text
stringlengths
81
112k
Decode from dataset on new checkpoint. def continuous_decode(self): """Decode from dataset on new checkpoint.""" for _ in next_checkpoint(self._hparams.model_dir, self._decode_hparams.decode_timeout_mins): self.decode()
Decode from dataset on new checkpoint. def continuous_decode_on_train_data(self): """Decode from dataset on new checkpoint.""" for _ in next_checkpoint(self._hparams.model_dir, self._decode_hparams.decode_timeout_mins): self.decode(dataset_split=tf.estimator.ModeKeys.TRAIN)
Decode from dataset on new checkpoint. def continuous_decode_on_eval_data(self): """Decode from dataset on new checkpoint.""" if self._hparams.mlperf_mode: ckpt_generator = next_undecoded_checkpoint( self._hparams.model_dir, self._decode_hparams.decode_timeout_mins) else: ckpt_generator = next_checkpoint(self._hparams.model_dir, self._decode_hparams.decode_timeout_mins) for ckpt in ckpt_generator: current_step = decoding.get_step_from_ckpt_path(ckpt) tf.logging.info("Decoding step %d" % current_step) # Skip checkpoint 0. if current_step == 0: continue # Decode the latest checkpoint by default. checkpoint_path = None if self._hparams.mlperf_mode: self._decode_hparams.mlperf_decode_step = current_step checkpoint_path = ckpt mlperf_log.transformer_print(key=mlperf_log.EVAL_START) self.decode( dataset_split=tf.estimator.ModeKeys.EVAL, checkpoint_path=checkpoint_path) d_hparams = self._decode_hparams if self._hparams.mlperf_mode and d_hparams.mlperf_success: mlperf_log.transformer_print( key=mlperf_log.RUN_STOP, value={"success": "true"}) break d_hparams = self._decode_hparams if self._hparams.mlperf_mode and not d_hparams.mlperf_success: mlperf_log.transformer_print( key=mlperf_log.RUN_STOP, value={"success": "false"})
Decode from file on new checkpoint. def continuous_decode_from_file(self): """Decode from file on new checkpoint.""" for _ in next_checkpoint(self._hparams.model_dir, self._decode_hparams.decode_timeout_mins): self.decode(decode_from_file=True)
Flatten dict of dicts into a single dict with appropriate prefixes. Handles only 2 levels of nesting in the original dict. Args: original_dict: Dict which may contain one or more dicts. Returns: flat_dict: Dict without any nesting. Any dicts in the original dict have their keys as prefixes in the new dict. Raises: ValueError if the original dict has more than two levels of nesting. def _flatten_dict(original_dict): """Flatten dict of dicts into a single dict with appropriate prefixes. Handles only 2 levels of nesting in the original dict. Args: original_dict: Dict which may contain one or more dicts. Returns: flat_dict: Dict without any nesting. Any dicts in the original dict have their keys as prefixes in the new dict. Raises: ValueError if the original dict has more than two levels of nesting. """ flat_dict = {} for key, value in original_dict.items(): if isinstance(value, dict): for name, tensor in value.items(): if isinstance(tensor, dict): raise ValueError("flatten_dict only handles 2 levels of nesting.") flat_key = "__" + key + "_" + name flat_dict[flat_key] = tensor else: flat_dict[key] = value return flat_dict
Returns a dict of dicts if any prefixes match keys in the flat dict. The function handles the case where the prefix may not be a dict. Args: flat_dict: A dict without any nesting. prefixes: A list of strings which may have been dicts in the original structure. def _unflatten_dict(flat_dict, prefixes): """Returns a dict of dicts if any prefixes match keys in the flat dict. The function handles the case where the prefix may not be a dict. Args: flat_dict: A dict without any nesting. prefixes: A list of strings which may have been dicts in the original structure. """ original_dict = {} for key, value in flat_dict.items(): prefix_found = False for prefix in prefixes: full_prefix = "__" + prefix + "_" if key.startswith(full_prefix): # Add a dict to the original dict with key=prefix if prefix not in original_dict: original_dict[prefix] = {} original_dict[prefix][key[len(full_prefix):]] = value prefix_found = True break if not prefix_found: # No key matched a prefix in the for loop. original_dict[key] = value return original_dict
Dummy vars for restore to work when not using TPU codepath. def create_dummy_vars(): """Dummy vars for restore to work when not using TPU codepath.""" var_names = set([v.name for v in tf.global_variables()]) if "losses_avg/problem_0/total_loss:0" in var_names: return with tf.variable_scope("losses_avg"): with tf.variable_scope("problem_0"): for var_name in ["total", "extra", "training"]: tf.get_variable( "%s_loss" % var_name, initializer=100.0, trainable=False) with tf.variable_scope("train_stats"): tf.get_variable("problem_0_steps", initializer=0, trainable=False)
Create the metrics_fn that TPUEstimatorSpec expects. def create_tpu_eval_metrics_fn(problem, model_hparams): """Create the metrics_fn that TPUEstimatorSpec expects.""" metric_fns = [] eval_metrics = problem.eval_metric_fns(model_hparams) tm = _create_target_modality(problem.get_hparams(model_hparams).modality) if isinstance(tm, dict): for k, v in six.iteritems(tm): weights_fn = modalities.get_weights_fn(v) def make_metric_fn(metric_fn): def wrapped_metric_fn(logits, labels, features, weights_fn=weights_fn): kwargs = {} args, _, keywords, _ = inspect.getargspec(metric_fn) if ("features" in args) or keywords: kwargs["features"] = features num, den = metric_fn(logits, labels, weights_fn=weights_fn, **kwargs) return tf.metrics.mean(num, den) return wrapped_metric_fn for metric, metric_fn in six.iteritems(eval_metrics): if metric in TPU_METRIC_BLACKLIST: log_warn("Skipping eval metric %s in TPU_METRIC_BLACKLIST", metric) continue name = "%s/metrics-%s/%s" % (k, problem.name, metric) metric_fns.append((name, make_metric_fn(metric_fn))) else: weights_fn = modalities.get_weights_fn(tm) def make_metric_fn(metric_fn): def wrapped_metric_fn(logits, labels, features): kwargs = {} args, _, keywords, _ = inspect.getargspec(metric_fn) if ("features" in args) or keywords: kwargs["features"] = features num, den = metric_fn(logits, labels, weights_fn=weights_fn, **kwargs) return tf.metrics.mean(num, den) return wrapped_metric_fn for metric, metric_fn in six.iteritems(eval_metrics): if metric in TPU_METRIC_BLACKLIST: log_warn("Skipping eval metric %s in TPU_METRIC_BLACKLIST", metric) continue name = "metrics-%s/%s" % (problem.name, metric) metric_fns.append((name, make_metric_fn(metric_fn))) def all_metrics_fn(**kwargs): """Construct metrics dictionary.""" original_kwargs = _unflatten_dict(kwargs, prefixes=["logits", "features"]) del kwargs logits = original_kwargs["logits"] labels = original_kwargs["labels"] features = original_kwargs["features"] del original_kwargs metrics_dict = {} for name, fn in metric_fns: if isinstance(logits, dict) and isinstance(labels, dict): for k, v in six.iteritems(logits): metrics_dict["%s/%s" % (k, name)] = fn(v, labels[k], features) elif isinstance(logits, dict): tf.logging.warning("Logits is a dict, but labels is not; only " "evaluating logits['targets'] against labels.") metrics_dict["%s/%s" % ("targets", name)] = fn(logits["targets"], labels, features) else: metrics_dict[name] = fn(logits, labels, features) return metrics_dict return all_metrics_fn
Remove summaries from the default graph. def remove_summaries(): """Remove summaries from the default graph.""" g = tf.get_default_graph() key = tf.GraphKeys.SUMMARIES log_debug("Remove summaries %s" % str(g.get_collection(key))) del g.get_collection_ref(key)[:] assert not g.get_collection(key)
Construct a host_call writing scalar summaries. Args: model_dir: String containing path to train Returns: (fn, args) Pair to be called by TPUEstimator as the host_call. def create_host_call(model_dir): """Construct a host_call writing scalar summaries. Args: model_dir: String containing path to train Returns: (fn, args) Pair to be called by TPUEstimator as the host_call. """ graph = tf.get_default_graph() summaries = graph.get_collection(tf.GraphKeys.SUMMARIES) gs_t = tf.reshape(tf.to_int32(tf.train.get_global_step()), [1]) summary_kwargs = collections.OrderedDict() for t in summaries: # TODO(aidangomez): enable ImageSummary support when we have a faster method # see @shibow's comment in cl/202344570 if t.op.type not in ["ScalarSummary"]: tf.logging.warn("Ignoring unsupported tf.Summary type %s" % t.op.type) continue name = t.op.name tensor = t.op.inputs[1] if t.op.type == "ScalarSummary": assert tensor.shape.is_compatible_with([]) if tensor.dtype == tf.int64: tensor = tf.to_int32(tensor) summary_kwargs["ScalarSummary" + name] = tf.reshape(tensor, [1]) elif t.op.type == "ImageSummary": # TODO(aidangomez): as we move to support more types, update # common_layers.tpu_safe_image_summary if tensor.dtype != tf.float32: tf.logging.warn( "Currently T2T on TPU only supports ImageSummary of " "tf.float32-type Tensors. Skipping Tensor " "%s with dtype %s..." % (tensor.name, tensor.dtype)) continue # tensor = tf.to_float(tensor) summary_kwargs["ImageSummary" + name] = tensor # When no supported summaries are found, don't create host_call. Otherwise, # TPU outfeed queue would enqueue global_step while host_call doesn't dequeue # it, eventually causing hang. if not summary_kwargs: return None summary_kwargs["global_step"] = gs_t log_info("summary_kwargs %s" % str(summary_kwargs)) def host_call_fn(**kwargs): """Training host call. Creates summaries for training metrics. Args: **kwargs: Dict of {str: Tensor} , with `Tensor` of shape `[batch]`. Must contain key "global_step" with value of current global_step Tensor. Returns: List of summary ops to run on the CPU host. """ gs = tf.to_int64(kwargs.pop("global_step")[0]) with tf.contrib.summary.create_file_writer(model_dir).as_default(): with tf.contrib.summary.always_record_summaries(): # We need to use tf.contrib.summary in order to feed the `step`. for name, value in sorted(six.iteritems(kwargs)): if name.startswith("ScalarSummary"): name = name[len("ScalarSummary"):] tf.contrib.summary.scalar( name, tf.reduce_mean(tf.to_float(value)), step=gs) elif name.startswith("ImageSummary"): name = name[len("ImageSummary"):] tf.contrib.summary.image(name, value, step=gs) return tf.contrib.summary.all_summary_ops() return (host_call_fn, summary_kwargs)
Average losses across datashards. Args: sharded_losses: list<dict<str loss_name, Tensor loss>>. The loss can be a single Tensor or a 2-tuple (numerator and denominator). Returns: losses: dict<str loss_name, Tensor avg_loss> def average_sharded_losses(sharded_losses): """Average losses across datashards. Args: sharded_losses: list<dict<str loss_name, Tensor loss>>. The loss can be a single Tensor or a 2-tuple (numerator and denominator). Returns: losses: dict<str loss_name, Tensor avg_loss> """ losses = {} for loss_name in sorted(sharded_losses[0]): all_shards = [shard_losses[loss_name] for shard_losses in sharded_losses] if isinstance(all_shards[0], tuple): sharded_num, sharded_den = zip(*all_shards) mean_loss = ( tf.add_n(sharded_num) / tf.maximum( tf.cast(1.0, sharded_den[0].dtype), tf.add_n(sharded_den))) else: mean_loss = tf.reduce_mean(all_shards) losses[loss_name] = mean_loss return losses
Generate summaries for features. def summarize_features(features, num_shards=1): """Generate summaries for features.""" if not common_layers.should_generate_summaries(): return with tf.name_scope("input_stats"): for (k, v) in sorted(six.iteritems(features)): if (isinstance(v, tf.Tensor) and (v.get_shape().ndims > 1) and (v.dtype != tf.string)): tf.summary.scalar("%s_batch" % k, tf.shape(v)[0] // num_shards) tf.summary.scalar("%s_length" % k, tf.shape(v)[1]) nonpadding = tf.to_float(tf.not_equal(v, 0)) nonpadding_tokens = tf.reduce_sum(nonpadding) tf.summary.scalar("%s_nonpadding_tokens" % k, nonpadding_tokens) tf.summary.scalar("%s_nonpadding_fraction" % k, tf.reduce_mean(nonpadding))
Compose two custom getters. Example use: tf.get_variable_scope().set_custom_getter( compose_custom_getters(tf.get_variable_scope().custom_getter, new_getter)) This composes getters in the same way as creating a new variable scope with the new_getter, but it does not actually create a new variable scope. Args: getter_a: a custom getter - generally from the existing variable scope. getter_b: a custom getter Returns: a custom getter def _compose_custom_getters(getter_a, getter_b): """Compose two custom getters. Example use: tf.get_variable_scope().set_custom_getter( compose_custom_getters(tf.get_variable_scope().custom_getter, new_getter)) This composes getters in the same way as creating a new variable scope with the new_getter, but it does not actually create a new variable scope. Args: getter_a: a custom getter - generally from the existing variable scope. getter_b: a custom getter Returns: a custom getter """ if not getter_a: return getter_b if not getter_b: return getter_a def getter_fn(getter, *args, **kwargs): return getter_b(functools.partial(getter_a, getter), *args, **kwargs) return getter_fn
Set a custom getter in the current variable scope. Do not overwrite the existing custom getter - rather compose with it. Args: custom_getter: a custom getter. def set_custom_getter_compose(custom_getter): """Set a custom getter in the current variable scope. Do not overwrite the existing custom getter - rather compose with it. Args: custom_getter: a custom getter. """ tf.get_variable_scope().set_custom_getter( _compose_custom_getters(tf.get_variable_scope().custom_getter, custom_getter))
Initialize variables from given directory. def initialize_from_ckpt(ckpt_dir, hparams): """Initialize variables from given directory.""" model_dir = hparams.get("model_dir", None) already_has_ckpt = ( model_dir and tf.train.latest_checkpoint(model_dir) is not None) if already_has_ckpt: return tf.logging.info("Checkpoint dir: %s", ckpt_dir) reader = tf.contrib.framework.load_checkpoint(ckpt_dir) variable_map = {} for var in tf.contrib.framework.get_trainable_variables(): var_name = var.name.split(":")[0] if reader.has_tensor(var_name): tf.logging.info("Loading variable from checkpoint: %s", var_name) variable_map[var_name] = var else: tf.logging.info("Cannot find variable in checkpoint, skipping: %s", var_name) tf.train.init_from_checkpoint(ckpt_dir, variable_map)
Whether the target modality is real-valued. def _target_modality_is_real(self): """Whether the target modality is real-valued.""" vocab_size = self._problem_hparams.vocab_size["targets"] if vocab_size is not None and hasattr(self._hparams, "vocab_divisor"): vocab_size += (-vocab_size) % self._hparams.vocab_divisor modality = self._problem_hparams.modality["targets"] modality_name = self._hparams.name.get( "targets", modalities.get_name(modality))(self._hparams, vocab_size) return modality_name.startswith("real")
