Home
last modified time | relevance | path

Searched full:checkpoint_dir (Results 1 – 25 of 65) sorted by relevance

123

/aosp_15_r20/external/tensorflow/tensorflow/python/training/
H A Dcheckpoint_utils_test.py41 def _create_checkpoints(sess, checkpoint_dir): argument
42 checkpoint_prefix = os.path.join(checkpoint_dir, "model")
60 def _create_partition_checkpoints(sess, checkpoint_dir): argument
61 checkpoint_prefix = os.path.join(checkpoint_dir, "model")
84 checkpoint_dir = self.get_temp_dir() + "/no_checkpoints"
87 checkpoint_utils.load_variable(checkpoint_dir, "var1"), [])
90 checkpoint_dir = self.get_temp_dir()
92 _, _, _, _ = _create_checkpoints(session, checkpoint_dir)
95 checkpoint_utils.load_variable(checkpoint_dir, "var5"), [])
98 checkpoint_dir = self.get_temp_dir()
[all …]
H A Dsession_manager_test.py86 checkpoint_dir = os.path.join(self.get_temp_dir(), "prepare_session")
89 gfile.DeleteRecursively(checkpoint_dir)
93 gfile.MakeDirs(checkpoint_dir)
104 checkpoint_dir=checkpoint_dir)
106 checkpoint_filename = os.path.join(checkpoint_dir,
112 os.rename(checkpoint_dir, checkpoint_dir2)
113 gfile.MakeDirs(checkpoint_dir)
127 checkpoint_dir=checkpoint_dir,
131 gfile.DeleteRecursively(checkpoint_dir)
132 os.rename(checkpoint_dir2, checkpoint_dir)
[all …]
H A Devaluation_test.py71 def _train_model(self, checkpoint_dir, num_steps): argument
78 checkpoint_dir: The directory where the checkpoint is written to.
94 checkpoint_dir=checkpoint_dir,
104 checkpoint_dir = os.path.join(self.get_temp_dir(),
108 self._train_model(checkpoint_dir, num_steps=300)
118 checkpoint_path = saver.latest_checkpoint(checkpoint_dir)
130 checkpoint_dir = os.path.join(self.get_temp_dir(),
134 self._train_model(checkpoint_dir, num_steps=300)
150 checkpoint_path = saver.latest_checkpoint(checkpoint_dir)
169 checkpoint_dir = os.path.join(self.get_temp_dir(), 'eval_ops_and_final_ops')
[all …]
H A Dsession_manager.py88 sess = sm.prepare_session(master, init_op, saver, checkpoint_dir)
182 checkpoint_dir=None, argument
193 checkpoint_dir: Path to the checkpoint files. The latest checkpoint in the
205 ValueError: If both checkpoint_dir and checkpoint_filename_with_path are
219 if checkpoint_dir and checkpoint_filename_with_path:
220 raise ValueError("Can not provide both checkpoint_dir and "
224 if not saver or not (checkpoint_dir or checkpoint_filename_with_path):
234 ckpt = checkpoint_management.get_checkpoint_state(checkpoint_dir)
240 ckpt = checkpoint_management.get_checkpoint_state(checkpoint_dir)
254 checkpoint_dir=None, argument
[all …]
H A Dmonitored_session.py318 checkpoint_dir=None, argument
366 summary_dir = summary_dir or checkpoint_dir
397 checkpoint_dir):
401 checkpoint_dir,
409 os.path.join(checkpoint_dir, tmpdir),
419 checkpoint_dir=checkpoint_dir,
431 checkpoint_dir=None, argument
471 checkpoint_dir: A string. Optional path to a directory where to restore
508 summaries. If None, checkpoint_dir is used instead.
510 `checkpoint_dir`. The GraphDef is saved after the session is created as
[all …]
H A Dmonitored_session_test.py294 is_chief=True, checkpoint_dir=logdir) as session:
300 is_chief=True, checkpoint_dir=logdir) as session:
310 checkpoint_dir=logdir,
317 is_chief=True, checkpoint_dir=logdir) as session:
327 checkpoint_dir=logdir,
336 is_chief=True, checkpoint_dir=logdir) as session:
364 checkpoint_dir=test_dir) as session:
373 checkpoint_dir=test_dir) as session:
386 checkpoint_dir=logdir,
404 checkpoint_dir=logdir,
[all …]
H A Dcheckpoint_utils.py122 def wait_for_new_checkpoint(checkpoint_dir, argument
129 checkpoint_dir: The directory in which checkpoints are saved.
