summation is not associative, meaning :math:`(a + b) + c` does not necessarily equal to :math:`a + (b + c)`, even though this property holds true for real number. As a result, whenever we change the order of a summation, the result can differ. This imposes the requirement that, in order to have reproducible output from XGBoost, the entire pipeline needs to be reproducible. - The software stack is the same for each runs. This goes without saying. XGBoost might generate different outputs between different versions. This is expected as we might change the default value of hyper-parameter, or the parallel strategy that generates different floating-point result. We guarantee the correctness the algorithms, but there are lots of wiggle room for the final output. The situation is similar for many dependencies, for instance, the random number generator might differ from platform to platform. - The hardware stack is the same for each runs. This includes the number of workers, and the amount of available resources on each worker. XGBoost can generate different results using different number of workers. This is caused by the approximation issue mentioned previously. - Similar to the hardware constraint, the network topology is also a factor in final output. If we change topology the workers might be ordered differently, leading to different ordering of floating-point operations. - The random seed used in various place of the pipeline. - The partitioning of data needs to be reproducible. This is related to the available resources on each worker. Dask might partition the data differently for each run according to its own scheduling policy. For instance, if there are some additional tasks in the cluster while you are running the second training session for XGBoost, some of the workers might have constrained memory and Dask may not push the training data for XGBoost to that worker. This change in data partitioning can lead to different output models. If you are using a shared Dask cluster, then the result is likely to vary between runs. - The operations performed on dataframes need to be reproducible. There are some operations like `DataFrame.merge` not being deterministic on parallel hardwares like GPU where the order of the index might differ from run to run. It's expected to have different results when training the model in a distributed environment than training the model using a single node due to aforementioned criteria. ************ Memory Usage ************ Here are some practices on reducing memory usage with dask and xgboost. - In a distributed work flow, data is best loaded by dask collections directly instead of loaded by client process. When loading with client process is unavoidable, use ``client.scatter`` to distribute data from client process to workers. See [2] for a nice summary. - When using GPU input, like dataframe loaded by ``dask_cudf``, you can try :py:class:`xgboost.dask.DaskQuantileDMatrix` as a drop in replacement for ``DaskDMatrix`` to reduce overall memory usage. See :ref:`sphx_glr_python_dask-examples_gpu_training.py` for an example. - Use in-place prediction when possible. References: #. https://github.com/dask/dask/issues/6833 #. https://stackoverflow.com/questions/45941528/how-to-efficiently-send-a-large-numpy-array-to-the-cluster-with-dask-array