| Commit message (Collapse) | Author | Age |
|
|
|
| |
Change: 126374056
|
|
|
|
|
|
| |
saved checkpoints.
Change: 126367935
|
|
|
|
| |
Change: 126366720
|
|
|
|
| |
Change: 126364522
|
|
|
|
|
|
| |
If the batch size is zero, we need to avoid calling into Eigen because Eigen
will explode. Zero classes is an error.
Change: 126359444
|
|
|
|
|
|
|
|
| |
Add GetAttr to shape_inference::InferenceContext.
Allow setting NodeDef in shape_inference_testutil INFER calls (with new
INFER*_WITH_DEF macro). Fix a bug that caused a crash when an INFER..ERROR
macro called a shape inference function that did not return an error.
Change: 126350221
|
|
|
|
| |
Change: 126349886
|
|
|
|
|
|
| |
This change adds the CPU kernel and Python ifaces. For now, assumes both
operands have the same shapes.
Change: 126348349
|
|
|
|
|
|
|
|
|
| |
Fixes #2099.
Tries to give Variables the same behavior as non-Variable tensors in
this respect. Useful for not having to special case e.g. coefficients
of a feature vector which may sometimes not have any features.
Change: 126347791
|
|
|
|
|
|
|
| |
This change has the unfortunate side-effect of preventing the use of
tf.InteractiveSession with a gRPC session. This is intended as a
temporary measure, while we fix the implementation of SimpleGraphExecutionState.
Change: 126344674
|
|
|
|
| |
Change: 126344587
|
|
|
|
|
|
| |
ptb word model from 6800 to 7800 words per second.
Change: 126342788
|
|
|
|
|
|
| |
inserted_count is used by RefCounted which is used by RefCountedVec.
So inserted_count must outlive RefCountedVec.
Change: 126342639
|
|
|
|
| |
Change: 126338283
|
|
|
|
| |
Change: 126335170
|
|
|
|
|
|
|
|
| |
This is the same logic as python, with an addition in the case where 1 value is
unknown and the other is unknown - in this case, we propagate the unknown input
dim instead of a new unknown input dim (this case did not apply in python where
None was used for the unknown input).
Change: 126308395
|
|
|
|
| |
Change: 126263834
|
|
|
|
| |
Change: 126255634
|
|
|
|
| |
Change: 126246458
|
|
|
|
| |
Change: 126219121
|
|
|
|
|
| |
Update the minimum required Bazel version to 0.3.0 which includes the bugfix.
Change: 126189429
|
|
|
|
|
|
|
| |
core tensorflow. This is done by moving the core of concat_lib_cpu.cc into a
new .h, and making it templated on a struct that defines the function to copy
a range of elements.
Change: 126147862
|
|
|
|
|
|
|
| |
- Proto definition for configuration
- Utility for converting from proto
- Visibility change
Change: 126145305
|
|
|
|
| |
Change: 126125286
|
|
|
|
| |
Change: 126119692
|
|
|
|
| |
Change: 126091563
|
|
|
|
| |
Change: 126086117
|
|
|
|
|
|
| |
fill_functor.h to its own compilation unit in fill_functor.cc.
Change: 126081539
|
|
|
|
|
| |
Useful to identify execution stages when looking at logs.
Change: 126022244
|
|
|
|
| |
Change: 126013328
|
|
|
|
|
|
| |
handle per-thread buffer allocation for the tileable executor without resorting to thread_local that is not fully supported on Android.
Change: 126009029
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
done by models distributed across many devices. A small
microbenchmark model that runs two banks (A and B) of 30 nodes with a
30x30 full shuffle between them, where each of the nodes in A and in B
run with one node on each of the 30 devices (so 30*29+30+30, or ~930
separate RPCs) was showing ~111,000 allocations per iteration of the graph.
With the changes here, this is now down to ~64,300 allocations per iteration.
Changes include:
o DeviceContext::CopyDeviceTensorToCPU and related helper routines:
use StringPiece instead of const string& for the tensor name (avoids
creating a string in some cases where the caller only has a
StringPiece available).
o Change some Rendezvous and BaseRemoteRendezvous interfaces to
take a 'const Rendezvous::ParsedKey& key', rather than 'const string& key'.
In many cases, the callers were already having to parse the key
into a ParsedKey, and so we were doing the parsing multiple times at
different levels as we processed receiving or sending of a tensor. This
reduces the number of times that we parse a key as it flows from a Send
node through to a Recv node on another worker.
o Changed Rendezvous::ParsedKey so that it makes a copy of the underlying
full key, and then uses StringPiece objects to point into this copy for
the src_device, dst_device, and edge_name pieces. This turns 3 string
allocations into 1 per Rendezvous::ParseKey call.
o Added new StringPiece Rendezvous::ParsedKey::FullKey() accessor to
return a StringPiece for the underlying full key, and used that in a
few places (mostly logging) where that is useful.
o In many places, used std::move(function_variable) when assigning to
an instance variable. This eliminates a very large number of excess
std::function allocations/initializations (~56000 of the baseline
allocations were related to std::function setup or cloning, and this
is now down to ~11000 after this cl).
o In the RPC-based remote workers (StubbyRemoteWorker and
GrpcRemoteWorker), changed the code path in RecvTensorAsync to avoid
creation of a std::function with 6 arguments unless necessary. There
are three cases now handled separately:
(a) We're not logging, and we didn't make a copy of the request that we
need to free: just use the passed in 'StatusCallback done' object
directly, without creating a wrapper std::function object at all
(b) We're not logging, but we made a copy of the request that we
need to free: we create a simple wrapper std::function that
invokes the passed in 'done' callback, and then frees the
req_copy request copy object.