140 logging.info("Waiting for new checkpoint at %s", checkpoint_dir)
143 checkpoint_path = checkpoint_management.latest_checkpoint(checkpoint_dir)
154 def checkpoints_iterator(checkpoint_dir, argument
184 checkpoint_dir: The directory in which checkpoints are saved.
199 checkpoint_dir, checkpoint_path, timeout=timeout)
/aosp_15_r20/external/tensorflow/tensorflow/python/distribute/failure_handling/
H A Dgce_failure_handler_test.py63 def _make_checkpoint_manager(checkpoint, checkpoint_dir, cluster_resolver): argument
71 checkpoint, directory=checkpoint_dir, max_to_keep=1)
76 checkpoint_dir, cluster_resolver.task_id),
105 checkpoint_dir, argument
157 fh_ckpt, checkpoint_dir, strategy.cluster_resolver)
161 checkpoint_dir, termination_config))
183 logging.info(gfile.ListDirectory(checkpoint_dir))
186 for a_file in gfile.ListDirectory(checkpoint_dir)
252 checkpoint_dir = os.path.join(self.get_temp_dir(), 'fh_ckpt/')
266 args=(checkpoint_dir, cluster_spec, input_arg, maintenance_event,
[all …]
H A Dfailure_handler_test.py63 def _make_checkpoint_manager(checkpoint, checkpoint_dir, cluster_resolver): argument
70 checkpoint, directory=checkpoint_dir, max_to_keep=1)
75 checkpoint_dir, cluster_resolver.task_id),
117 checkpoint_dir, argument
154 fh_ckpt, checkpoint_dir, strategy.cluster_resolver)
158 checkpoint_dir, termination_config))
190 for a_file in gfile.ListDirectory(checkpoint_dir)
238 checkpoint_dir = os.path.join(self.get_temp_dir(), 'fh_ckpt')
251 args=(checkpoint_dir, cluster_spec, input_arg,
298 self.worker_fn(checkpoint_dir, cluster_spec, input_arg,
[all …]
H A Dfailure_handling.py55 def _non_chief_checkpoint_dir(checkpoint_dir, task_id): argument
57 dirpath = os.path.dirname(checkpoint_dir)
58 base = os.path.basename(checkpoint_dir)
340 that's passed as the `checkpoint_dir` argument. In this way, at the program
365 …f.distribute.experimental.PreemptionCheckpointHandler(cluster_resolver, checkpoint, checkpoint_dir)
394 checkpoint_dir=None, argument
408 a `tf.train.CheckpointManager` to manage it in the `checkpoint_dir`.
409 checkpoint_dir: a directory where the `PreemptionCheckpointHandler` saves
411 created, the latest checkpoint in the `checkpoint_dir` will be restored.