Estimator model_fn sharded along batch dimension. Args: sharded_features: {str: [Tensor]}. Features sharded along batch dimension. Each list is the same length (== number of shards). Returns: sharded_logits: [Tensor]. Logits for each shard of examples. losses: {str: 0-D Tensor}. Loss averaged across shards. def model_fn_sharded(self, sharded_features): """Estimator model_fn sharded along batch dimension. Args: sharded_features: {str: [Tensor]}. Features sharded along batch dimension. Each list is the same length (== number of shards). Returns: sharded_logits: [Tensor]. Logits for each shard of examples. losses: {str: 0-D Tensor}. Loss averaged across shards. """ dp = self._data_parallelism # [{str: Tensor}]. Transpose of 'sharded_features'. datashard_to_features = self._to_features_per_datashard(sharded_features) if self.use_body_sharded(): if self.hparams.scheduled_sampling_prob > 0.0: raise NotImplementedError( "Scheduled sampling for non-sharded body only.") # MoE models override body_sharded transformed_features = dp(self.bottom, datashard_to_features) body_out = self.body_sharded( self._to_single_features_dict(transformed_features)) body_out, losses = self._normalize_body_output(body_out) if "training" in losses: log_info("Skipping T2TModel top and loss because training loss " "returned from body") sharded_logits = body_out else: if isinstance(body_out, dict): sharded_logits = collections.OrderedDict() sharded_losses = collections.OrderedDict() for k, v in sorted(six.iteritems(body_out)): sharded_logits[k] = dp(self.top, v, datashard_to_features) sharded_losses[k] = dp(self.loss, sharded_logits[k], datashard_to_features) training_loss_dict = average_sharded_losses([({ "training": l } for l in loss) for loss in sharded_losses.values()]) losses.update(training_loss_dict) else: sharded_logits = dp(self.top, body_out, datashard_to_features) sharded_losses = dp(self.loss, sharded_logits, datashard_to_features) if isinstance(sharded_losses, tuple): nums, dens = sharded_losses sharded_losses = zip(nums, dens) training_loss_dict = average_sharded_losses([{ "training": loss } for loss in sharded_losses]) losses.update(training_loss_dict) else: sharded_logits, sharded_losses = dp(self.model_fn, datashard_to_features) sharded_logits, sharded_losses = dp( self.maybe_scheduled_sampling, datashard_to_features, sharded_logits, sharded_losses) if isinstance(sharded_logits[0], dict): temp_dict = {k: [] for k, _ in six.iteritems(sharded_logits[0])} for k, _ in six.iteritems(sharded_logits[0]): for l in sharded_logits: temp_dict[k].append(l[k]) sharded_logits = temp_dict losses = average_sharded_losses(sharded_losses) return sharded_logits, losses
Transforms features to feed into body. Args: features: dict of str to Tensor. Typically it is the preprocessed data batch after Problem's preprocess_example(). Returns: transformed_features: dict of same key-value pairs as features. The value Tensors are newly transformed. def bottom(self, features): """Transforms features to feed into body. Args: features: dict of str to Tensor. Typically it is the preprocessed data batch after Problem's preprocess_example(). Returns: transformed_features: dict of same key-value pairs as features. The value Tensors are newly transformed. """ if not self._problem_hparams: log_warn("Without a Problem, T2TModel.bottom is a passthrough.") return features transformed_features = collections.OrderedDict() all_previous_modalities = [] target_modality = _create_target_modality(self._problem_hparams.modality) # Transform features via its corresponding modality. for feature_name, modality in sorted( six.iteritems(self._problem_hparams.modality)): if feature_name not in features: tf.logging.warning("Missing feature %s - ignoring." % feature_name) continue vocab_size = self._problem_hparams.vocab_size[feature_name] if vocab_size is not None and hasattr(self._hparams, "vocab_divisor"): vocab_size += (-vocab_size) % self._hparams.vocab_divisor modality_name = self._hparams.name.get( feature_name, modalities.get_name(modality))(self._hparams, vocab_size) # Use if-else clauses to preserve behavior of previous changes: namely, # the variable scope name for the targets feature if there is only one # target modality; and to reuse variable scopes for only input modalities. if feature_name in target_modality: if len(target_modality) > 1: variable_scope_name = "%s/%s" % (modality_name, feature_name) else: variable_scope_name = modality_name bottom = self._hparams.bottom.get( feature_name, modalities.get_targets_bottom(modality)) # TODO(aidangomez): share variables? with tf.variable_scope(variable_scope_name) as vs: self._add_variable_scope(variable_scope_name, vs) log_info("Transforming feature '%s' with %s.targets_bottom", feature_name, modality_name) transformed_features[feature_name] = bottom(features[feature_name], self._hparams, vocab_size) else: bottom = self._hparams.bottom.get(feature_name, modalities.get_bottom(modality)) do_reuse = modality_name in all_previous_modalities with tf.variable_scope(modality_name, reuse=do_reuse) as vs: self._add_variable_scope(modality_name, vs) log_info("Transforming feature '%s' with %s.bottom", feature_name, modality_name) transformed_features[feature_name] = bottom(features[feature_name], self._hparams, vocab_size) all_previous_modalities.append(modality_name) for key in features: if key not in transformed_features: # For features without a modality, we pass them along as is transformed_features[key] = features[key] else: # Other features get passed along with the "raw" suffix transformed_features[key + "_raw"] = features[key] return transformed_features
Computes logits given body output and features. Args: body_output: dict of str to Tensor, comprising one key-value pair for each target. Each value denotes the target's pre-logit activations. Alternatively, it may be a single Tensor denoting the pre-logits for that target. features: dict of str to Tensor. Typically it is the preprocessed data batch after Problem's preprocess_example(). Returns: logits: dict of str to Tensor, denoting each logits for each target; or a single Tensor denoting the logits for that target. When targets are generated at training time: logits == { "self_generated_targets": <generated targets tensor> "logits": <original logits Tensor or dict> } def top(self, body_output, features): """Computes logits given body output and features. Args: body_output: dict of str to Tensor, comprising one key-value pair for each target. Each value denotes the target's pre-logit activations. Alternatively, it may be a single Tensor denoting the pre-logits for that target. features: dict of str to Tensor. Typically it is the preprocessed data batch after Problem's preprocess_example(). Returns: logits: dict of str to Tensor, denoting each logits for each target; or a single Tensor denoting the logits for that target. When targets are generated at training time: logits == { "self_generated_targets": <generated targets tensor> "logits": <original logits Tensor or dict> } """ if isinstance(body_output, dict): logits = {} for k, v in six.iteritems(body_output): # TODO(aidangomez): share variables here? with tf.variable_scope(k) as top_vs: self._add_variable_scope("top_%s" % k, top_vs) logits[k] = self._top_single(v, k, features) return logits else: return self._top_single(body_output, "targets", features)
Return a training op minimizing loss. def optimize(self, loss, num_async_replicas=1, use_tpu=False): """Return a training op minimizing loss.""" lr = learning_rate.learning_rate_schedule(self.hparams) if num_async_replicas > 1: log_info("Dividing learning rate by num_async_replicas: %d", num_async_replicas) lr /= math.sqrt(float(num_async_replicas)) train_op = optimize.optimize(loss, lr, self.hparams, use_tpu=use_tpu) return train_op
Set hparams with the given mode. def set_mode(self, mode): """Set hparams with the given mode.""" log_info("Setting T2TModel mode to '%s'", mode) hparams = hparams_lib.copy_hparams(self._original_hparams) hparams.add_hparam("mode", mode) # When not in training mode, set all forms of dropout to zero. if mode != tf.estimator.ModeKeys.TRAIN: for key in hparams.values(): if key.endswith("dropout") or key == "label_smoothing": log_info("Setting hparams.%s to 0.0", key) setattr(hparams, key, 0.0) self._hparams = hparams
Autoregressive eval. Quadratic time in decode_length. Args: features: an map of string to `Tensor` decode_length: an integer. How many additional timesteps to decode. Returns: logits: `Tensor` losses: a dictionary: {loss-name (string): floating point `Scalar`}. Contains a single key "training". def eval_autoregressive(self, features=None, decode_length=50): """Autoregressive eval. Quadratic time in decode_length. Args: features: an map of string to `Tensor` decode_length: an integer. How many additional timesteps to decode. Returns: logits: `Tensor` losses: a dictionary: {loss-name (string): floating point `Scalar`}. Contains a single key "training". """ results = self._slow_greedy_infer(features, decode_length=decode_length) return results["logits"], results["losses"]
A inference method. Quadratic time in decode_length. Args: features: an map of string to `Tensor` decode_length: an integer. How many additional timesteps to decode. beam_size: number of beams. top_beams: an integer. How many of the beams to return. alpha: Float that controls the length penalty. larger the alpha, stronger the preference for longer translations. use_tpu: bool, whether to build the inference graph for TPU. Returns: A dict of decoding results { "outputs": integer `Tensor` of decoded ids of shape [batch_size, <= decode_length] if beam_size == 1 or [batch_size, top_beams, <= decode_length] "scores": decoding log probs from the beam search, None if using greedy decoding (beam_size=1) } if slow greedy decoding is used then the dict will also contain { "logits": `Tensor` of shape [batch_size, time, 1, 1, vocab_size]. "losses": a dictionary: {loss-name (string): floating point `Scalar` } def infer(self, features=None, decode_length=50, beam_size=1, top_beams=1, alpha=0.0, use_tpu=False): """A inference method. Quadratic time in decode_length. Args: features: an map of string to `Tensor` decode_length: an integer. How many additional timesteps to decode. beam_size: number of beams. top_beams: an integer. How many of the beams to return. alpha: Float that controls the length penalty. larger the alpha, stronger the preference for longer translations. use_tpu: bool, whether to build the inference graph for TPU. Returns: A dict of decoding results { "outputs": integer `Tensor` of decoded ids of shape [batch_size, <= decode_length] if beam_size == 1 or [batch_size, top_beams, <= decode_length] "scores": decoding log probs from the beam search, None if using greedy decoding (beam_size=1) } if slow greedy decoding is used then the dict will also contain { "logits": `Tensor` of shape [batch_size, time, 1, 1, vocab_size]. "losses": a dictionary: {loss-name (string): floating point `Scalar` } """ set_custom_getter_compose(self._custom_getter) with self._eager_var_store.as_default(): # TODO(rsepassi): Make decoding work with real-valued model outputs # (i.e. if the target modality is RealModality). self.prepare_features_for_infer(features) if not self.has_input and beam_size > 1: log_warn("Beam searching for a model with no inputs.") if not self.has_input and self.hparams.sampling_method != "random": log_warn("Non-random sampling for a model with no inputs.") self._fill_problem_hparams_features(features) if self._problem_hparams: target_modality = self._problem_hparams.modality["targets"] if target_modality == modalities.ModalityType.CLASS_LABEL: beam_size = 1 # No use to run beam-search for a single class. if beam_size == 1: log_info("Greedy Decoding") results = self._greedy_infer(features, decode_length, use_tpu) else: log_info("Beam Decoding with beam size %d" % beam_size) results = self._beam_decode(features, decode_length, beam_size, top_beams, alpha, use_tpu) return results
Beam search decoding. Models should ideally implement a more efficient version of this function. Args: features: an map of string to `Tensor` decode_length: an integer. How many additional timesteps to decode. beam_size: number of beams. top_beams: an integer. How many of the beams to return. alpha: Float that controls the length penalty. larger the alpha, stronger the preference for longer translations. use_tpu: A bool, whether to do beam decode on TPU. Returns: samples: an integer `Tensor`. Top samples from the beam search def _beam_decode(self, features, decode_length, beam_size, top_beams, alpha, use_tpu=False): """Beam search decoding. Models should ideally implement a more efficient version of this function. Args: features: an map of string to `Tensor` decode_length: an integer. How many additional timesteps to decode. beam_size: number of beams. top_beams: an integer. How many of the beams to return. alpha: Float that controls the length penalty. larger the alpha, stronger the preference for longer translations. use_tpu: A bool, whether to do beam decode on TPU. Returns: samples: an integer `Tensor`. Top samples from the beam search """ return self._beam_decode_slow(features, decode_length, beam_size, top_beams, alpha, use_tpu)
