Python tensorflow.app.run() Examples
The following are 30
code examples of tensorflow.app.run().
You can vote up the ones you like or vote down the ones you don't like,
and go to the original project or source file by following the links above each example.
You may also want to check out all available functions/classes of the module
tensorflow.app
, or try the search function
.
Example #1
Source File: model.py From ad-versarial with MIT License | 5 votes |
def match(self, img, verbose=False, img_name=None): img = to_alpha(img) h, w, _ = img.shape if h < 5 or w < 5: if verbose: print("Skipping {}, too small".format(img_name)) return False, len(self.target_txt) feed_dict = { self.img_var: img, self.h_orig_var: [h], self.w_orig_var: [w], self.h_resized_var: [h], self.w_resized_var: [w] } mul = 1 if self.target_height and h < self.target_height: mul = get_size_mul(h, w, self.target_height) feed_dict[self.h_resized_var] = [mul * h] feed_dict[self.w_resized_var] = [mul * w] output2 = self.sess.run(self.text_output, feed_dict=feed_dict) s2 = decode(output2, self.char_map)[0] print(s2) dist2 = levenshtein(s2.lower(), self.target_txt.lower()) if self.target_txt.lower() in s2.lower(): dist2 = 0 dist = dist2 if dist <= self.sim_threshold: if verbose: print("{} matched with distance {}".format(img_name, dist)) return True, dist else: if verbose: print("no match for {}! Distance {}".format(img_name, dist)) return False, dist
Example #2
Source File: train.py From Y8M with Apache License 2.0 | 5 votes |
def run(self): """Starts the parameter server.""" logging.info("%s: Starting parameter server within cluster %s.", task_as_string(self.task), self.cluster.as_dict()) server = start_server(self.cluster, self.task) server.join()
Example #3
Source File: train.py From object_detection_with_tensorflow with MIT License | 5 votes |
def train(loss, init_fn, hparams): """Wraps slim.learning.train to run a training loop. Args: loss: a loss tensor init_fn: A callable to be executed after all other initialization is done. hparams: a model hyper parameters """ optimizer = create_optimizer(hparams) if FLAGS.sync_replicas: replica_id = tf.constant(FLAGS.task, tf.int32, shape=()) optimizer = tf.LegacySyncReplicasOptimizer( opt=optimizer, replicas_to_aggregate=FLAGS.replicas_to_aggregate, replica_id=replica_id, total_num_replicas=FLAGS.total_num_replicas) sync_optimizer = optimizer startup_delay_steps = 0 else: startup_delay_steps = 0 sync_optimizer = None train_op = slim.learning.create_train_op( loss, optimizer, summarize_gradients=True, clip_gradient_norm=FLAGS.clip_gradient_norm) slim.learning.train( train_op=train_op, logdir=FLAGS.train_log_dir, graph=loss.graph, master=FLAGS.master, is_chief=(FLAGS.task == 0), number_of_steps=FLAGS.max_number_of_steps, save_summaries_secs=FLAGS.save_summaries_secs, save_interval_secs=FLAGS.save_interval_secs, startup_delay_steps=startup_delay_steps, sync_optimizer=sync_optimizer, init_fn=init_fn)
Example #4
Source File: inference.py From AttentionCluster with Apache License 2.0 | 5 votes |
def main(unused_argv): logging.set_verbosity(tf.logging.INFO) if FLAGS.input_model_tgz: if FLAGS.train_dir: raise ValueError("You cannot supply --train_dir if supplying " "--input_model_tgz") # Untar. if not file_io.file_exists(FLAGS.untar_model_dir): os.makedirs(FLAGS.untar_model_dir) tarfile.open(FLAGS.input_model_tgz).extractall(FLAGS.untar_model_dir) FLAGS.train_dir = FLAGS.untar_model_dir flags_dict_file = os.path.join(FLAGS.train_dir, "model_flags.json") if not file_io.file_exists(flags_dict_file): raise IOError("Cannot find %s. Did you run eval.py?" % flags_dict_file) flags_dict = json.loads(file_io.FileIO(flags_dict_file, "r").read()) # convert feature_names and feature_sizes to lists of values feature_names, feature_sizes = utils.GetListOfFeatureNamesAndSizes( flags_dict["feature_names"], flags_dict["feature_sizes"]) if flags_dict["frame_features"]: reader = readers.YT8MFrameFeatureReader(feature_names=feature_names, feature_sizes=feature_sizes) else: reader = readers.YT8MAggregatedFeatureReader(feature_names=feature_names, feature_sizes=feature_sizes) if FLAGS.output_file is "": raise ValueError("'output_file' was not specified. " "Unable to continue with inference.") if FLAGS.input_data_pattern is "": raise ValueError("'input_data_pattern' was not specified. " "Unable to continue with inference.") inference(reader, FLAGS.train_dir, FLAGS.input_data_pattern, FLAGS.output_file, FLAGS.batch_size, FLAGS.top_k)
Example #5
Source File: train.py From object_detection_kitti with Apache License 2.0 | 5 votes |