(c) We're logging: we create the std::function object with all the
necessary state to log when the recv has finished.
o Changed DeviceMgr::LookupDevice to take a StringPiece, rather than a
const string&, and changed the hash table to use StringPiece keys.
This allows clients that just have a StringPiece device name in their
hand to avoid a string creation to lookup the Device* object.
o Changed ExecutorState to use a specialized TaggedNodeReadyQueue that
internally uses a gtl::InlinedVector<TaggedNode, 16>, rather than
using a std::deque<TaggedNode> for keeping track of nodes ready to
execute. This is faster because it avoids allocations entirely if the
ready node queue doesn't get bigger than 16, and inlined vectors are
generally faster than std::deque, at a minor risk of using more memory
if this queue grows to very large numbers of ready nodes (mostly imaginable
only in pathological graphs).
o In ExecutorState::Process, allocated a single ExecutorState::AsyncState
object to keep track of all the state we need to preserve for an asynchronously
executed node, rather than keeping this state implicitly via a very large
number of arguments to a lamda function.
o Added new atomic std::atomic<bool> status_is_ok_ in
BaseRemoteRendezvous. This allows us to avoid acquiring the lock when
we just want to check if the status is non-OK in
BaseRemoteRendezvous::Send and BaseRemoteRendezvous::ValidateDevices.
o In GraphMgr::RunAllDone, changed assignment of args.runner to avoid
one extra level of std::function indirection (binding the function directly
to the ThreadPool::Schedule routine, rather than creating an intermediate
lambda function that invokes this inside the body of the lambda.
o Added freelist of RpcRecvTensorCall objects in
third_party/tensorflow/core/distributed_runtime/rpc/rpc_rendezvous_mgr.cc
o Changed third_party/tensorflow/core/framework/rendezvous.cc to keep the
hashtable of Item* objects keyed by uint64 (hash of the tensor name), rather
than the full-string tensor name. Collisions in the 64-bit hash space
should basically never happen.
o Sped up DeviceNameUtils::ParseFullName by optimizing for the common
ordering of parts of /job, /replica, /task, /device. The parsing code
was general enough to handle any order, but did so by comparing the
prefixes 4, 3, 2, and 1 times, respectively, rather than 1, 1, 1, and 1 times.
o Sped up DeviceNameUtils::SplitDeviceName to avoid extra string copies.
Change: 125991891
|
|
|
|
| |
Change: 125975221
|
|
|
|
| |
Change: 125964943
|
|
|
|
| |
Change: 125963264
|
|
|
|
| |
Change: 125961539
|
|
|
|
| |
Change: 125901975
|
|
|
|
|
| |
Add support for exporting the contents of a table.
Change: 125901929
|
|
|
|
|
|
|
| |
This is the first of a series of CLs aimed at implementing the TF debugger (tfdb).
This C++ CL adds the node-outputs callback to ProcessOutputs() in ExecutorImpl and provides access to it via the DebugGateway class. This makes it possible to observe intermediate node outputs during a DirectSession.Run() call.
Change: 125882979
|
|
|
|
| |
Change: 125838604
|
|
|
|
|
|
| |
corner case that an op has multiple inputs, one of the input is of ref type from a different device, and one other input is also remote but cached for some other reasons.
Change: 125837438
|
|
|
|
| |
Change: 125837171
|
|
|
|
| |
Change: 125835079
|
|
|
|
| |
Change: 125829994
|
|
|
|
|
|
| |
fill_functor. This is needed because the templates in fill_functor.h are instantiated in constant_op.cc.
Change: 125816082
|
|
|
|
|
|
|
|
|
| |
Added Python interfaces to reset resource containers.
pywrap_tensorflow.TF_Reset(target, [containers]) will release resources
in all listed containers, while pywrap_tensorflow.TF_Reset(target) will
release resources in all the containers. In particular, Variables cached
on the devices will be cleared by this call.
Change: 125790643
|
|
|
|
| |
Change: 125790131
|
|
|
|
|
|
|
|
|
| |
StridedSliceGrad op implements the gradient of StridedSlice.
Also implement python benchmark for StridedSlice and simple Slice.
Fix bugs in special case optimizations in StridedSlice.
(Toward resolving bug #206)
Change: 125789921
|
|
|
|
|
|
| |
- Prevent unnecessary tensorflow.Example copies when this method
is used in serving.
Change: 125788885
|
|
|
|
|
|
|
| |
- adds support of paths pointing to the root of a bucket, e.g gs://bucket
- adds pagination support for GetChildren
- makes a more optimal HTTP request in GetChildren
Change: 125785863
|