Slow version of Beam search decoding. Quadratic time in decode_length. Args: features: an map of string to `Tensor` decode_length: an integer. How many additional timesteps to decode. beam_size: number of beams. top_beams: an integer. How many of the beams to return. alpha: Float that controls the length penalty. larger the alpha, stronger the preference for longer translations. use_tpu: A bool, whether to do slow beam decode on TPU. Returns: samples: an integer `Tensor`. Top samples from the beam search. Raises: NotImplementedError: If use_tpu is set to true. def _beam_decode_slow(self, features, decode_length, beam_size, top_beams, alpha, use_tpu=False): """Slow version of Beam search decoding. Quadratic time in decode_length. Args: features: an map of string to `Tensor` decode_length: an integer. How many additional timesteps to decode. beam_size: number of beams. top_beams: an integer. How many of the beams to return. alpha: Float that controls the length penalty. larger the alpha, stronger the preference for longer translations. use_tpu: A bool, whether to do slow beam decode on TPU. Returns: samples: an integer `Tensor`. Top samples from the beam search. Raises: NotImplementedError: If use_tpu is set to true. """ batch_size = common_layers.shape_list(features["inputs"])[0] def symbols_to_logits_fn(ids, i=None): """Go from ids to logits.""" ids = tf.expand_dims(tf.expand_dims(ids, axis=2), axis=3) ids = tf.pad(ids[:, 1:], [[0, 0], [0, 1], [0, 0], [0, 0]]) if "partial_targets" in features: pt = features["partial_targets"] pt_length = common_layers.shape_list(pt)[1] pt = tf.tile(pt, [1, beam_size]) pt = tf.reshape(pt, [batch_size * beam_size, pt_length, 1, 1]) ids = tf.concat([pt, ids], axis=1) features["targets"] = ids if i is not None: features["decode_loop_step"] = i self._coverage = None logits, _ = self(features) # pylint: disable=not-callable # now self._coverage is a coverage tensor for the first datashard. # it has shape [batch_size] and contains floats between 0 and # source_length. if self._problem_hparams: modality = self._problem_hparams.modality["targets"] top = self._hparams.top.get("targets", modalities.get_top(modality)) if getattr(top, "pointwise", False): return tf.squeeze(logits, axis=[1, 2, 3]) # -1 due to the pad above. current_output_position = common_layers.shape_list(ids)[1] - 1 logits = logits[:, current_output_position, :, :] return tf.squeeze(logits, axis=[1, 2]) def _clone_examples_for_beam(old_feature, n): """Clone each example n times.""" old_shape = common_layers.shape_list(old_feature) assert len(old_shape) >= 1 # Expand the inputs in to the beam size. feature = tf.expand_dims(old_feature, 1) feature = tf.tile(feature, [1, n] + [1] * (len(old_shape) - 1)) new_shape = common_layers.shape_list(feature) feature = tf.reshape(feature, [new_shape[0] * new_shape[1]] + new_shape[2:]) return feature initial_ids = tf.zeros([batch_size], dtype=tf.int32) # Clone select features multiple times to account for beam size. old_features = {} for feature_name in ["inputs", "knowledge"]: if feature_name not in features: continue old_features[feature_name] = features[feature_name] features[feature_name] = _clone_examples_for_beam( features[feature_name], beam_size) vocab_size = self._problem_hparams.vocab_size["targets"] if vocab_size is not None and hasattr(self._hparams, "vocab_divisor"): vocab_size += (-vocab_size) % self._hparams.vocab_divisor # Setting decode length to input length + decode_length if "partial_targets" not in features: inputs = features["inputs"] decode_length = (common_layers.shape_list(inputs)[1] + features.get("decode_length", decode_length)) ids, scores, _ = beam_search.beam_search( symbols_to_logits_fn, initial_ids, beam_size, decode_length, vocab_size, alpha, stop_early=(top_beams == 1), use_tpu=use_tpu) # Set features back to the unexpanded form to not to confuse the # Estimator! features.update(old_features) # Return `top_beams` decodings (also remove initial id from the beam search) # TODO(lukaszkaiser): make it work multi-problem. if top_beams == 1: samples = ids[:, 0, 1:] else: samples = ids[:, :top_beams, 1:] return {"outputs": samples, "scores": scores}
A greedy inference method. Models should ideally implement a more efficient version of this function. Args: features: an map of string to `Tensor` decode_length: an integer. How many additional timesteps to decode. use_tpu: A bool, whether to build the inference graph for TPU. Returns: A dict of decoding results { "outputs": integer `Tensor` of decoded ids of shape [batch_size, <= decode_length] if beam_size == 1 or [batch_size, top_beams, <= decode_length] "scores": None "logits": `Tensor` of shape [batch_size, time, 1, 1, vocab_size]. "losses": a dictionary: {loss-name (string): floating point `Scalar`} } def _greedy_infer(self, features, decode_length, use_tpu=False): """A greedy inference method. Models should ideally implement a more efficient version of this function. Args: features: an map of string to `Tensor` decode_length: an integer. How many additional timesteps to decode. use_tpu: A bool, whether to build the inference graph for TPU. Returns: A dict of decoding results { "outputs": integer `Tensor` of decoded ids of shape [batch_size, <= decode_length] if beam_size == 1 or [batch_size, top_beams, <= decode_length] "scores": None "logits": `Tensor` of shape [batch_size, time, 1, 1, vocab_size]. "losses": a dictionary: {loss-name (string): floating point `Scalar`} } """ if use_tpu: return self._slow_greedy_infer_tpu(features, decode_length) return self._slow_greedy_infer(features, decode_length)
A slow greedy inference method on TPU. Quadratic time in decode_length. Args: features: An map of string to `Tensor`. decode_length: An integer, how many additional timesteps to decode. Returns: A dict of decoding results { "outputs": integer `Tensor` of decoded ids of shape [batch_size, <= decode_length] if beam_size == 1 or [batch_size, top_beams, <= decode_length] "scores": None "logits": `Tensor` of shape [batch_size, time, 1, 1, vocab_size]. "losses": a dictionary: {loss-name (string): floating point `Scalar`} } def _slow_greedy_infer_tpu(self, features, decode_length): """A slow greedy inference method on TPU. Quadratic time in decode_length. Args: features: An map of string to `Tensor`. decode_length: An integer, how many additional timesteps to decode. Returns: A dict of decoding results { "outputs": integer `Tensor` of decoded ids of shape [batch_size, <= decode_length] if beam_size == 1 or [batch_size, top_beams, <= decode_length] "scores": None "logits": `Tensor` of shape [batch_size, time, 1, 1, vocab_size]. "losses": a dictionary: {loss-name (string): floating point `Scalar`} } """ if not features: features = {} inputs_old = None if "inputs" in features and len(features["inputs"].shape) < 4: inputs_old = features["inputs"] features["inputs"] = tf.expand_dims(features["inputs"], 2) if not self.has_input: # Prepare partial targets. # In either features["inputs"] or features["targets"]. # We force the outputs to begin with these sequences. partial_targets = features.get("inputs") if partial_targets is None: partial_targets = features["targets"] features["partial_targets"] = tf.to_int64(partial_targets) # Save the targets in a var and reassign it after the tf.while loop to avoid # having targets being in a 'while' frame. This ensures targets when used # in metric functions stays in the same frame as other vars. targets_old = features.get("targets", None) target_modality = self._problem_hparams.modality["targets"] def infer_step(i, recent_output, recent_logits, unused_loss): """Inference step.""" if not tf.executing_eagerly(): recent_output.set_shape([None, None, None, 1]) padded = tf.pad(recent_output, [[0, 0], [0, 1], [0, 0], [0, 0]]) features["targets"] = padded # This is inefficient in that it generates samples at all timesteps, # not just the last one, except if target_modality is pointwise. features["decode_loop_step"] = i samples, logits, losses = self.sample(features) # Concatenate the already-generated recent_output with last timestep # of the newly-generated samples.z top = self._hparams.top.get("targets", modalities.get_top(target_modality)) if getattr(top, "pointwise", False): cur_sample = samples[:, -1, :, :] else: cur_sample = samples[:, i, :, :] samples = tf.transpose(recent_output, perm=[1, 0, 2, 3]) samples = inplace_ops.alias_inplace_update(samples, i, tf.to_int64(cur_sample)) samples = tf.transpose(samples, perm=[1, 0, 2, 3]) if not tf.executing_eagerly(): samples.set_shape([None, None, None, 1]) # Assuming we have one shard for logits. recent_logits = tf.transpose(recent_logits, perm=[1, 0, 2, 3, 4]) recent_logits = inplace_ops.alias_inplace_update( recent_logits, i, tf.squeeze(logits[:, -1:], axis=1)) logits = tf.transpose(recent_logits, perm=[1, 0, 2, 3, 4]) loss = sum([l for l in losses.values() if l is not None]) return i + 1, samples, logits, loss # Create an initial output tensor. This will be passed # to the infer_step, which adds one timestep at every iteration. if "partial_targets" in features: initial_output = tf.to_int64(features["partial_targets"]) while len(initial_output.get_shape().as_list()) < 4: initial_output = tf.expand_dims(initial_output, 2) batch_size = common_layers.shape_list(initial_output)[0] else: batch_size = common_layers.shape_list(features["inputs"])[0] initial_output = tf.zeros((batch_size, 0, 1, 1), dtype=tf.int64) # Hack: foldl complains when the output shape is less specified than the # input shape, so we confuse it about the input shape. initial_output = tf.slice(initial_output, [0, 0, 0, 0], common_layers.shape_list(initial_output)) target_modality = self._problem_hparams.modality["targets"] if target_modality == modalities.ModalityType.CLASS_LABEL: decode_length = 1 else: if "partial_targets" in features: prefix_length = common_layers.shape_list(features["partial_targets"])[1] else: prefix_length = common_layers.shape_list(features["inputs"])[1] decode_length = prefix_length + decode_length # Initial values of result, logits and loss. result = tf.concat( [initial_output, tf.zeros([batch_size, decode_length, 1, 1], tf.int64)], axis=1) # tensor padded to [batch_size, decode_length, 1, 1, vocab_size] vocab_size = self._problem_hparams.vocab_size["targets"] if vocab_size is not None and hasattr(self._hparams, "vocab_divisor"): vocab_size += (-vocab_size) % self._hparams.vocab_divisor logits = tf.zeros((batch_size, decode_length, 1, 1, vocab_size)) if not tf.executing_eagerly(): logits.set_shape([None, None, None, None, None]) loss = 0.0 def while_exit_cond(i, result, logits, loss): # pylint: disable=unused-argument """Exit the loop either if reach decode_length or EOS.""" not_overflow = i < decode_length if self._problem_hparams.stop_at_eos: def fn_not_eos(): # Check if the last predicted element is a EOS return tf.reduce_any( tf.not_equal( tf.squeeze(result[:, -1, :, :]), text_encoder.EOS_ID)) not_eos = tf.cond( # We only check for early stopping if there is at least 1 element ( # otherwise not_eos will crash). tf.not_equal(i, 0), fn_not_eos, lambda: True, ) return tf.cond( tf.equal(batch_size, 1), # If batch_size == 1, we check EOS for early stopping. lambda: tf.logical_and(not_overflow, not_eos), # Else, just wait for max length lambda: not_overflow) return not_overflow _, result, logits, loss = tf.while_loop( while_exit_cond, infer_step, [tf.constant(0), result, logits, loss], shape_invariants=[ tf.TensorShape([]), tf.TensorShape([batch_size, decode_length, 1, 1]), tf.TensorShape([batch_size, decode_length, 1, 1, vocab_size]), tf.TensorShape([]), ], back_prop=False, parallel_iterations=1) if inputs_old is not None: # Restore to not confuse Estimator. features["inputs"] = inputs_old # Reassign targets back to the previous value. if targets_old is not None: features["targets"] = targets_old losses = {"training": loss} if "partial_targets" in features: partial_target_length = common_layers.shape_list( features["partial_targets"])[1] result = tf.slice(result, [0, partial_target_length, 0, 0], [-1, -1, -1, -1]) return { "outputs": result, "scores": None, "logits": logits, "losses": losses, }
Run the model and extract samples. Args: features: an map of string to `Tensor`. Returns: samples: an integer `Tensor`. logits: a list of `Tensor`s, one per datashard. losses: a dictionary: {loss-name (string): floating point `Scalar`}. def sample(self, features): """Run the model and extract samples. Args: features: an map of string to `Tensor`. Returns: samples: an integer `Tensor`. logits: a list of `Tensor`s, one per datashard. losses: a dictionary: {loss-name (string): floating point `Scalar`}. """ logits, losses = self(features) # pylint: disable=not-callable if self._target_modality_is_real: return logits, logits, losses # Raw numbers returned from real modality. if self.hparams.sampling_method == "argmax": samples = tf.argmax(logits, axis=-1) else: assert self.hparams.sampling_method == "random" def multinomial_squeeze(logits, temperature=1.0): logits_shape = common_layers.shape_list(logits) reshaped_logits = ( tf.reshape(logits, [-1, logits_shape[-1]]) / temperature) choices = tf.multinomial(reshaped_logits, 1) choices = tf.reshape(choices, logits_shape[:-1]) return choices samples = multinomial_squeeze(logits, self.hparams.sampling_temp) return samples, logits, losses
Model fn for Estimator. Args: hparams: HParams, model hyperparameters features: dict<str name, Tensor feature> labels: Tensor mode: tf.estimator.ModeKeys config: RunConfig, possibly with data_parallelism attribute params: dict, may include batch_size, use_tpu decode_hparams: HParams, used when mode == PREDICT. use_tpu: A bool, whether to build the inference graph for TPU. Returns: TPUEstimatorSpec if use tpu else EstimatorSpec def estimator_model_fn(cls, hparams, features, labels, mode, config=None, params=None, decode_hparams=None, use_tpu=False): """Model fn for Estimator. Args: hparams: HParams, model hyperparameters features: dict<str name, Tensor feature> labels: Tensor mode: tf.estimator.ModeKeys config: RunConfig, possibly with data_parallelism attribute params: dict, may include batch_size, use_tpu decode_hparams: HParams, used when mode == PREDICT. use_tpu: A bool, whether to build the inference graph for TPU. Returns: TPUEstimatorSpec if use tpu else EstimatorSpec """ if mode == tf.estimator.ModeKeys.TRAIN: create_dummy_vars() hparams = hparams_lib.copy_hparams(hparams) # Instantiate model data_parallelism = None if not use_tpu and config: data_parallelism = config.data_parallelism reuse = tf.get_variable_scope().reuse model = cls( hparams, mode, data_parallelism=data_parallelism, decode_hparams=decode_hparams, _reuse=reuse) # PREDICT mode if mode == tf.estimator.ModeKeys.PREDICT: if use_tpu: inputs = features.get("inputs") if inputs is None: inputs = features["targets"] shape = inputs.get_shape().as_list() if shape[0] is None: shape[0] = decode_hparams.batch_size or hparams.batch_size if shape[1] is None: shape[1] = hparams.max_input_seq_length or hparams.max_length inputs.set_shape(shape) return model.estimator_spec_predict(features, use_tpu=use_tpu) # TRAIN and EVAL modes if hparams.eval_run_autoregressive and mode == tf.estimator.ModeKeys.EVAL: logits, losses_dict = model.eval_autoregressive(features) else: logits, losses_dict = model(features) # pylint: disable=not-callable # Support model-generated labels by overriding features["targets"] with # logits["self_generated_targets"]. if isinstance(logits, dict) and "self_generated_targets" in logits: # Overwrite 'features["targets"]' and 'labels' # by logits["self_generated_targets"]. tf.logging.info("Replacing targets with model-provided targets.") features["targets"] = labels = logits.pop("self_generated_targets") assert list(logits.keys()) == ["logits"], ( # See "Returns" in the "top" method docstring for the expected # "logits" format when targets are generated at training time. "Expect only key 'logits' when there is 'self_generated_targets'. " "Found {}".format(logits.keys()) ) # Recover the original logits tensor from the logits dict. logits = logits["logits"] # Can be a tf.Tensor or a dict. # Set known shapes if common_layers.is_xla_compiled(): if isinstance(logits, dict): for k, v in sorted(six.iteritems(logits)): if "scalar/" in k: continue shape = v.get_shape().as_list() if shape[0] is None: shape[0] = params["batch_size"] if shape[1] is None: shape[1] = hparams.max_length v.set_shape(shape) else: shape = logits.get_shape().as_list() if shape[0] is None: shape[0] = params["batch_size"] if shape[1] is None: shape[1] = hparams.max_length logits.set_shape(shape) assert "training" in losses_dict # Attack mode if mode == "attack": return logits # Summarize losses model._summarize_losses(losses_dict) # pylint: disable=protected-access # Accumulate losses loss = sum(losses_dict[key] for key in sorted(losses_dict.keys())) # EVAL mode if mode == tf.estimator.ModeKeys.EVAL: return model.estimator_spec_eval(features, logits, labels, loss, losses_dict) # TRAIN mode assert mode == tf.estimator.ModeKeys.TRAIN num_async_replicas = 1 if config and not use_tpu: num_async_replicas = config.t2t_device_info["num_async_replicas"] return model.estimator_spec_train( loss, num_async_replicas=num_async_replicas, use_tpu=use_tpu)
