- Sort Score
- Result 10 results
- Languages All
Results 151 - 160 of 245 for tpu0 (0.04 sec)
-
tensorflow/compiler/mlir/tf2xla/internal/passes/clustering_passes.h
// is faster. std::unique_ptr<mlir::OperationPass<mlir::func::FuncOp>> CreateXlaBroadcastPass(); // Creates a pass that identifies XLASharding ops in launch op for TPU // computation. std::unique_ptr<mlir::OperationPass<mlir::ModuleOp>> CreateTPUShardingIdentificationPass(); #define GEN_PASS_REGISTRATION #define GEN_PASS_DECL_MARKOPSFOROUTSIDECOMPILATIONPASS
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Tue Apr 30 02:01:13 UTC 2024 - 3.5K bytes - Viewed (0) -
tensorflow/compiler/mlir/tensorflow/tests/rewrite_tpu_embedding_ops.mlir
// RUN: tf-opt -tf-rewrite-tpu-embedding-ops %s | FileCheck %s // CHECK-LABEL: func @recv_tpu_embedding_activations func.func @recv_tpu_embedding_activations() -> (tensor<512x256xf32>) { // CHECK: %[[DATA:.*]] = "tf.XlaRecvTPUEmbeddingDeduplicationData"() <{config = {{.*}}}> : () -> tensor<!tf_type.variant> // CHECK: %[[RESULT:.*]] = "tf.XlaRecvTPUEmbeddingActivations"(%[[DATA]]) <{config = {{.*}}}> : (tensor<!tf_type.variant>) -> tensor<512x256xf32>
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Mon Oct 30 06:52:55 UTC 2023 - 4.2K bytes - Viewed (0) -
tensorflow/compiler/mlir/tf2xla/api/v2/cluster_tf.h
// is_supported_by_replicated_brige - If the graph targets the replicated // bridge. Set it to true for replicated/partitioned graphs. e.g. replicated // and single-core TPU graphs. Set this to false if the graph is not // replicated, e.g. CPU/GPU graphs. is_in_fallback_enabled_mode - Whether this // was called with fallback to the non-MLIR Bridge. This is just for logging
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Fri Feb 16 23:11:04 UTC 2024 - 2.9K bytes - Viewed (0) -
tensorflow/compiler/mlir/tfrt/function/function.cc
// for training. if (absl::StrContains(pass_options.default_device, "CPU")) { pass_options.skip_fold_transpose_in_ops = true; } pass_options.enable_optimizer = options.enable_optimizer; // Use TFRT TPU OpKernel for training. pass_options.target_tpurt = false; pass_options.tpu_use_core_selector = options.tpu_use_core_selector; pass_options.tpu_use_bundled_transfer = options.tpu_use_bundled_transfer;
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Mon Apr 22 08:13:15 UTC 2024 - 3.8K bytes - Viewed (0) -
tensorflow/compiler/mlir/tensorflow/transforms/replicate_to_island.cc
// and logical core (`logical_core`). // `replica_id` is the index of the ancestor ReplicateOp in [0, num_replicas). // `logical_core` is the index of the TPU core in [0, num_cores_per_replica). // `device_ordinal` is the index of the TPU core relative to its host. LogicalResult GetDeviceOrdinal(const std::optional<DictionaryAttr>& devices, const unsigned replica_id,
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Mon Jul 24 21:01:40 UTC 2023 - 16.9K bytes - Viewed (0) -
tensorflow/compiler/mlir/tensorflow/transforms/tpu_resource_partitioning.cc
return str_attr && (str_attr.getValue().find("COMPOSITE") != llvm::StringRef::npos); } } // namespace // Rewrites unpartitioned resource reads and writes to partitioned resource // reads and writes. The TPU computation from the frontend is generated in such // a way that resource operations operate on the unpartitioned resource handle // (from a `tf.TPUReplicatedInput`). This results in resource reads and writes
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Wed Sep 06 19:12:29 UTC 2023 - 11.8K bytes - Viewed (0) -
tensorflow/compiler/mlir/tensorflow/tests/tpu_identity_pruning.mlir
// RUN: tf-opt %s -tf-tpu-identity-pruning | FileCheck %s --dump-input=always // Tests Identity op in cluster is pruned away. // CHECK-LABEL: func @testIdentity // CHECK-SAME: ([[ARG0:%.*]]: tensor<i32>) func.func @testIdentity(%arg0: tensor<i32>) { // CHECK-NOT: "tf.Identity" // CHECK: "tf_device.cluster" // CHECK-NEXT: tf_device.return [[ARG0]] %0 = "tf_device.cluster"() ({ %1 = "tf.Identity"(%arg0) : (tensor<i32>) -> tensor<i32>
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Mon Mar 28 12:06:33 UTC 2022 - 3.5K bytes - Viewed (0) -
tensorflow/compiler/mlir/tensorflow/transforms/executor_tpuv1_inline_tpu_island.cc
#include "tensorflow/compiler/mlir/tensorflow/ir/tf_ops.h" #include "tensorflow/compiler/mlir/tensorflow/transforms/passes.h" #include "tensorflow/compiler/mlir/tensorflow/utils/error_util.h" #define DEBUG_TYPE "tf-executor-tpu-v1-island-inlining" namespace mlir { namespace tf_executor { namespace { constexpr llvm::StringRef kNestedModule = "_tpu_v1_compat_outlined"; #define GEN_PASS_DEF_EXECUTORTPUV1ISLANDINLININGPASS
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Tue Dec 19 08:06:04 UTC 2023 - 4K bytes - Viewed (0) -
tensorflow/compiler/mlir/tensorflow/transforms/canonicalize_compile_and_replicate_attributes.cc
// This transformation pass converts existing compilation and replication // attributes into unified attributes. For example, A _tpu_replicate=X // should be replaced with _xla_compile_device_type=TPU and // _replication_info=X attributes by the conversion. An _XlaMustCompile=true // should be replaced with _xla_compile_device_type with the value of device // attribute. #include "llvm/ADT/StringRef.h"
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Wed Oct 05 23:50:19 UTC 2022 - 4.5K bytes - Viewed (0) -
tensorflow/compiler/mlir/tensorflow/tests/extract_outside_compilation.mlir
module attributes {tf.versions = {producer = 888 : i32}, tf.devices = ["/job:worker/replica:0/task:0/device:CPU:0", "/job:worker/replica:0/task:0/device:TPU_SYSTEM:0", "/job:worker/replica:0/task:0/device:TPU:0"]} { // Tests that TPU cluster with no outside compilation does not generate parallel_execute. // CHECK-LABEL: func @no_outside_compilation func.func @no_outside_compilation() -> tensor<2xi32> { // CHECK-NOT: "tf_device.parallel_execute"
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Tue Oct 31 08:59:10 UTC 2023 - 129.6K bytes - Viewed (0)