core
core copied to clipboard
[FEA] Add compute_with_dask utility for down-stream Merlin libriaries
Proposal
I propose that we add a single compute_dask_object
(or compute_with_dask
) utility to merlin.core.utils
, and use that utility for all Dask computation within Merlin. The purpose of this utility would be to check the global_dask_client()
utility, and utilize the appropriate default client/scheduler for Merlin (and not the default for dask/distributed).
More Background
While starting to look into merlin-models#339, I noticed that there is at least one place where Merlin-models uses a bare compute()
statement to compute a Dask collection. I suspect that this is also done in several other places across the Merlin ecosystem.
When there is no global Dask client in the current python context, using compute()
will typically result in execution with Dask's "multi-threaded" scheduler. This may be fine for CPU-backed data, but will result in many python threads thrashing the same GPU (device 0) when the data is GPU backed.
For compute
operations in NVTabular (which only operate on Delayed
Dask objects), the merlin.core.utils.global_dask_client
utility is used to query to current Dask client. If this function returns None
, the convention is to use the "sychronous" scheduler (compute(scheduler="sychronous")
), otherwise the distributed client is used. I propose that this same convention be used everywhere in Merlin (besides special cases where scheduler="synchronous"
can be hard coded.
Note that these changes are also required for Merlin's Serial
and Distributed
context managers to work correctly.
Proposed Implementation
from merlin.core.utils import global_dask_client
from dask.base import is_dask_collection
from dask.delayed import Delayed
def compute_dask_object(dask_obj):
"""Compute a Dask collection using Merlin's dask-client settings"""
# Check global client
dask_client = global_dask_client()
if is_dask_collection(dask_obj) or isinstance(dask_obj, Delayed):
# Compute simple Dask collections
# (Use distributed client, or fall back to "synchronous" scheduler)
scheduler = dask_client.get if dask_client else "synchronous"
return dask_obj.compute(scheduler=scheduler)
elif isinstance(dask_obj, list):
# Maybe check that all elements of list are collections or Delayed?
# Compute entire list at once:
if dask_client:
return [r.result() for r in dask_client.compute(dask_obj)]
else:
return dask.compute(dask_obj, scheduler="synchronous")[0]
else:
raise ValueError
This looks good! My only comment here is that I wonder if this function might belong in the io
package, which is where most of the Dask-related functionality currently lives.
This looks good! My only comment here is that I wonder if this function might belong in the io package, which is where most of the Dask-related functionality currently lives.
Computing a Dask object/graph is definitely used extensively outside IO, so my gut feeling is "no". Also, all other utilities related to distributed/serial Dask execution are already defined in merlin.core.utils
. That said, I'm open to re-organizing these utilities in some other way if you have something cleaner in mind.
Probably fine to add it there for now then. I would like to organize the utils into other more clearly named sub-packages, but that's out of scope for this issue.
Thinking about it, I wonder if having a dask
sub-package outside of utils
would make sense? Maybe the thing to do is migrate parts of io
that are for Dask but not specific to IO into a new package along with this util.
Thinking about it, I wonder if having a dask sub-package outside of utils would make sense? Maybe the thing to do is migrate parts of io that are for Dask but not specific to IO into a new package along with this util.
Sure - Do you have an intuition for what would qualify a piece of code as something that belongs under dask/
? Would it be code that explicitly imports a dask-related package? Most of the code in io
uses dask directly, or is called by dask-specific code, so finding an intuitive division may be a bit tricky (but is probably doable).
I don't have a clear line in mind, but having looked at what's in io
, it seemed like DaskSubgraph
, DataFrameIter
, and shuffle.py
are pure Dask code that might be applicable outside of io
. (I could be totally wrong about this too though, thinking out loud here.)
I don't have a clear line in mind, but having looked at what's in io, it seemed like DaskSubgraph, DataFrameIter, and shuffle.py are pure Dask code that might be applicable outside of io. (I could be totally wrong about this too though, thinking out loud here.)
Okay - I definitely agree that it makes sense to add a new merlin module for dask-specific code (as long as we all understand that a lot of dask-related code will still need to live in many other places. Perhaps I'll try to throw up a PR with a prospective re-org today or tomorrow.