Constructs `tf.estimator.EstimatorSpec` for TRAIN (training) mode. def estimator_spec_train(self, loss, num_async_replicas=1, use_tpu=False): """Constructs `tf.estimator.EstimatorSpec` for TRAIN (training) mode.""" train_op = self.optimize(loss, num_async_replicas=num_async_replicas, use_tpu=use_tpu) if use_tpu: if self._hparams.warm_start_from: def scaffold_fn(): self.initialize_from_ckpt(self._hparams.warm_start_from) return tf.train.Scaffold() else: scaffold_fn = None # Note: important to call this before remove_summaries() if self.hparams.tpu_enable_host_call: host_call = self.create_train_host_call() else: host_call = None remove_summaries() return tf.contrib.tpu.TPUEstimatorSpec( tf.estimator.ModeKeys.TRAIN, loss=loss, train_op=train_op, host_call=host_call, scaffold_fn=scaffold_fn) else: if self._hparams.warm_start_from: self.initialize_from_ckpt(self._hparams.warm_start_from) # When loading weights from a pre-trained model, you want to be able to # load separate weights into the encoder and decoder. if self._hparams.warm_start_from_second: self.initialize_from_ckpt(self._hparams.warm_start_from_second) return tf.estimator.EstimatorSpec( tf.estimator.ModeKeys.TRAIN, loss=loss, train_op=train_op)
Constructs `tf.estimator.EstimatorSpec` for EVAL (evaluation) mode. def estimator_spec_eval(self, features, logits, labels, loss, losses_dict): """Constructs `tf.estimator.EstimatorSpec` for EVAL (evaluation) mode.""" del losses_dict hparams = self.hparams if not hasattr(hparams, "problem"): raise NotImplementedError(_no_problem_err("estimator_spec_eval")) problem = hparams.problem if common_layers.is_xla_compiled(): # Note: important to call this before remove_summaries() if self.hparams.tpu_enable_host_call: host_call = self.create_eval_host_call() else: host_call = None remove_summaries() eval_metrics_fn = create_tpu_eval_metrics_fn(problem, hparams) batch_size = [feature.shape.as_list()[0] for _, feature in features.items() if feature.shape.ndims][0] # Add batch dimension to all features since tpu requires the batch # dimension on all tensors. for name, feature in features.items(): if not feature.shape.as_list(): # All features must have a batch dimension feature = tf.tile(tf.expand_dims(feature, 0), [batch_size]) features[name] = feature eval_metrics_fn_args = dict( logits=logits, # possibly a dict labels=labels, features=features, # dict ) eval_metrics_fn_flat_args = _flatten_dict(eval_metrics_fn_args) return tf.contrib.tpu.TPUEstimatorSpec( tf.estimator.ModeKeys.EVAL, eval_metrics=(eval_metrics_fn, eval_metrics_fn_flat_args), host_call=host_call, loss=loss) else: task_list = [problem] if hasattr(problem, "task_list"): task_list = problem.task_list eval_metrics_fns = metrics.create_evaluation_metrics(task_list, hparams) eval_metrics = {} for metric_name, metric_fn in six.iteritems(eval_metrics_fns): if isinstance(logits, dict): # the key is located in the center of metric_name: "metrics-%s/%s/%s" k = metric_name.split("/")[1] if k in logits: eval_metrics[metric_name] = metric_fn(logits[k], features, features[k]) else: # We do not make it an error because we sometimes run models that # predict only parts of the targets defined by the Problem class. # For example, an autoencoder or pure-video model can run on a gym # problem even if another model is also predicting other things, # like actions or rewards. tf.logging.warning("No key %s in logits for evaluation." % k) else: eval_metrics[metric_name] = metric_fn(logits, features, features["targets"]) if isinstance(logits, dict): predictions = logits else: predictions = {"predictions": logits} evaluation_hooks = [] # Create a SummarySaverHook eval_dir = os.path.join( self.hparams.model_dir, self.hparams.get("eval_dir_name", "eval")) eval_summary_hook = tf.train.SummarySaverHook( save_steps=1, output_dir=eval_dir, summary_op=tf.summary.merge_all()) evaluation_hooks.append(eval_summary_hook) evaluation_hooks += problem.eval_hooks(features, logits, hparams) return tf.estimator.EstimatorSpec( tf.estimator.ModeKeys.EVAL, predictions=predictions, eval_metric_ops=eval_metrics, evaluation_hooks=evaluation_hooks, loss=loss)
Constructs `tf.estimator.EstimatorSpec` for PREDICT (inference) mode. def estimator_spec_predict(self, features, use_tpu=False): """Constructs `tf.estimator.EstimatorSpec` for PREDICT (inference) mode.""" decode_hparams = self._decode_hparams top_beams = decode_hparams.beam_size if decode_hparams.return_beams else 1 infer_out = self.infer( features, beam_size=decode_hparams.beam_size, top_beams=top_beams, alpha=decode_hparams.alpha, decode_length=decode_hparams.extra_length, use_tpu=use_tpu) if isinstance(infer_out, dict): outputs = infer_out["outputs"] scores = infer_out["scores"] else: outputs = infer_out scores = None inputs = features.get("inputs") if inputs is None: inputs = features["targets"] predictions = { "outputs": outputs, "scores": scores, "inputs": inputs, "targets": features.get("infer_targets"), } # Pass through remaining features for name, feature in features.items(): if name not in list(predictions.keys()) + ["infer_targets"]: if name == "decode_loop_step": continue if not feature.shape.as_list(): # All features must have a batch dimension batch_size = common_layers.shape_list(outputs)[0] feature = tf.tile(tf.expand_dims(feature, 0), [batch_size]) predictions[name] = feature _del_dict_non_tensors(predictions) export_out = {"outputs": predictions["outputs"]} if "scores" in predictions: export_out["scores"] = predictions["scores"] # Necessary to rejoin examples in the correct order with the Cloud ML Engine # batch prediction API. if "batch_prediction_key" in predictions: export_out["batch_prediction_key"] = predictions["batch_prediction_key"] export_outputs = { tf.saved_model.signature_constants.DEFAULT_SERVING_SIGNATURE_DEF_KEY: tf.estimator.export.PredictOutput(export_out) } if use_tpu: # Note: important to call this before remove_summaries() if self.hparams.tpu_enable_host_call: host_call = self.create_eval_host_call() else: host_call = None remove_summaries() return tf.contrib.tpu.TPUEstimatorSpec( tf.estimator.ModeKeys.PREDICT, predictions=predictions, host_call=host_call, export_outputs=export_outputs) else: return tf.estimator.EstimatorSpec( tf.estimator.ModeKeys.PREDICT, predictions=predictions, export_outputs=export_outputs)
Adds `tf.summary`s to all terms in the losses dictionary. def _summarize_losses(self, losses_dict): """Adds `tf.summary`s to all terms in the losses dictionary.""" if common_layers.should_generate_summaries(): with tf.name_scope("losses"): for loss_name, loss_val in sorted(losses_dict.items()): tf.summary.scalar(loss_name, loss_val)
Scheduled sampling. Performs forward inference again with "targets" feature replaced with values sampled from the model. This is the identity unless self.hparams.scheduled_sampling_prob > 0 (default). **WARNING**: This is not a faithful implementation of scheduled sampling. This implementation samples tokens for timestep t condtioned on gold tokens 1...t-1. A proper implementation must condition on a mix of gold and sampled tokens. Doing so is not efficient for models such like Transformer. Args: features: {str: Tensor}. Features sharded along batch dimension. logits: Tensor. Logits for each shard of data. losses: 0-D Tensor or (num: 0-D Tensor, denom: 0-D Tensor). Loss Tensor Returns: new_logits: Tensor. new_losses: {str: loss} where loss is one of (i) a 0-D Tensor or (ii) a (num: 0-D Tensor, denom: 0-D Tensor) pair to be used in a weighted average. def maybe_scheduled_sampling(self, features, logits, losses): """Scheduled sampling. Performs forward inference again with "targets" feature replaced with values sampled from the model. This is the identity unless self.hparams.scheduled_sampling_prob > 0 (default). **WARNING**: This is not a faithful implementation of scheduled sampling. This implementation samples tokens for timestep t condtioned on gold tokens 1...t-1. A proper implementation must condition on a mix of gold and sampled tokens. Doing so is not efficient for models such like Transformer. Args: features: {str: Tensor}. Features sharded along batch dimension. logits: Tensor. Logits for each shard of data. losses: 0-D Tensor or (num: 0-D Tensor, denom: 0-D Tensor). Loss Tensor Returns: new_logits: Tensor. new_losses: {str: loss} where loss is one of (i) a 0-D Tensor or (ii) a (num: 0-D Tensor, denom: 0-D Tensor) pair to be used in a weighted average. """ hparams = self.hparams problem_hparams = self._problem_hparams # Only do scheduled sampling if requested. if hparams.scheduled_sampling_prob == 0.0: return (logits, losses) # Only do scheduled sampling on language tasks. modality = problem_hparams.modality["targets"] if modality != modalities.ModalityType.SYMBOL: assert hparams.scheduled_sampling_prob == 0, ( "Scheduled sampling only applies to ModalityType.SYMBOL. Set " "hparams.scheduled_sampling_prob == 0.0.") return (logits, losses) # Only do scheduled sampling when training. is_training = (hparams.mode == tf.estimator.ModeKeys.TRAIN) if not is_training: tf.logging.info("Running in %s mode. Not using scheduled sampling.", hparams.mode) return (logits, losses) # Pad vocabulary if vocab size must be evenly divisible by vocab_divisor. vocab_size = problem_hparams.vocab_size["targets"] assert vocab_size is not None assert hparams.vocab_divisor == 1 def sample(x): """Multinomial sampling from a n-dimensional tensor.""" samples = tf.multinomial(tf.reshape(x, [-1, vocab_size]), 1) reshaped_samples = tf.reshape(samples, common_layers.shape_list(x)[:-1]) return tf.to_int32(reshaped_samples) def mix_gold_sampled(gold_targets, sampled_targets, mixin_prob): """Interleave sampled and gold tokens randomly.""" return tf.where( tf.less( tf.random_uniform(common_layers.shape_list(sampled_targets)), mixin_prob), sampled_targets, gold_targets) def sampled_results(features, logits, mixin_prob): """Generate scheduled sampling results.""" sampled_targets = sample(logits) new_targets = mix_gold_sampled(features["targets"], sampled_targets, mixin_prob) new_targets = tf.stop_gradient(new_targets) # Treat new_targets as given. new_features = copy.copy(features) new_features["targets"] = new_targets with tf.variable_scope(tf.get_variable_scope(), reuse=True): # Compute bottom() for new_targets. # # TODO(duckworthd): Only apply bottom to 'new_targets'. new_transformed_features = self.bottom(new_features) # Compute body. with tf.variable_scope("body"): new_body_outputs, new_losses = self._normalize_body_output( self.body(new_transformed_features)) assert "training" not in new_losses # Compute top. new_logits = self.top(new_body_outputs, new_features) # Compute loss. Use original features (== labels). if (hparams.mode != tf.estimator.ModeKeys.PREDICT and hparams.mode != "attack"): new_losses["training"] = self.loss(new_logits, features) else: new_losses["training"] = 0.0 return new_logits, new_losses tf.logging.info("Using scheduled sampling.") assert hparams.scheduled_sampling_prob == 1.0, ( "hparams.scheduled_sampling_prob must be 0 or 1.") # Gradually increase over a warmup period. Lower numbers mean more gold # tokens. mixin_prob = ( hparams.scheduled_sampling_gold_mixin_prob * common_layers.inverse_exp_decay( hparams.scheduled_sampling_warmup_steps, min_value=0.001) ) # Apply scheduled sampling over N passes. The logits from the (n-1)-th pass # will be mixed with gold tokens for conditioning in the n-th pass. scheduled_sampling_num_passes = getattr( hparams, "scheduled_sampling_num_passes", 1) assert scheduled_sampling_num_passes > 0, ( "hparams.scheduled_sampling_num_passes must be > 0 if " "hparams.scheduled_sampling_prob > 0.0") new_logits = logits new_losses = losses for _ in range(scheduled_sampling_num_passes): new_logits, new_losses = sampled_results(features, new_logits, mixin_prob) return new_logits, new_losses
Prepare one shard of the model for the decoder. Args: targets: a Tensor. hparams: run hyperparameters Returns: decoder_input: a Tensor, bottom of decoder stack decoder_self_attention_bias: a Tensor, containing large negative values to implement masked attention and possibly biases for diagonal alignments pad_remover (expert_utils.PadRemover): an util object to remove padding def attention_lm_moe_prepare_decoder(targets, hparams): """Prepare one shard of the model for the decoder. Args: targets: a Tensor. hparams: run hyperparameters Returns: decoder_input: a Tensor, bottom of decoder stack decoder_self_attention_bias: a Tensor, containing large negative values to implement masked attention and possibly biases for diagonal alignments pad_remover (expert_utils.PadRemover): an util object to remove padding """ targets_pad_mask = common_attention.embedding_to_padding(targets) with tf.name_scope("pad_remover"): # Because of the shift_right, the <eos> token will be considered as # padding. In practice, it doesn't really matter, due to the triangular # mask, this token should never be attended. pad_remover = expert_utils.PadRemover(targets_pad_mask) if hparams.prepend_mode == "prepend_inputs_full_attention": decoder_self_attention_bias = ( common_attention.attention_bias_prepend_inputs_full_attention( targets_pad_mask)) else: decoder_self_attention_bias = ( common_attention.attention_bias_lower_triangle(tf.shape(targets)[1])) decoder_input = common_layers.shift_right_3d(targets) if hparams.pos == "timing": decoder_input = common_attention.add_timing_signal_1d(decoder_input) return (decoder_input, decoder_self_attention_bias, pad_remover)