442 checkpoint_lib.Checkpoint) and not checkpoint_dir:
[all …]
/aosp_15_r20/external/pytorch/test/distributed/checkpoint/
H A Dtest_dtensor_resharding.py20 CHECKPOINT_DIR = "checkpoint" variable
53 CHECKPOINT_DIR = self.temp_dir
68 storage_writer=dist_cp.FileSystemWriter(path=CHECKPOINT_DIR),
79 storage_reader=dist_cp.FileSystemReader(CHECKPOINT_DIR),
104 CHECKPOINT_DIR = self.temp_dir
120 storage_writer=dist_cp.FileSystemWriter(path=CHECKPOINT_DIR),
129 storage_reader=dist_cp.FileSystemReader(CHECKPOINT_DIR),
158 CHECKPOINT_DIR = self.temp_dir
170 storage_writer=dist_cp.FileSystemWriter(path=CHECKPOINT_DIR),
185 storage_reader=dist_cp.FileSystemReader(CHECKPOINT_DIR),
[all …]
H A Dtest_tp_checkpoint.py45 CHECKPOINT_DIR = self.temp_dir
62 storage_writer=dcp.FileSystemWriter(CHECKPOINT_DIR),
80 storage_reader=dcp.FileSystemReader(CHECKPOINT_DIR),
92 CHECKPOINT_DIR = self.temp_dir
111 storage_writer=dcp.FileSystemWriter(CHECKPOINT_DIR),
122 storage_reader=dcp.FileSystemReader(CHECKPOINT_DIR),
H A Dtest_hsdp_checkpoint.py79 CHECKPOINT_DIR = self.temp_dir
99 storage_writer=dist_cp.FileSystemWriter(CHECKPOINT_DIR),
118 storage_reader=dist_cp.FileSystemReader(CHECKPOINT_DIR),
138 CHECKPOINT_DIR = self.temp_dir
155 storage_writer=dist_cp.FileSystemWriter(CHECKPOINT_DIR),
187 storage_reader=dist_cp.FileSystemReader(CHECKPOINT_DIR),
H A Dtest_fsdp_optim_state.py59 CHECKPOINT_DIR = self.temp_dir
82 storage_writer=dcp.FileSystemWriter(CHECKPOINT_DIR),
103 storage_reader=dcp.FileSystemReader(CHECKPOINT_DIR),
110 storage_reader=dcp.FileSystemReader(CHECKPOINT_DIR),
/aosp_15_r20/external/tensorflow/tensorflow/python/distribute/
H A Dcheckpoint_utils_test.py38 def _create_checkpoints(sess, checkpoint_dir): argument
39 checkpoint_prefix = os.path.join(checkpoint_dir, "model")
58 checkpoint_dir = self.get_temp_dir()
60 v1, v2 = _create_checkpoints(session, checkpoint_dir)
61 return checkpoint_dir, v1, v2
76 checkpoint_dir, v1_value, v2_value = self._get_test_object()
84 checkpoint_utils.init_from_checkpoint(checkpoint_dir, {
112 checkpoint_dir, v1_value, _ = self._get_test_object()
120 checkpoint_utils.init_from_checkpoint(checkpoint_dir, {
H A Dtpu_strategy_model_parallelism_test.py151 checkpoint_dir = self.get_temp_dir()
152 checkpoint_prefix = os.path.join(checkpoint_dir, "ckpt")
165 checkpoint_management.latest_checkpoint(checkpoint_dir))
334 checkpoint_dir = self.get_temp_dir()
335 checkpoint_prefix = os.path.join(checkpoint_dir, "ckpt")
356 checkpoint_management.latest_checkpoint(checkpoint_dir))
/aosp_15_r20/external/tensorflow/tensorflow/core/kernels/
H A Dcheckpoint_callback_manager.cc49 absl::string_view checkpoint_dir, in TriggerSaveCallbackIfFileNotExist() argument
53 checkpoint_dir, absl::StrCat(checkpoint_id, ".", file_extension)); in TriggerSaveCallbackIfFileNotExist()
84 absl::string_view checkpoint_dir, in TriggerRestoreCallbackIfFileExists() argument
88 checkpoint_dir, absl::StrCat(checkpoint_id, ".", file_extension)); in TriggerRestoreCallbackIfFileExists()
165 std::string checkpoint_dir; in RegisterSaveCallback() local
178 checkpoint_dir = last_saved_checkpoint_id_and_dir_.second; in RegisterSaveCallback()
183 TriggerSaveCallbackIfFileNotExist(checkpoint_id, checkpoint_dir, in RegisterSaveCallback()
199 std::string checkpoint_dir; in RegisterRestoreCallback() local
212 checkpoint_dir = last_restored_checkpoint_id_and_dir_.second; in RegisterRestoreCallback()
217 TriggerRestoreCallbackIfFileExists(checkpoint_id, checkpoint_dir, in RegisterRestoreCallback()
/aosp_15_r20/external/tensorflow/tensorflow/python/checkpoint/
H A Dcheckpoint_management.py249 def get_checkpoint_state(checkpoint_dir, latest_filename=None): argument
256 checkpoint_dir: The directory of checkpoints.
