MCPcopy Create free account
hub / github.com/DeepRec-AI/DeepRec / freeze

Method freeze

tensorflow/python/tpu/tpu_feed.py:434–458  ·  view source on GitHub ↗

Freezes the InfeedQueue so it can no longer be modified. The configuration is implicitly frozen before any host-side or device-side Ops are generated. The configuration cannot be frozen until the types and shapes of the tuple elements have been set. Raises: ValueError: if the

(self)

Source from the content-addressed store, hash-verified

432 self.set_tuple_types([t.dtype for t in input_tensors[0]])
433
434 def freeze(self):
435 """Freezes the InfeedQueue so it can no longer be modified.
436
437 The configuration is implicitly frozen before any host-side or
438 device-side Ops are generated. The configuration cannot be frozen
439 until the types and shapes of the tuple elements have been set.
440
441 Raises:
442 ValueError: if the types or shapes of the tuple elements have not been
443 set.
444 """
445 self._frozen = True
446 if self._tuple_types is None:
447 raise ValueError(
448 "Can't freeze an InfeedQueue without setting all tuple types.")
449 if self._tuple_shapes is None:
450 raise ValueError(
451 "Can't freeze an InfeedQueue without setting all tuple shapes.")
452 for shape in self._tuple_shapes:
453 if shape.dims is None:
454 raise ValueError(
455 "Can't freeze an InfeedQueue without setting all tuple shapes.")
456 for policy in self._sharding_policies:
457 policy.freeze()
458 self._validate()
459
460 def generate_dequeue_op(self, tpu_device=0):
461 """Generates the device-side Op to dequeue a tuple from the queue.

Callers 5

generate_dequeue_opMethod · 0.95
generate_enqueue_opsMethod · 0.95
testFreezingMethod · 0.95
generate_dequeue_opMethod · 0.45

Calls 1

_validateMethod · 0.95

Tested by 1

testFreezingMethod · 0.76