Return a flat int32 tensor of shape [1, batch_size*length, 1]. def get_batch_coordinate(x, axis=0): """Return a flat int32 tensor of shape [1, batch_size*length, 1].""" # Compute the batch coordinate before flattening all batches batch_coordinate = tf.expand_dims( common_attention.coordinate_tensor(tf.shape(x)[:-1], axis=axis), axis=-1) return batch_coordinate
Duplicate elements of bc by length_factor. Args: bc (tf.Tensor): int32 tensor of shape [1, length, 1] length_factor (int): Returns: tf.Tensor: of shape [1, length*length_factor, 1] where every elements has been duplicated length_factor times. def expand_batch_coordinates(bc, length_factor): """Duplicate elements of bc by length_factor. Args: bc (tf.Tensor): int32 tensor of shape [1, length, 1] length_factor (int): Returns: tf.Tensor: of shape [1, length*length_factor, 1] where every elements has been duplicated length_factor times. """ assert bc.get_shape().as_list() == [1, None, 1] # bc has shape [1, length, 1] bc *= tf.constant([[1] * length_factor]) # bc has shape [1, length, length_factor] bc = tf.reshape(bc, [1, -1, 1]) # bc has shape [1, length*length_factor] return bc
Remove padding by concatenating all dimension into one. Args: x (tf.Tensor): input of shape [batch_size, length, depth] pad_remover (obj): a PadRemover object mode (ModeKeys): infer, train or eval. If inference, the padding remover is not applied Returns: tf.Tensor of shape [1,length_nonpad,depth] where length_nonpad <= batch_size*length def remove_pad(x, pad_remover, mode): """Remove padding by concatenating all dimension into one. Args: x (tf.Tensor): input of shape [batch_size, length, depth] pad_remover (obj): a PadRemover object mode (ModeKeys): infer, train or eval. If inference, the padding remover is not applied Returns: tf.Tensor of shape [1,length_nonpad,depth] where length_nonpad <= batch_size*length """ # Concatenate all tokens (without padding) x = expert_utils.flatten_all_but_last(x) # Remove padding for training and eval if mode != ModeKeys.PREDICT: # This is a hack to allows inference when the <go> token # is detected as padding and removed. This works for now because there is # no padding at inference. x = pad_remover.remove(x) x = tf.expand_dims(x, axis=0) # Now batch_size=1 return x
Set of hyperparameters. suitable for 1 gpu. on lm1b_32k: ~229M params 0.9 steps/sec on [GeForce GTX TITAN X] Returns: a hparams object def attention_lm_moe_base(): """Set of hyperparameters. suitable for 1 gpu. on lm1b_32k: ~229M params 0.9 steps/sec on [GeForce GTX TITAN X] Returns: a hparams object """ hparams = common_hparams.basic_params1() hparams.hidden_size = 1024 hparams.batch_size = 8192 hparams.max_length = 256 hparams.dropout = 0.0 hparams.clip_grad_norm = 0. # i.e. no gradient clipping hparams.optimizer_adam_epsilon = 1e-9 hparams.learning_rate_decay_scheme = "noam" hparams.learning_rate = 0.1 hparams.learning_rate_warmup_steps = 2000 hparams.initializer_gain = 1.0 hparams.num_hidden_layers = 4 hparams.initializer = "uniform_unit_scaling" hparams.weight_decay = 0.0 hparams.optimizer_adam_beta1 = 0.9 hparams.optimizer_adam_beta2 = 0.98 hparams.num_sampled_classes = 0 hparams.label_smoothing = 0.0 hparams.shared_embedding_and_softmax_weights = False hparams.add_hparam("filter_size", 2048) # Add new ones like this. hparams.moe_num_experts = 32 # attention-related flags hparams.add_hparam("num_heads", 8) hparams.add_hparam("attention_key_channels", 0) hparams.add_hparam("attention_value_channels", 0) # All hyperparameters ending in "dropout" are automatically set to 0.0 # when not in training mode. hparams.add_hparam("attention_dropout", 0.0) hparams.add_hparam("relu_dropout", 0.0) hparams.add_hparam("pos", "timing") # timing, none hparams.add_hparam("moe_layers", "2") # comma separated list of layer numbers # moe params. local attention moe. # If attention_layers is set, the num_hidden_layers parameter will be ignored # and each caracter of the string will correspond to one attention # layer type hparams.add_hparam("attention_layers", "") hparams.add_hparam("attention_type", AttentionType.MULTIHEAD) hparams.add_hparam("attention_local", False) hparams.add_hparam("attention_moe_k", 2) hparams.add_hparam("attention_num_head", 1) hparams.add_hparam("attention_num_experts", 16) hparams.add_hparam("attention_split_batch", False) hparams.add_hparam("attention_red_factor", 3) hparams.add_hparam("attention_block_length", 128) hparams.add_hparam("attention_reduction_type", "conv") # Non linearity for the attention reduction. Either "none", or "silu" ( # Sigmoid Linear-Unit described in https://arxiv.org/abs/1710.05941) hparams.add_hparam("attention_nonlinearity", "none") # If attention_exp_factor is set, each input to local_expert_attention (of # dimensionality hidden size) is projected into attention_exp_factor smaller # inputs, each of dimensionality attention_exp_inputdim. (otherwise # attention_exp_inputdim is ignored) hparams.add_hparam("attention_exp_factor", 0) hparams.add_hparam("attention_exp_inputdim", 128) # Key, query and value dimensions for the attention hparams.add_hparam("attention_kq_size", 128) hparams.add_hparam("attention_v_size", 256) # Loss coef for load balancing hparams.add_hparam("attention_load_balance", 2e-2) # Locality-sensitive hashing params hparams.add_hparam("lsh_num_hyperplanes", 4) hparams.add_hparam("lsh_use_map_fn", False) hparams.add_hparam("use_sepconv", False) hparams.add_hparam("diet_experts", False) hparams.add_hparam("memory_efficient_ffn", False) # if True, we learn a non-autoregressive model from "inputs" to "targets". # if False, we learn an autoregressive model to generate "targets" hparams.add_hparam("use_inputs", False) return hparams
Hyper parameters specifics for long sequence generation. def attention_lm_moe_base_long_seq(): """Hyper parameters specifics for long sequence generation.""" hparams = attention_lm_moe_base() hparams.max_length = 0 # max_length == batch_size hparams.eval_drop_long_sequences = True hparams.min_length_bucket = 256 # Avoid cyclic problems for big batches hparams.use_sepconv = True return hparams
Base model with attention expert. def attention_lm_moe_base_ae(): """Base model with attention expert.""" hparams = attention_lm_moe_base_long_seq() hparams.attention_type = AttentionType.LOCAL_EXPERTS hparams.learning_rate = 0.05 hparams.learning_rate_warmup_steps = 10000 # According to noam, ("n", "da") seems better for harder-to-learn models # hparams.layer_preprocess_sequence = "n" # hparams.layer_postprocess_sequence = "da" return hparams
Experiment with the exp_factor params. def attention_lm_ae_extended(): """Experiment with the exp_factor params.""" hparams = attention_lm_moe_base_long_seq() hparams.attention_layers = "eeee" hparams.attention_local = True # hparams.factored_logits=1 # Necessary when the number of expert grow bigger hparams.attention_moe_k = 2 hparams.attention_exp_factor = 4 # hparams.attention_exp_inputdim = 128 hparams.layer_preprocess_sequence = "n" hparams.layer_postprocess_sequence = "da" return hparams
Base model with attention expert. def attention_lm_moe_base_memeff(): """Base model with attention expert.""" hparams = attention_lm_moe_base_long_seq() hparams.use_sepconv = False hparams.diet_experts = True hparams.layer_preprocess_sequence = "n" hparams.layer_postprocess_sequence = "da" hparams.layer_prepostprocess_dropout = 0.0 hparams.memory_efficient_ffn = True hparams.attention_type = AttentionType.MEMORY_EFFICIENT hparams.num_heads = 8 hparams.factored_logits = True return hparams
Cheap model for single-gpu training. on lm1b_32k: ~312M params 1.6 steps/sec on [GeForce GTX TITAN X] After 50K steps on 8 GPUs (synchronous): eval_log_ppl_per_token = 3.31 Returns: an hparams object. def attention_lm_moe_small(): """Cheap model for single-gpu training. on lm1b_32k: ~312M params 1.6 steps/sec on [GeForce GTX TITAN X] After 50K steps on 8 GPUs (synchronous): eval_log_ppl_per_token = 3.31 Returns: an hparams object. """ hparams = attention_lm_moe_base() hparams.num_hidden_layers = 4 hparams.hidden_size = 512 hparams.filter_size = 2048 hparams.moe_num_experts = 128 hparams.moe_layers = "2" return hparams
Cheap model for debugging. Returns: an hparams object. def attention_lm_attention_moe_tiny(): """Cheap model for debugging. Returns: an hparams object. """ hparams = attention_lm_moe_small() hparams.moe_layers = "" hparams.attention_num_experts = 128 hparams.filter_size = 8192 hparams.attention_type = AttentionType.LOCAL_EXPERTS return hparams
Large model for distributed training. Over 1B parameters, so requires multi-gpu training due to memory requirements. on lm1b_32k: After 45K steps on 8 GPUs (synchronous): eval_log_ppl_per_token = 3.18 eval_ppl_per_word = exp(1.107893 * eval_log_ppl_per_token) = 33.9 Returns: an hparams object. def attention_lm_moe_large(): """Large model for distributed training. Over 1B parameters, so requires multi-gpu training due to memory requirements. on lm1b_32k: After 45K steps on 8 GPUs (synchronous): eval_log_ppl_per_token = 3.18 eval_ppl_per_word = exp(1.107893 * eval_log_ppl_per_token) = 33.9 Returns: an hparams object. """ hparams = attention_lm_moe_base() hparams.num_hidden_layers = 5 hparams.moe_layers = "3" hparams.hidden_size = 1024 hparams.num_heads = 16 hparams.filter_size = 4096 hparams.moe_hidden_sizes = "4096" hparams.moe_num_experts = 128 hparams.layer_prepostprocess_dropout = 0.2 return hparams
Memory-efficient version. def attention_lm_moe_memory_efficient(): """Memory-efficient version.""" hparams = attention_lm_moe_large() hparams.diet_experts = True hparams.layer_preprocess_sequence = "n" hparams.layer_postprocess_sequence = "da" hparams.layer_prepostprocess_dropout = 0.0 hparams.memory_efficient_ffn = True hparams.attention_type = AttentionType.MEMORY_EFFICIENT hparams.num_heads = 8 hparams.factored_logits = True return hparams
Unnecessarily large model with 24B params - because we can. def attention_lm_moe_24b_diet(): """Unnecessarily large model with 24B params - because we can.""" hparams = attention_lm_moe_large_diet() hparams.moe_hidden_sizes = "12288" hparams.moe_num_experts = 1024 hparams.batch_size = 4096 return hparams
Version to use for seq2seq. def attention_lm_moe_translation(): """Version to use for seq2seq.""" hparams = attention_lm_moe_base() hparams.layer_preprocess_sequence = "n" hparams.layer_postprocess_sequence = "da" hparams.learning_rate = 0.4 hparams.prepend_mode = "prepend_inputs_masked_attention" hparams.max_length = 512 hparams.label_smoothing = 0.1 hparams.layer_prepostprocess_dropout = 0.2 hparams.num_hidden_layers = 6 hparams.moe_layers = "0,1,2,3,4,5" hparams.shared_embedding_and_softmax_weights = True return hparams
Version to use with languagemodel_wiki_scramble1k50. def attention_lm_moe_unscramble_base(): """Version to use with languagemodel_wiki_scramble1k50.""" hparams = attention_lm_no_moe_small() hparams.use_inputs = True hparams.min_length_bucket = 1024 hparams.max_length = 1024 hparams.batch_size = 5000 hparams.layer_prepostprocess_dropout = 0.0 hparams.layer_preprocess_sequence = "n" hparams.layer_postprocess_sequence = "da" return hparams
Transform input from data space to model space. Args: x: A Tensor with shape [batch, ...] model_hparams: HParams, model hyperparmeters. vocab_size: int, vocabulary size. Returns: body_input: A Tensor with shape [batch, ?, ?, model_hparams.hidden_size]. def audio_bottom(x, model_hparams, vocab_size): """Transform input from data space to model space. Args: x: A Tensor with shape [batch, ...] model_hparams: HParams, model hyperparmeters. vocab_size: int, vocabulary size. Returns: body_input: A Tensor with shape [batch, ?, ?, model_hparams.hidden_size]. """ del vocab_size # unused arg inputs = x with tf.variable_scope("audio_modality"): # TODO(aidangomez): Will need to sort out a better audio pipeline def xnet_resblock(x, filters, res_relu, name): """Xception block.""" with tf.variable_scope(name): # Typically audio samples are >100k samples in length and have a width # of 2 or 4. Mono audio has a single channel while stereo has 2. y = common_layers.separable_conv_block( x, filters, [((1, 1), (3, 3)), ((1, 1), (3, 3))], first_relu=True, padding="SAME", force2d=True, name="sep_conv_block") y = common_layers.pool(y, (3, 3), "MAX", "SAME", strides=(2, 2)) return y + common_layers.conv_block( x, filters, [((1, 1), (1, 1))], padding="SAME", strides=(2, 2), first_relu=res_relu, force2d=True, name="res_conv0") x = tf.to_float(inputs) / 255. x.set_shape([None, None, None, 1]) for i in range(model_hparams.audio_compression): x = xnet_resblock(x, 2**(i + 1), True, "compress_block_%d" % i) return xnet_resblock(x, model_hparams.hidden_size, False, "compress_block_final")
Bottom transformation for target images. def image_targets_bottom(x, model_hparams, vocab_size): """Bottom transformation for target images.""" pixel_embedding_size = 64 inputs = x with tf.variable_scope("image_modality"): if not tf.executing_eagerly(): tf.summary.image( "targets_bottom", common_layers.tpu_safe_image_summary(inputs), max_outputs=1) inputs_shape = common_layers.shape_list(inputs) if len(inputs_shape) != 4: raise ValueError("Assuming images given as int tensors in the format " "[batch, height, width, channels] (256 values).") # We embed each of 256=vocab_size possible pixel values. embedding_var = tf.get_variable( "pixel_embedding", [vocab_size, pixel_embedding_size]) hot_inputs = tf.one_hot(tf.to_int32(inputs), vocab_size) hot_inputs = tf.reshape(hot_inputs, [-1, vocab_size]) embedded = tf.matmul(hot_inputs, embedding_var) # Let's now merge all channels that were embedded into a single vector. merged_size = pixel_embedding_size * inputs_shape[3] embedded = tf.reshape(embedded, inputs_shape[:3] + [merged_size]) merged = tf.layers.dense( embedded, model_hparams.hidden_size, name="merge_pixel_embedded_channels") return merged
Compresses channel-wise input pixels into whole pixel representions. Perform conversion of RGB pixel values to a real number in the range -1 to 1. This combines pixel channels to form a representation of shape [img_len, img_len]. Args: inputs: Tensor representing RGB pixel intensities as integers, of shape [batch, img_len, img_len, channels]. model_hparams: HParams, model hyperparmeters. name: string, scope. Returns: body_input: Tensor of shape [batch, img_len, img_len, model_hparams.hidden_size]. def _image_channel_compress_bottom(inputs, model_hparams, name="bottom"): """Compresses channel-wise input pixels into whole pixel representions. Perform conversion of RGB pixel values to a real number in the range -1 to 1. This combines pixel channels to form a representation of shape [img_len, img_len]. Args: inputs: Tensor representing RGB pixel intensities as integers, of shape [batch, img_len, img_len, channels]. model_hparams: HParams, model hyperparmeters. name: string, scope. Returns: body_input: Tensor of shape [batch, img_len, img_len, model_hparams.hidden_size]. """ num_channels = 3 with tf.variable_scope(name): inputs = tf.to_float(inputs) hp = model_hparams if hp.mode != tf.estimator.ModeKeys.PREDICT: tf.summary.image( "inputs", common_layers.tpu_safe_image_summary(inputs), max_outputs=2) inputs = common_layers.convert_rgb_to_symmetric_real(inputs) # Reshape inputs to apply convolutions across [img_len, img_len*channels]. inputs_shape = common_layers.shape_list(inputs) inputs = tf.reshape( inputs, [-1, inputs_shape[1], inputs_shape[2] * inputs_shape[3], 1]) # Compress RGB intensities for each pixel using a convolution. outputs = tf.layers.conv2d( inputs, model_hparams.hidden_size, kernel_size=(1, num_channels), padding="VALID", strides=(1, num_channels), activation=tf.nn.relu, name="conv_input") return outputs