267 if isinstance(checkpoint_dir, os.PathLike):
268 checkpoint_dir = os.fspath(checkpoint_dir)
270 coord_checkpoint_filename = _GetCheckpointFilename(checkpoint_dir,
283 + checkpoint_dir)
285 # prepend checkpoint_dir.
287 ckpt.model_checkpoint_path = os.path.join(checkpoint_dir,
291 ckpt.all_model_checkpoint_paths[i] = os.path.join(checkpoint_dir, p)
327 def latest_checkpoint(checkpoint_dir, latest_filename=None): argument
[all …]
/aosp_15_r20/external/pytorch/torch/distributed/checkpoint/examples/
H A Dfsdp_checkpoint_example.py23 CHECKPOINT_DIR = f"/scratch/{os.environ['LOGNAME']}/checkpoint" variable
71 # Save the model to CHECKPOINT_DIR
80 storage_writer=dist_cp.FileSystemWriter(CHECKPOINT_DIR),
90 # Load model_2 with parameters saved in CHECKPOINT_DIR
99 storage_reader=dist_cp.FileSystemReader(CHECKPOINT_DIR),
106 storage_reader=dist_cp.FileSystemReader(CHECKPOINT_DIR),
125 shutil.rmtree(CHECKPOINT_DIR, ignore_errors=True)
H A Dstateful_example.py23 CHECKPOINT_DIR = f"~/{os.environ['LOGNAME']}/checkpoint" variable
86 checkpoint_id=CHECKPOINT_DIR,
93 checkpoint_id=CHECKPOINT_DIR,
101 shutil.rmtree(CHECKPOINT_DIR, ignore_errors=True)
/aosp_15_r20/external/executorch/examples/models/llama/
H A Dexport_llama_lib.py183 "--checkpoint_dir",
185 … with a sharded checkpoint, not for the standard llama2 model. Note, checkpoint_dir takes preceden…
524 checkpoint_dir = (
525 canonical_path(args.checkpoint_dir) if args.checkpoint_dir else None
543 checkpoint_dir=checkpoint_dir,
850 checkpoint_dir: Optional[str] = None,
878 checkpoint or checkpoint_dir
879 ) and params_path, "Both checkpoint/checkpoint_dir and params can't be empty"
901 checkpoint_dir=checkpoint_dir,
/aosp_15_r20/external/tensorflow/tensorflow/python/keras/distribute/
H A Dworker_training_state.py44 def __init__(self, model, checkpoint_dir): argument
62 # If this is single-worker training, checkpoint_dir are the same for
66 # save checkpoint, we replace the write_checkpoint_manager's checkpoint_dir
71 # But all workers should restore from the same checkpoint_dir as passed in
75 directory=os.path.join(checkpoint_dir, 'chief'),
78 checkpoint_dir, self._model.distribute_strategy)
/aosp_15_r20/device/google/zuma/
Dcopy_efs_files_to_data.sh3 CHECKPOINT_DIR=/data/vendor/copied
7 $BIN_DIR/mkdir -p $CHECKPOINT_DIR
14 tmpdir=$CHECKPOINT_DIR/$partition_name.img
15 build_checkpoint=$CHECKPOINT_DIR/$partition_name
/aosp_15_r20/device/google/zumapro/
Dcopy_efs_files_to_data.sh3 CHECKPOINT_DIR=/data/vendor/copied
7 $BIN_DIR/mkdir -p $CHECKPOINT_DIR
14 tmpdir=$CHECKPOINT_DIR/$partition_name.img
15 build_checkpoint=$CHECKPOINT_DIR/$partition_name
/aosp_15_r20/external/libopus/dnn/torch/lpcnet/
H A Dtrain_lpcnet.py126 checkpoint_dir = os.path.join(args.output, 'checkpoints') variable
127 os.makedirs(checkpoint_dir, exist_ok=True)
253 torch.save(checkpoint, os.path.join(checkpoint_dir, checkpoint_prefix + f'_best.pth'))
256 torch.save(checkpoint, os.path.join(checkpoint_dir, checkpoint_prefix + f'_epoch_{ep}.pth'))
257 torch.save(checkpoint, os.path.join(checkpoint_dir, checkpoint_prefix + f'_last.pth'))

123