Bottom transformation for image targets. def image_channel_embeddings_bottom(x, model_hparams, vocab_size): """Bottom transformation for image targets.""" del vocab_size # unused arg inputs = tf.to_int32(x) io_depth = model_hparams.num_channels tshape = common_layers.shape_list(inputs) hidden_size = model_hparams.hidden_size target_embeddings = cia.get_channel_embeddings( io_depth, inputs, hidden_size, "input_bottom") return tf.reshape(target_embeddings, [tshape[0], tshape[1], tshape[2] * io_depth, hidden_size])
Use batchnorm instead of CMVN and shorten the stft with strided convs. Args: x: float32 tensor with shape [batch_size, len, 1, freqs * channels] model_hparams: HParams, model hyperparmeters. vocab_size: int, vocabulary size. Returns: float32 tensor with shape [batch_size, shorter_len, 1, hidden_size] def speech_recognition_bottom(x, model_hparams, vocab_size): """Use batchnorm instead of CMVN and shorten the stft with strided convs. Args: x: float32 tensor with shape [batch_size, len, 1, freqs * channels] model_hparams: HParams, model hyperparmeters. vocab_size: int, vocabulary size. Returns: float32 tensor with shape [batch_size, shorter_len, 1, hidden_size] """ del vocab_size # unused arg inputs = x p = model_hparams num_mel_bins = p.audio_num_mel_bins num_channels = 3 if p.audio_add_delta_deltas else 1 with tf.variable_scope("speech_recognition_modality"): if p.audio_preproc_in_bottom: # Compute filterbanks with tf.variable_scope("fbanks"): waveforms = tf.squeeze(inputs, [2, 3]) mel_fbanks = common_audio.compute_mel_filterbank_features( waveforms, sample_rate=p.audio_sample_rate, dither=p.audio_dither, preemphasis=p.audio_preemphasis, frame_length=p.audio_frame_length, frame_step=p.audio_frame_step, lower_edge_hertz=p.audio_lower_edge_hertz, upper_edge_hertz=p.audio_upper_edge_hertz, num_mel_bins=p.audio_num_mel_bins, apply_mask=True) if p.audio_add_delta_deltas: mel_fbanks = common_audio.add_delta_deltas(mel_fbanks) x = tf.reshape(mel_fbanks, common_layers.shape_list(mel_fbanks)[:2] + [num_mel_bins, num_channels]) nonpadding_mask = 1. - common_attention.embedding_to_padding(x) num_of_nonpadding_elements = tf.reduce_sum( nonpadding_mask) * num_mel_bins * num_channels # This replaces CMVN estimation on data var_epsilon = 1e-09 mean = tf.reduce_sum( x, axis=[1], keepdims=True) / num_of_nonpadding_elements variance = (num_of_nonpadding_elements * mean**2. - 2. * mean * tf.reduce_sum(x, axis=[1], keepdims=True) + tf.reduce_sum(x**2, axis=[1], keepdims=True) ) / num_of_nonpadding_elements x = (x - mean) * tf.rsqrt(variance + var_epsilon) * tf.expand_dims( nonpadding_mask, -1) else: x = inputs # The convention is that the models are flattened along the spatial, # dimensions, thus the speech preprocessor treats frequencies and # channels as image colors (last axis) x.set_shape([None, None, num_mel_bins, num_channels]) # TODO(chorowski): how to specify bottom's hparams and avoid hardcoding? x = tf.pad(x, [[0, 0], [0, 8], [0, 0], [0, 0]]) for _ in range(2): x = tf.layers.conv2d( x, 128, (3, 3), (2, 2), use_bias=False) x = common_layers.layer_norm(x) x = tf.nn.relu(x) xshape = common_layers.shape_list(x) # apply a conv that will remove all frequencies and at the same time # project the output into desired hidden_size x = tf.pad(x, [[0, 0], [0, 2], [0, 0], [0, 0]]) x = tf.layers.conv2d(x, p.hidden_size, (3, xshape[2]), use_bias=False) assert common_layers.shape_list(x)[2] == 1 x = common_layers.layer_norm(x) x = tf.nn.relu(x) return x
Create or get concatenated embedding or softmax variable. Args: model_hparams: HParams, model hyperparmeters. vocab_size: int, vocabulary size. hidden_dim: dim of the variable. Defaults to _model_hparams' hidden_size Returns: a list of num_shards Tensors. def get_weights(model_hparams, vocab_size, hidden_dim=None): """Create or get concatenated embedding or softmax variable. Args: model_hparams: HParams, model hyperparmeters. vocab_size: int, vocabulary size. hidden_dim: dim of the variable. Defaults to _model_hparams' hidden_size Returns: a list of num_shards Tensors. """ if hidden_dim is None: hidden_dim = model_hparams.hidden_size num_shards = model_hparams.symbol_modality_num_shards shards = [] for i in range(num_shards): shard_size = (vocab_size // num_shards) + ( 1 if i < vocab_size % num_shards else 0) var_name = "weights_%d" % i shards.append( tf.get_variable( var_name, [shard_size, hidden_dim], initializer=tf.random_normal_initializer(0.0, hidden_dim**-0.5))) if num_shards == 1: ret = shards[0] else: ret = tf.concat(shards, 0) # Convert ret to tensor. if not tf.executing_eagerly(): ret = common_layers.convert_gradient_to_tensor(ret) return ret
Bottom transformation for symbols. def _symbol_bottom_simple(x, model_hparams, vocab_size, name, reuse): """Bottom transformation for symbols.""" with tf.variable_scope(name, reuse=reuse): # Ensure the inputs are 3-D if len(x.get_shape()) == 4: x = tf.squeeze(x, axis=3) while len(x.get_shape()) < 3: x = tf.expand_dims(x, axis=-1) var = get_weights(model_hparams, vocab_size) x = common_layers.dropout_no_scaling( x, 1.0 - model_hparams.symbol_dropout) ret = common_layers.gather(var, x) if model_hparams.multiply_embedding_mode == "sqrt_depth": ret *= model_hparams.hidden_size**0.5 ret *= tf.expand_dims( common_layers.cast_like(tf.not_equal(x, 0), ret), -1) return ret
Bottom transformation for target symbols. def symbol_targets_bottom(x, model_hparams, vocab_size): """Bottom transformation for target symbols.""" if (model_hparams.shared_embedding_and_softmax_weights or model_hparams.get("shared_embedding")): try: return _symbol_bottom_simple( x, model_hparams, vocab_size, "shared", reuse=True) except ValueError: # perhaps there were no inputs, and this is a new variable. return _symbol_bottom_simple( x, model_hparams, vocab_size, "shared", reuse=None) else: return _symbol_bottom_simple( x, model_hparams, vocab_size, "target_emb", reuse=None)
Bottom transformation for embedding video bitwise. def video_bitwise_bottom(x, model_hparams, vocab_size): """Bottom transformation for embedding video bitwise.""" pixel_embedding_size = 64 inputs = x with tf.variable_scope("video_modality_bitwise", reuse=tf.AUTO_REUSE): common_layers.summarize_video(inputs, "bottom") # Embed bitwise. assert vocab_size == 256 embedded = discretization.int_to_bit_embed(inputs, 8, pixel_embedding_size) # Project. return tf.layers.dense( embedded, model_hparams.hidden_size, name="merge_pixel_embedded_frames")
Bottom transformation for video. def video_pixel_noise_bottom(x, model_hparams, vocab_size): """Bottom transformation for video.""" input_noise = getattr(model_hparams, "video_modality_input_noise", 0.25) inputs = x if model_hparams.mode == tf.estimator.ModeKeys.TRAIN: background = tfp.stats.percentile(inputs, 50., axis=[0, 1, 2, 3]) input_shape = common_layers.shape_list(inputs) input_size = tf.reduce_prod(input_shape[:-1]) input_mask = tf.multinomial( tf.log([[input_noise, 1.-input_noise]]), input_size) input_mask = tf.reshape(tf.cast(input_mask, tf.int32), input_shape[:-1]+[1]) inputs = inputs * input_mask + background * (1 - input_mask) return video_bottom(inputs, model_hparams, vocab_size)
Convert prediction and target from rgb to real. def convert_rgb_to_real(prediction, targets): """Convert prediction and target from rgb to real.""" prediction = tf.squeeze(prediction, axis=-1) prediction = common_layers.convert_rgb_to_real(prediction) targets = common_layers.convert_rgb_to_real(targets) return prediction, targets
Compute the CTC loss. def ctc_symbol_loss(top_out, targets, model_hparams, vocab_size, weight_fn): """Compute the CTC loss.""" del model_hparams, vocab_size # unused arg logits = top_out with tf.name_scope("ctc_loss", values=[logits, targets]): # For CTC we assume targets are 1d, [batch, length, 1, 1] here. targets_shape = targets.get_shape().as_list() assert len(targets_shape) == 4 assert targets_shape[2] == 1 assert targets_shape[3] == 1 targets = tf.squeeze(targets, axis=[2, 3]) logits = tf.squeeze(logits, axis=[2, 3]) targets_mask = 1 - tf.to_int32(tf.equal(targets, 0)) targets_lengths = tf.reduce_sum(targets_mask, axis=1) sparse_targets = tf.keras.backend.ctc_label_dense_to_sparse( targets, targets_lengths) xent = tf.nn.ctc_loss( sparse_targets, logits, targets_lengths, time_major=False, preprocess_collapse_repeated=False, ctc_merge_repeated=False) weights = weight_fn(targets) return tf.reduce_sum(xent), tf.reduce_sum(weights)
Compute loss numerator and denominator for one shard of output. def generic_loss(top_out, targets, model_hparams, vocab_size, weights_fn): """Compute loss numerator and denominator for one shard of output.""" del vocab_size # unused arg logits = top_out logits = common_attention.maybe_upcast(logits, hparams=model_hparams) cutoff = getattr(model_hparams, "video_modality_loss_cutoff", 0.0) return common_layers.padded_cross_entropy( logits, targets, model_hparams.label_smoothing, cutoff=cutoff, weights_fn=weights_fn)
Average loss over the labels. def multi_label_loss(top_out, targets, model_hparams, vocab_size, weights_fn): """Average loss over the labels.""" del vocab_size # unused arg logits = top_out num_labels = tf.shape(targets)[1] logits = tf.tile(logits, [1, num_labels, 1, 1, 1]) xent, weights = common_layers.padded_cross_entropy( logits, targets, model_hparams.label_smoothing, weights_fn=weights_fn, reduce_sum=False, ) xent = tf.squeeze(xent, [2, 3]) weights = tf.squeeze(weights, [2, 3]) # average loss over all labels loss = tf.reduce_sum(xent, axis=1) weights = tf.reduce_sum(weights, axis=1) loss /= (weights + 1e-8) weights = tf.to_float(tf.greater(weights, 0.)) return tf.reduce_sum(loss*weights), tf.reduce_sum(weights)
Apply softmax cross-entropy between outputs and targets. Args: top_out: logits Tensor with shape [batch, ?, ?, num_classes] targets: one-hot encoding Tensor with shape [batch, ?, ?, num_classes] model_hparams: HParams, model hyperparmeters. vocab_size: int, vocabulary size. weights_fn: Returns: loss_scale (cross-entropy), loss_denom def one_hot_class_label_loss(top_out, targets, model_hparams, vocab_size, weights_fn): """Apply softmax cross-entropy between outputs and targets. Args: top_out: logits Tensor with shape [batch, ?, ?, num_classes] targets: one-hot encoding Tensor with shape [batch, ?, ?, num_classes] model_hparams: HParams, model hyperparmeters. vocab_size: int, vocabulary size. weights_fn: Returns: loss_scale (cross-entropy), loss_denom """ del model_hparams, vocab_size # unused arg loss_scale = tf.losses.softmax_cross_entropy( onehot_labels=targets, logits=top_out) weights = weights_fn(targets) loss_denom = tf.reduce_sum(weights) return loss_scale, loss_denom
Poisson loss for real. def real_log_poisson_loss(top_out, targets, model_hparams, vocab_size, weights_fn): """Poisson loss for real.""" del model_hparams, vocab_size # unused arg predictions = top_out if (len(common_layers.shape_list(top_out)) != len( common_layers.shape_list(targets))): predictions = tf.squeeze(top_out, axis=[-1]) with tf.name_scope("log_possion"): weights = weights_fn(targets) lp_loss = tf.nn.log_poisson_loss(targets, predictions) return tf.reduce_sum(lp_loss * weights), tf.reduce_sum(weights)
Loss for class label. def sigmoid_class_label_loss(top_out, targets, model_hparams, vocab_size, weights_fn): """Loss for class label.""" # Expect inputs of size [batch-size, timesteps, 1, num-classes], where the # last dimension of num-classes represents logits for binary labels del model_hparams, vocab_size # unused arg loss_scale = tf.losses.sigmoid_cross_entropy( multi_class_labels=targets, logits=top_out) weights = weights_fn(targets) loss_denom = tf.reduce_sum(weights) return loss_scale, loss_denom
Compute loss numerator and denominator for one shard of output. def video_loss(top_out, targets, model_hparams, vocab_size, weights_fn): """Compute loss numerator and denominator for one shard of output.""" del vocab_size # unused arg logits = top_out logits = tf.reshape(logits, [-1] + common_layers.shape_list(logits)[2:]) targets = tf.reshape(targets, [-1] + common_layers.shape_list(targets)[2:]) cutoff = getattr(model_hparams, "video_modality_loss_cutoff", 0.01) return common_layers.padded_cross_entropy( logits, targets, model_hparams.label_smoothing, cutoff=cutoff, weights_fn=weights_fn)
Compute loss numerator and denominator for one shard of output. def video_l1_loss(top_out, targets, model_hparams, vocab_size, weights_fn): """Compute loss numerator and denominator for one shard of output.""" del vocab_size # unused arg logits = top_out logits = tf.reshape(logits, [-1] + common_layers.shape_list(logits)[2:-1]) targets = tf.reshape(targets, [-1] + common_layers.shape_list(targets)[2:]) weights = weights_fn(targets) # Shift targets by 0.5 so later just casting to int gives the prediction. # So for int targets, say 0 and 7, we actually train to predict 0.5 and 7.5. # Later (in merics or infer) this is cast to int anyway. Also, we have no # loss beyond cutoff = 0.2 as these are already correct predictions. targets = tf.to_float(targets) + 0.5 loss = video_l1_internal_loss(logits, targets, model_hparams) return tf.reduce_sum(loss * weights), tf.reduce_sum(weights)
Compute loss numerator and denominator for one shard of output. def video_l2_loss(top_out, targets, model_hparams, vocab_size, weights_fn): """Compute loss numerator and denominator for one shard of output.""" del vocab_size # unused arg logits = top_out logits = tf.reshape(logits, [-1] + common_layers.shape_list(logits)[2:-1]) targets = tf.reshape(targets, [-1] + common_layers.shape_list(targets)[2:]) weights = weights_fn(targets) # Shift targets by 0.5 so later just casting to int gives the prediction. # So for int targets, say 0 and 7, we actually train to predict 0.5 and 7.5. # Later (in merics or infer) this is cast to int anyway. Also, we have no # loss beyond cutoff = 0.2 as these are already correct predictions. targets = tf.to_float(targets) + 0.5 loss = video_l2_internal_loss(logits, targets, model_hparams) return tf.reduce_sum(loss * weights), tf.reduce_sum(weights)
Transform inputs from model space to target space. Average over inner dims and a linear layer to logits. Args: body_output: A Tensor with shape [batch, ?, ?, body_output_size]. targets: model_hparams: HParams, model hyperparmeters. vocab_size: int, vocabulary size. Returns: a Tensors, each with shape [batch_size, 1, 1, 1, vocab_size] def class_label_top(body_output, targets, model_hparams, vocab_size): """Transform inputs from model space to target space. Average over inner dims and a linear layer to logits. Args: body_output: A Tensor with shape [batch, ?, ?, body_output_size]. targets: model_hparams: HParams, model hyperparmeters. vocab_size: int, vocabulary size. Returns: a Tensors, each with shape [batch_size, 1, 1, 1, vocab_size] """ del targets # unused arg with tf.variable_scope("class_label_modality_%d_%d" % ( vocab_size, model_hparams.hidden_size)): x = body_output x = tf.reduce_mean(x, axis=[1, 2], keepdims=True) res = tf.layers.dense(x, vocab_size) return tf.expand_dims(res, 3)
Top transformation for images. def image_top(body_output, targets, model_hparams, vocab_size): """Top transformation for images.""" del targets # unused arg # TODO(lukaszkaiser): is this a universal enough way to get channels? num_channels = model_hparams.problem.num_channels with tf.variable_scope("rgb_softmax"): body_output_shape = common_layers.shape_list(body_output) reshape_shape = body_output_shape[:3] reshape_shape.extend([num_channels, vocab_size]) res = tf.layers.dense(body_output, vocab_size * num_channels) res = tf.reshape(res, reshape_shape) if not tf.get_variable_scope().reuse: res_argmax = tf.argmax(res, axis=-1) tf.summary.image( "result", common_layers.tpu_safe_image_summary(res_argmax), max_outputs=1) return res
Transforms body output to return logits. Args: body_output: Tensor of shape [batch, img_len, img_len, depth]. targets: model_hparams: HParams, model hyperparmeters. vocab_size: int, vocabulary size. Returns: Tensor of shape [batch, img_len, img_len, channels, vocab_size]. def image_channel_compress_top(body_output, targets, model_hparams, vocab_size): """Transforms body output to return logits. Args: body_output: Tensor of shape [batch, img_len, img_len, depth]. targets: model_hparams: HParams, model hyperparmeters. vocab_size: int, vocabulary size. Returns: Tensor of shape [batch, img_len, img_len, channels, vocab_size]. """ del targets # unused arg with tf.variable_scope("image_channel_compress_modality"): hidden_size = model_hparams.hidden_size img_len = model_hparams.img_len channels = 3 # RGB batch = common_layers.shape_list(body_output)[0] x = tf.layers.conv2d( body_output, hidden_size * channels, kernel_size=(1, 1), strides=(1, 1), padding="VALID", activation=tf.nn.relu, name="decompress_conv") x = tf.reshape(x, [batch, img_len, img_len * channels, hidden_size]) x = common_layers.layer_preprocess(x, model_hparams) x = tf.layers.dense(x, vocab_size, use_bias=True, activation=None, name="output_conv") x = tf.reshape( x, [batch, img_len, img_len, channels, vocab_size]) return x
Top transformation for images. def image_channel_embeddings_top(body_output, targets, model_hparams, vocab_size): """Top transformation for images.""" del targets # unused arg with tf.variable_scope("image_channel_embeddings_bottom"): img_len = model_hparams.img_len channels = model_hparams.num_channels x = tf.layers.dense( body_output, 256, use_bias=True, activation=None, name="output_conv") x = tf.reshape(x, [-1, img_len, img_len, channels, vocab_size]) return x
Loss for class label. def softmax_average_pooling_class_label_top(body_output, targets, model_hparams, vocab_size): """Loss for class label.""" del targets # unused arg with tf.variable_scope( "softmax_average_pooling_onehot_class_label_modality_%d_%d" % ( vocab_size, model_hparams.hidden_size)): x = body_output x = tf.reduce_mean(x, axis=1, keepdims=True) return tf.layers.dense(x, vocab_size)
Loss for class label. def softmax_last_timestep_class_label_top(body_output, targets, model_hparams, vocab_size): """Loss for class label.""" del targets # unused arg with tf.variable_scope( "softmax_last_timestep_onehot_class_label_modality_%d_%d" % ( vocab_size, model_hparams.hidden_size)): x = body_output x = tf.expand_dims(x[:, -1], 1) # Pick the last timestep return tf.layers.dense(x, vocab_size)
Loss for class label. def softmax_max_pooling_class_label_top(body_output, targets, model_hparams, vocab_size): """Loss for class label.""" del targets # unused arg with tf.variable_scope( "softmax_max_pooling_onehot_class_label_modality_%d_%d" % ( vocab_size, model_hparams.hidden_size)): x = body_output x = tf.reduce_max(x, axis=1, keepdims=True) return tf.layers.dense(x, vocab_size)
Generate logits. Args: body_output: A Tensor with shape [batch, p0, p1, model_hparams.hidden_size]. targets: Unused. model_hparams: HParams, model hyperparmeters. vocab_size: int, vocabulary size. Returns: logits: A Tensor with shape [batch, p0, p1, ?, vocab_size]. def symbol_top(body_output, targets, model_hparams, vocab_size): """Generate logits. Args: body_output: A Tensor with shape [batch, p0, p1, model_hparams.hidden_size]. targets: Unused. model_hparams: HParams, model hyperparmeters. vocab_size: int, vocabulary size. Returns: logits: A Tensor with shape [batch, p0, p1, ?, vocab_size]. """ del targets # unused arg if model_hparams.shared_embedding_and_softmax_weights: scope_name = "shared" reuse = tf.AUTO_REUSE else: scope_name = "softmax" reuse = False with tf.variable_scope(scope_name, reuse=reuse): body_output_shape = common_layers.shape_list(body_output) var = get_weights(model_hparams, vocab_size, body_output_shape[-1]) if (model_hparams.factored_logits and model_hparams.mode == tf.estimator.ModeKeys.TRAIN): # insert channels dimension body_output = tf.expand_dims(body_output, 3) return common_layers.FactoredTensor(body_output, var) else: body_output = tf.reshape(body_output, [-1, body_output_shape[-1]]) logits = tf.matmul(body_output, var, transpose_b=True) return tf.reshape(logits, body_output_shape[:-1] + [1, vocab_size])
Top transformation for video. def video_top(body_output, targets, model_hparams, vocab_size): """Top transformation for video.""" del targets # unused arg num_channels = model_hparams.problem.num_channels shape = common_layers.shape_list(body_output) reshape_shape = shape[:-1] + [num_channels, vocab_size] res = tf.reshape(body_output, reshape_shape) # Calculate argmax so as to have a summary with the produced images. x = tf.argmax(tf.reshape(res, [-1, vocab_size]), axis=-1) x = tf.reshape(x, shape[:-1] + [num_channels]) common_video.gif_summary("results", x, max_outputs=1) return res
Top transformation for video. def video_l1_top(body_output, targets, model_hparams, vocab_size): """Top transformation for video.""" del targets, vocab_size # unused arg num_channels = model_hparams.problem.num_channels num_frames = model_hparams.video_num_target_frames with tf.variable_scope("rgb"): body_output_shape = common_layers.shape_list(body_output) res = tf.layers.dense(body_output, num_channels * num_frames, name="cast") res = tf.reshape(res, body_output_shape[:3] + [num_channels, num_frames]) res = tf.transpose(res, [0, 4, 1, 2, 3]) # Move frames next to batch. if not tf.get_variable_scope().reuse: res_argmax = res[:, -1, :, :, :] tf.summary.image( "result", common_layers.tpu_safe_image_summary(res_argmax), max_outputs=1) return tf.expand_dims(res, axis=-1)
Gets default bottom transformation; if none available, return value. def get_bottom(modality_type, value=None): """Gets default bottom transformation; if none available, return value.""" if modality_type == ModalityType.AUDIO: return audio_bottom elif modality_type == ModalityType.AUDIO_SPECTRAL: return audio_spectral_bottom elif modality_type in (ModalityType.CLASS_LABEL, ModalityType.MULTI_LABEL, ModalityType.ONE_HOT_CLASS_LABEL, ModalityType.SIGMOID_CLASS_LABEL, ModalityType.SIGMOID_MAX_POOLING_CLASS_LABEL, ModalityType.SOFTMAX_AVERAGE_POOLING_CLASS_LABEL, ModalityType.SOFTMAX_LAST_TIMESTEP_CLASS_LABEL, ModalityType.SOFTMAX_MAX_POOLING_CLASS_LABEL): return class_label_bottom elif modality_type in (ModalityType.CTC_SYMBOL, ModalityType.SYMBOL, ModalityType.SYMBOL_WEIGHTS_ALL): return symbol_bottom elif modality_type in (ModalityType.GENERIC_L2_LOSS, ModalityType.IDENTITY, ModalityType.IDENTITY_SYMBOL, ModalityType.IMAGE_CHANNEL_EMBEDDINGS_BOTTOM): return identity_bottom elif modality_type == ModalityType.IMAGE: return image_bottom elif modality_type in (ModalityType.IMAGE_CHANNEL_BOTTOM_IDENTITY, ModalityType.IMAGE_CHANNEL_COMPRESS): return image_channel_compress_bottom elif modality_type in (ModalityType.REAL, ModalityType.REAL_L2_LOSS, ModalityType.REAL_LOG_POISSON_LOSS): return real_bottom elif modality_type == ModalityType.SPEECH_RECOGNITION: return speech_recognition_bottom elif modality_type == ModalityType.SYMBOL_ONE_HOT: return symbol_one_hot_bottom elif modality_type in (ModalityType.VIDEO, ModalityType.VIDEO_L1, ModalityType.VIDEO_L2): return video_bottom elif modality_type == ModalityType.VIDEO_BITWISE: return video_bitwise_bottom elif modality_type == ModalityType.VIDEO_IDENTITY: return video_identity_bottom elif modality_type in (ModalityType.VIDEO_L1_RAW, ModalityType.VIDEO_L2_RAW): return video_raw_bottom elif modality_type == ModalityType.VIDEO_PIXEL_NOISE: return video_pixel_noise_bottom return value
Gets default loss transformation; if none available, return value. def get_loss(modality_type, value=None): """Gets default loss transformation; if none available, return value.""" if modality_type in (ModalityType.AUDIO, ModalityType.AUDIO_SPECTRAL, ModalityType.CLASS_LABEL, ModalityType.IDENTITY, ModalityType.IDENTITY_SYMBOL, ModalityType.IMAGE, ModalityType.IMAGE_CHANNEL_BOTTOM_IDENTITY, ModalityType.IMAGE_CHANNEL_COMPRESS, ModalityType.IMAGE_CHANNEL_EMBEDDINGS_BOTTOM, ModalityType.REAL, ModalityType.SPEECH_RECOGNITION, ModalityType.SYMBOL, ModalityType.SYMBOL_WEIGHTS_ALL): return generic_loss elif modality_type == ModalityType.CTC_SYMBOL: return ctc_symbol_loss elif modality_type == ModalityType.GENERIC_L2_LOSS: return generic_l2_loss elif modality_type == ModalityType.MULTI_LABEL: return multi_label_loss elif modality_type in (ModalityType.ONE_HOT_CLASS_LABEL, ModalityType.SOFTMAX_AVERAGE_POOLING_CLASS_LABEL, ModalityType.SOFTMAX_LAST_TIMESTEP_CLASS_LABEL, ModalityType.SOFTMAX_MAX_POOLING_CLASS_LABEL): return one_hot_class_label_loss elif modality_type == ModalityType.REAL_L2_LOSS: return real_l2_loss elif modality_type == ModalityType.REAL_LOG_POISSON_LOSS: return real_log_poisson_loss elif modality_type == ModalityType.SIGMOID_CLASS_LABEL: return sigmoid_class_label_loss elif modality_type == ModalityType.SIGMOID_MAX_POOLING_CLASS_LABEL: return sigmoid_max_pooling_class_label_loss elif modality_type == ModalityType.SYMBOL_ONE_HOT: return symbol_one_hot_loss elif modality_type in (ModalityType.VIDEO, ModalityType.VIDEO_BITWISE, ModalityType.VIDEO_PIXEL_NOISE): return video_loss elif modality_type == ModalityType.VIDEO_IDENTITY: return video_identity_loss elif modality_type == ModalityType.VIDEO_L1: return video_l1_loss elif modality_type == ModalityType.VIDEO_L1_RAW: return video_l1_raw_loss elif modality_type == ModalityType.VIDEO_L2: return video_l2_loss elif modality_type == ModalityType.VIDEO_L2_RAW: return video_l2_raw_loss return value
Gets default name for transformations; if none available, return value. def get_name(modality_type, value=None): """Gets default name for transformations; if none available, return value.""" # For legacy reasons, modalities vary in their naming scheme. Future plans are # to remove any need for get_name. We do not recommend using it. if modality_type == ModalityType.AUDIO: return lambda model_hparams, vocab_size: "audio_modality" elif modality_type == ModalityType.AUDIO_SPECTRAL: return lambda model_hparams, vocab_size: "audio_spectral_modality" elif modality_type == ModalityType.GENERIC_L2_LOSS: return lambda model_hparams, vocab_size: "generic_l2_loss_modality" elif modality_type == ModalityType.IDENTITY: return lambda model_hparams, vocab_size: "identity_modality" elif modality_type == ModalityType.IMAGE: return lambda model_hparams, vocab_size: "image_modality" elif modality_type == ModalityType.IMAGE_CHANNEL_BOTTOM_IDENTITY: return (lambda model_hparams, vocab_size: # pylint: disable=g-long-lambda "image_channel_bottom_identity_modality") elif modality_type == ModalityType.IMAGE_CHANNEL_COMPRESS: return lambda model_hparams, vocab_size: "image_channel_compress_modality" elif modality_type == ModalityType.IMAGE_CHANNEL_EMBEDDINGS_BOTTOM: return lambda model_hparams, vocab_size: "image_channel_embeddings_bottom" elif modality_type == ModalityType.REAL: return lambda model_hparams, vocab_size: "real_modality" elif modality_type == ModalityType.REAL_L2_LOSS: return lambda model_hparams, vocab_size: "real_l2_loss_modality" elif modality_type == ModalityType.REAL_LOG_POISSON_LOSS: return lambda model_hparams, vocab_size: "real_log_poisson_loss_modality" elif modality_type == ModalityType.SPEECH_RECOGNITION: return lambda model_hparams, vocab_size: "speech_recognition_modality" elif modality_type == ModalityType.VIDEO: return lambda model_hparams, vocab_size: "video_modality" elif modality_type == ModalityType.VIDEO_BITWISE: return lambda model_hparams, vocab_size: "video_modality_bitwise" elif modality_type == ModalityType.VIDEO_IDENTITY: return lambda model_hparams, vocab_size: "video_modality_identity" elif modality_type == ModalityType.VIDEO_L1: return lambda model_hparams, vocab_size: "video_modality_l1" elif modality_type == ModalityType.VIDEO_L1_RAW: return lambda model_hparams, vocab_size: "video_modality_l1_raw" elif modality_type == ModalityType.VIDEO_L2: return lambda model_hparams, vocab_size: "video_modality_l2" elif modality_type == ModalityType.VIDEO_L2_RAW: return lambda model_hparams, vocab_size: "video_modality_l2_raw" elif modality_type == ModalityType.VIDEO_PIXEL_NOISE: return lambda model_hparams, vocab_size: "video_modality_pixel_noise" elif modality_type in (ModalityType.CLASS_LABEL, ModalityType.MULTI_LABEL, ModalityType.ONE_HOT_CLASS_LABEL): def name(model_hparams, vocab_size): return "class_label_modality_%d_%d" % (vocab_size, model_hparams.hidden_size) return name elif modality_type in (ModalityType.CTC_SYMBOL, ModalityType.IDENTITY_SYMBOL, ModalityType.SYMBOL, ModalityType.SYMBOL_WEIGHTS_ALL, ModalityType.SYMBOL_ONE_HOT): def name(model_hparams, vocab_size): return "symbol_modality_%d_%d" % (vocab_size, model_hparams.hidden_size) return name elif modality_type == ModalityType.SIGMOID_CLASS_LABEL: def name(model_hparams, vocab_size): return "sigmoid_class_symbol_modality_%d_%d" % (vocab_size, model_hparams.hidden_size) return name elif modality_type == ModalityType.SIGMOID_MAX_POOLING_CLASS_LABEL: def name(model_hparams, vocab_size): return "sigmoid_max_pooling_class_symbol_modality_%d_%d" % ( vocab_size, model_hparams.hidden_size) return name elif modality_type == ModalityType.SOFTMAX_AVERAGE_POOLING_CLASS_LABEL: def name(model_hparams, vocab_size): return "softmax_average_pooling_onehot_class_label_modality_%d_%d" % ( vocab_size, model_hparams.hidden_size) return name elif modality_type == ModalityType.SOFTMAX_LAST_TIMESTEP_CLASS_LABEL: def name(model_hparams, vocab_size): return "softmax_last_timestep_onehot_class_label_modality_%d_%d" % ( vocab_size, model_hparams.hidden_size) return name elif modality_type == ModalityType.SOFTMAX_MAX_POOLING_CLASS_LABEL: def name(model_hparams, vocab_size): return "softmax_max_pooling_onehot_class_label_modality_%d_%d" % ( vocab_size, model_hparams.hidden_size) return name return value
Gets default bottom transformation for targets; if none, return value. def get_targets_bottom(modality_type, value=None): """Gets default bottom transformation for targets; if none, return value.""" if modality_type == ModalityType.AUDIO: return make_targets_bottom(audio_bottom) elif modality_type == ModalityType.AUDIO_SPECTRAL: return make_targets_bottom(audio_spectral_bottom) elif modality_type in (ModalityType.CLASS_LABEL, ModalityType.MULTI_LABEL, ModalityType.ONE_HOT_CLASS_LABEL, ModalityType.SIGMOID_CLASS_LABEL, ModalityType.SIGMOID_MAX_POOLING_CLASS_LABEL, ModalityType.SOFTMAX_AVERAGE_POOLING_CLASS_LABEL, ModalityType.SOFTMAX_LAST_TIMESTEP_CLASS_LABEL, ModalityType.SOFTMAX_MAX_POOLING_CLASS_LABEL): return class_label_targets_bottom elif modality_type in (ModalityType.CTC_SYMBOL, ModalityType.SYMBOL, ModalityType.SYMBOL_WEIGHTS_ALL): return symbol_targets_bottom elif modality_type in (ModalityType.GENERIC_L2_LOSS, ModalityType.IDENTITY_SYMBOL): return identity_bottom elif modality_type == ModalityType.IDENTITY: return make_targets_bottom(identity_bottom) elif modality_type == ModalityType.IMAGE: return image_targets_bottom elif modality_type in (ModalityType.IMAGE_CHANNEL_BOTTOM_IDENTITY, ModalityType.IMAGE_CHANNEL_COMPRESS): return image_channel_compress_targets_bottom elif modality_type == ModalityType.IMAGE_CHANNEL_EMBEDDINGS_BOTTOM: return image_channel_embeddings_bottom elif modality_type in (ModalityType.REAL, ModalityType.REAL_L2_LOSS, ModalityType.REAL_LOG_POISSON_LOSS): return make_targets_bottom(real_bottom) elif modality_type == ModalityType.SPEECH_RECOGNITION: return make_targets_bottom(speech_recognition_bottom) elif modality_type == ModalityType.SYMBOL_ONE_HOT: return symbol_one_hot_bottom elif modality_type in (ModalityType.VIDEO, ModalityType.VIDEO_L1, ModalityType.VIDEO_L2): return video_targets_bottom elif modality_type == ModalityType.VIDEO_BITWISE: return video_bitwise_targets_bottom elif modality_type == ModalityType.VIDEO_IDENTITY: return video_identity_targets_bottom elif modality_type in (ModalityType.VIDEO_L1_RAW, ModalityType.VIDEO_L2_RAW): return video_raw_targets_bottom elif modality_type == ModalityType.VIDEO_PIXEL_NOISE: return make_targets_bottom(video_pixel_noise_bottom) return value
Gets default top transformation; if none available, return value. def get_top(modality_type, value=None): """Gets default top transformation; if none available, return value.""" if modality_type in (ModalityType.AUDIO, ModalityType.AUDIO_SPECTRAL, ModalityType.GENERIC_L2_LOSS, ModalityType.IDENTITY, ModalityType.IDENTITY_SYMBOL, ModalityType.IMAGE_CHANNEL_BOTTOM_IDENTITY, ModalityType.SPEECH_RECOGNITION, ModalityType.VIDEO_IDENTITY): return identity_top elif modality_type in (ModalityType.CLASS_LABEL, ModalityType.MULTI_LABEL, ModalityType.ONE_HOT_CLASS_LABEL, ModalityType.SIGMOID_CLASS_LABEL): return class_label_top elif modality_type in (ModalityType.CTC_SYMBOL, ModalityType.SYMBOL, ModalityType.SYMBOL_WEIGHTS_ALL): return symbol_top elif modality_type == ModalityType.IMAGE: return image_top elif modality_type == ModalityType.IMAGE_CHANNEL_COMPRESS: return image_channel_compress_top elif modality_type == ModalityType.IMAGE_CHANNEL_EMBEDDINGS_BOTTOM: return image_channel_embeddings_top elif modality_type in (ModalityType.REAL, ModalityType.REAL_L2_LOSS, ModalityType.REAL_LOG_POISSON_LOSS): return real_top elif modality_type == ModalityType.SIGMOID_MAX_POOLING_CLASS_LABEL: return sigmoid_max_pooling_class_label_top elif modality_type == ModalityType.SOFTMAX_AVERAGE_POOLING_CLASS_LABEL: return softmax_average_pooling_class_label_top elif modality_type == ModalityType.SOFTMAX_LAST_TIMESTEP_CLASS_LABEL: return softmax_last_timestep_class_label_top elif modality_type == ModalityType.SOFTMAX_MAX_POOLING_CLASS_LABEL: return softmax_max_pooling_class_label_top elif modality_type == ModalityType.SYMBOL_ONE_HOT: return symbol_one_hot_top elif modality_type in (ModalityType.VIDEO, ModalityType.VIDEO_BITWISE, ModalityType.VIDEO_PIXEL_NOISE): return video_top elif modality_type in (ModalityType.VIDEO_L1, ModalityType.VIDEO_L2): return video_l1_top elif modality_type in (ModalityType.VIDEO_L1_RAW, ModalityType.VIDEO_L2_RAW): return video_raw_top return value
Gets default weights function; if none available, return value. def get_weights_fn(modality_type, value=None): """Gets default weights function; if none available, return value.""" if modality_type in (ModalityType.CTC_SYMBOL, ModalityType.IDENTITY_SYMBOL, ModalityType.MULTI_LABEL, ModalityType.SYMBOL, ModalityType.SYMBOL_ONE_HOT): return common_layers.weights_nonzero elif modality_type in ModalityType.get_choices(): return common_layers.weights_all return value
Generates all possible pair combinations for the input list of sentences. For example: input = ["paraphrase1", "paraphrase2", "paraphrase3"] output = [("paraphrase1", "paraphrase2"), ("paraphrase1", "paraphrase3"), ("paraphrase2", "paraphrase3")] Args: list_of_sentences: the list of input sentences. Returns: the list of all possible sentence pairs. def create_combination(list_of_sentences): """Generates all possible pair combinations for the input list of sentences. For example: input = ["paraphrase1", "paraphrase2", "paraphrase3"] output = [("paraphrase1", "paraphrase2"), ("paraphrase1", "paraphrase3"), ("paraphrase2", "paraphrase3")] Args: list_of_sentences: the list of input sentences. Returns: the list of all possible sentence pairs. """ num_sentences = len(list_of_sentences) - 1 combinations = [] for i, _ in enumerate(list_of_sentences): if i == num_sentences: break num_pairs = num_sentences - i populated = num_pairs * [list_of_sentences[i]] zipped = list(zip(populated, list_of_sentences[i + 1:])) combinations += zipped return combinations
Set of hyperparameters. def image_transformer2d_base(): """Set of hyperparameters.""" hparams = common_hparams.basic_params1() hparams.hidden_size = 512 hparams.batch_size = 1 hparams.max_length = 256 hparams.dropout = 0.0 hparams.clip_grad_norm = 0. # i.e. no gradient clipping hparams.optimizer_adam_epsilon = 1e-9 hparams.learning_rate_decay_scheme = "noam" hparams.learning_rate = 0.1 hparams.learning_rate_warmup_steps = 4000 hparams.initializer_gain = 0.2 hparams.initializer = "uniform_unit_scaling" hparams.weight_decay = 0.0 hparams.optimizer_adam_beta1 = 0.9 hparams.optimizer_adam_beta2 = 0.98 hparams.label_smoothing = 0.0 hparams.bottom["targets"] = modalities.make_targets_bottom( modalities.image_channel_embeddings_bottom) hparams.top["targets"] = modalities.identity_top hparams.norm_type = "layer" hparams.layer_prepostprocess_dropout = 0.0 hparams.add_hparam("filter_size", 512) # Add new ones like this. # attention-related flags hparams.add_hparam("num_heads", 8) hparams.add_hparam("attention_key_channels", 0) hparams.add_hparam("attention_value_channels", 0) hparams.add_hparam("ffn_layer", "conv_hidden_relu") # All hyperparameters ending in "dropout" are automatically set to 0.0 # when not in training mode. hparams.add_hparam("attention_dropout", 0.0) hparams.add_hparam("relu_dropout", 0.0) hparams.add_hparam("pos", "timing") # timing, none hparams.add_hparam("nbr_decoder_problems", 1) hparams.add_hparam("num_output_layers", 3) hparams.add_hparam("block_size", 1) # image size related flags # assuming that the image has same height and width hparams.add_hparam("img_len", 32) hparams.add_hparam("num_channels", 3) # Local attention params hparams.add_hparam("local_and_global_att", False) hparams.add_hparam("block_length", 256) hparams.add_hparam("block_width", 128) # Local 2D attention params hparams.add_hparam("query_shape", (16, 16)) hparams.add_hparam("memory_flange", (16, 32)) hparams.add_hparam("num_encoder_layers", 4) hparams.add_hparam("num_decoder_layers", 8) # attention type related params hparams.add_hparam("enc_attention_type", cia.AttentionType.GLOBAL) hparams.add_hparam("dec_attention_type", cia.AttentionType.LOCAL_2D) hparams.add_hparam("block_raster_scan", False) # multipos attention params hparams.add_hparam("q_filter_width", 1) hparams.add_hparam("kv_filter_width", 1) hparams.add_hparam("unconditional", False) # unconditional generation # relative embedding hparams hparams.add_hparam("shared_rel", False) return hparams
hparams fo 8 layer big 2d model for cifar 10. def imagetransformer2d_base_8l_8_32_big(): """hparams fo 8 layer big 2d model for cifar 10.""" hparams = image_transformer2d_base() hparams.num_heads = 16 hparams.hidden_size = 1024 hparams.filter_size = 2048 hparams.num_decoder_layers = 8 hparams.batch_size = 1 hparams.layer_prepostprocess_dropout = 0.3 hparams.query_shape = (8, 16) hparams.memory_flange = (0, 32) hparams.unconditional = int(False) return hparams
big 1d model for unconditional generation on imagenet. def imagetransformer_base_10l_8h_big_uncond_dr03_dan_64_2d(): """big 1d model for unconditional generation on imagenet.""" hparams = image_transformer2d_base() hparams.unconditional = True hparams.hidden_size = 512 hparams.batch_size = 1 hparams.img_len = 64 hparams.num_heads = 8 hparams.filter_size = 2048 hparams.batch_size = 1 hparams.max_length = 3075 hparams.max_length = 14000 hparams.layer_preprocess_sequence = "none" hparams.layer_postprocess_sequence = "dan" hparams.layer_prepostprocess_dropout = 0.1 hparams.dec_attention_type = cia.AttentionType.LOCAL_2D hparams.query_shape = (16, 16) hparams.memory_flange = (8, 8) return hparams
Base params for img2img 2d attention. def img2img_transformer2d_base(): """Base params for img2img 2d attention.""" hparams = image_transformer2d_base() # learning related flags hparams.layer_preprocess_sequence = "n" hparams.layer_postprocess_sequence = "da" # This version seems to benefit from a higher learning rate. hparams.learning_rate = 0.2 hparams.layer_prepostprocess_dropout = 0.1 hparams.learning_rate_warmup_steps = 12000 hparams.filter_size = 2048 hparams.num_encoder_layers = 4 hparams.num_decoder_layers = 8 hparams.bottom["inputs"] = modalities.image_channel_embeddings_bottom hparams.dec_attention_type = cia.AttentionType.LOCAL_2D hparams.block_raster_scan = True return hparams
Current best hparams for local 2d. def img2img_transformer2d_q3(): """Current best hparams for local 2d.""" hparams = img2img_transformer2d_q1() hparams.batch_size = 2 hparams.query_shape = (8, 16) hparams.memory_flange = (8, 32) return hparams
Base params for local1d attention. def img2img_transformer_base(): """Base params for local1d attention.""" hparams = image_transformer2d_base() # learning related flags hparams.layer_preprocess_sequence = "n" hparams.layer_postprocess_sequence = "da" # This version seems to benefit from a higher learning rate. hparams.learning_rate = 0.2 hparams.layer_prepostprocess_dropout = 0.1 hparams.learning_rate_warmup_steps = 12000 hparams.filter_size = 2048 hparams.num_encoder_layers = 4 hparams.num_decoder_layers = 8 hparams.block_length = 256 hparams.block_width = 256 hparams.dec_attention_type = cia.AttentionType.LOCAL_1D hparams.block_raster_scan = False return hparams
Current best hparams for local 1d. def img2img_transformer_b3(): """Current best hparams for local 1d.""" hparams = img2img_transformer_base() hparams.batch_size = 2 hparams.layer_preprocess_sequence = "none" hparams.layer_postprocess_sequence = "dan" hparams.block_length = 128 hparams.sampling_temp = 0.9 return hparams
Try dilated. def img2img_transformer_dilated(): """Try dilated.""" hparams = img2img_transformer_base() hparams.add_hparam("num_memory_blocks", 1) hparams.num_heads = 8 hparams.attention_key_channels = hparams.attention_value_channels = 0 hparams.hidden_size = 512 hparams.filter_size = 2048 hparams.num_decoder_layers = 8 hparams.sampling_method = "random" hparams.gap_sizes = [0, 16, 64, 0, 16, 64, 128, 0] hparams.dec_attention_type = cia.AttentionType.DILATED hparams.img_len = 64 hparams.block_length = 128 hparams.block_width = 128 return hparams
Hparams for training img2img_transformer on tpu. def img2img_transformer_base_tpu(): """Hparams for training img2img_transformer on tpu.""" hparams = img2img_transformer_base() update_hparams_for_tpu(hparams) hparams.batch_size = 2 hparams.num_heads = 4 # heads are expensive on tpu hparams.num_decoder_layers = 8 hparams.num_encoder_layers = 4 hparams.shared_embedding_and_softmax_weights = False return hparams
Set of hyperparameters. def img2img_transformer2d_n31(): """Set of hyperparameters.""" hparams = img2img_transformer2d_base() hparams.batch_size = 1 hparams.num_encoder_layers = 6 hparams.num_decoder_layers = 12 hparams.num_heads = 8 hparams.query_shape = (16, 32) hparams.memory_flange = (16, 32) return hparams
Set of hyperparameters. def img2img_transformer2d_n24(): """Set of hyperparameters.""" hparams = img2img_transformer2d_base() hparams.batch_size = 1 hparams.hidden_size = 1024 hparams.filter_size = 2048 hparams.layer_prepostprocess_dropout = 0.2 hparams.num_decoder_layers = 8 hparams.query_shape = (8, 16) hparams.memory_flange = (8, 32) return hparams
Tiny params. def img2img_transformer2d_tiny(): """Tiny params.""" hparams = img2img_transformer2d_base() hparams.num_decoder_layers = 2 hparams.hidden_size = 128 hparams.batch_size = 4 hparams.max_length = 128 hparams.attention_key_channels = hparams.attention_value_channels = 0 hparams.filter_size = 128 hparams.num_heads = 4 hparams.pos = "timing" hparams.img_len = 32 return hparams
Tiny params. def img2img_transformer_tiny(): """Tiny params.""" hparams = img2img_transformer2d_base() hparams.num_hidden_layers = 2 hparams.hidden_size = 128 hparams.batch_size = 4 hparams.max_length = 128 hparams.attention_key_channels = hparams.attention_value_channels = 0 hparams.filter_size = 128 hparams.num_heads = 1 hparams.pos = "timing" return hparams