def train(loss, init_fn, hparams): """Wraps slim.learning.train to run a training loop. Args: loss: a loss tensor init_fn: A callable to be executed after all other initialization is done. hparams: a model hyper parameters """ optimizer = create_optimizer(hparams) if FLAGS.sync_replicas: replica_id = tf.constant(FLAGS.task, tf.int32, shape=()) optimizer = tf.LegacySyncReplicasOptimizer( opt=optimizer, replicas_to_aggregate=FLAGS.replicas_to_aggregate, replica_id=replica_id, total_num_replicas=FLAGS.total_num_replicas) sync_optimizer = optimizer startup_delay_steps = 0 else: startup_delay_steps = 0 sync_optimizer = None train_op = slim.learning.create_train_op( loss, optimizer, summarize_gradients=True, clip_gradient_norm=FLAGS.clip_gradient_norm) slim.learning.train( train_op=train_op, logdir=FLAGS.train_log_dir, graph=loss.graph, master=FLAGS.master, is_chief=(FLAGS.task == 0), number_of_steps=FLAGS.max_number_of_steps, save_summaries_secs=FLAGS.save_summaries_secs, save_interval_secs=FLAGS.save_interval_secs, startup_delay_steps=startup_delay_steps, sync_optimizer=sync_optimizer, init_fn=init_fn)
Example #6
Source File: train.py From hands-detection with MIT License | 5 votes |
def train(loss, init_fn, hparams): """Wraps slim.learning.train to run a training loop. Args: loss: a loss tensor init_fn: A callable to be executed after all other initialization is done. hparams: a model hyper parameters """ optimizer = create_optimizer(hparams) if FLAGS.sync_replicas: replica_id = tf.constant(FLAGS.task, tf.int32, shape=()) optimizer = tf.LegacySyncReplicasOptimizer( opt=optimizer, replicas_to_aggregate=FLAGS.replicas_to_aggregate, replica_id=replica_id, total_num_replicas=FLAGS.total_num_replicas) sync_optimizer = optimizer startup_delay_steps = 0 else: startup_delay_steps = 0 sync_optimizer = None train_op = slim.learning.create_train_op( loss, optimizer, summarize_gradients=True, clip_gradient_norm=FLAGS.clip_gradient_norm) slim.learning.train( train_op=train_op, logdir=FLAGS.train_log_dir, graph=loss.graph, master=FLAGS.master, is_chief=(FLAGS.task == 0), number_of_steps=FLAGS.max_number_of_steps, save_summaries_secs=FLAGS.save_summaries_secs, save_interval_secs=FLAGS.save_interval_secs, startup_delay_steps=startup_delay_steps, sync_optimizer=sync_optimizer, init_fn=init_fn)
Example #7
Source File: train.py From AttentionCluster with Apache License 2.0 | 5 votes |
def run(self): """Starts the parameter server.""" logging.info("%s: Starting parameter server within cluster %s.", task_as_string(self.task), self.cluster.as_dict()) server = start_server(self.cluster, self.task) server.join()
Example #8
Source File: train.py From Y8M with Apache License 2.0 | 5 votes |
def get_input_data_tensors(reader, data_files, batch_size=1000, num_epochs=None, num_readers=1): """Creates the section of the graph which reads the training data. Args: reader: A class which parses the training data. data_pattern: A 'glob' style path to the data files. batch_size: How many examples to process at a time. num_epochs: How many passes to make over the training data. Set to 'None' to run indefinitely. num_readers: How many I/O threads to use. Returns: A tuple containing the features tensor, labels tensor, and optionally a tensor containing the number of frames per video. The exact dimensions depend on the reader being used. Raises: IOError: If no files matching the given pattern were found. """ logging.info("Using batch size of " + str(batch_size) + " for training.") with tf.name_scope("train_input"): logging.info("Fold number {}, use files {}.".format( FLAGS.fold, len(data_files))) filename_queue = tf.train.string_input_producer( data_files, num_epochs=num_epochs, shuffle=True) training_data = [ reader.prepare_reader(filename_queue) for _ in range(num_readers) ] return tf.train.shuffle_batch_join( training_data, batch_size=batch_size, capacity=batch_size * 5, min_after_dequeue=batch_size, allow_smaller_final_batch=True, enqueue_many=True)
Example #9
Source File: train.py From Y8M with Apache License 2.0 | 5 votes |
def run(self): """Starts the parameter server.""" logging.info("%s: Starting parameter server within cluster %s.", task_as_string(self.task), self.cluster.as_dict()) server = start_server(self.cluster, self.task) server.join()
Example #10
Source File: train.py From Y8M with Apache License 2.0 | 5 votes |
def run(self): """Starts the parameter server.""" logging.info("%s: Starting parameter server within cluster %s.", task_as_string(self.task), self.cluster.as_dict()) server = start_server(self.cluster, self.task) server.join()
Example #11
Source File: train.py From Y8M with Apache License 2.0 | 5 votes |
def main(unused_argv): # Load the environment. env = json.loads(os.environ.get("TF_CONFIG", "{}")) # Load the cluster data from the environment. cluster_data = env.get("cluster", None) cluster = tf.train.ClusterSpec(cluster_data) if cluster_data else None # Load the task data from the environment. task_data = env.get("task", None) or {"type": "master", "index": 0} task = type("TaskSpec", (object,), task_data) # Logging the version. logging.set_verbosity(tf.logging.INFO) logging.info("%s: Tensorflow version: %s.", task_as_string(task), tf.__version__) # Dispatch to a master, a worker, or a parameter server. if not cluster or task.type == "master" or task.type == "worker": model = find_class_by_name(FLAGS.model, [frame_level_models, video_level_models])() reader = get_reader() model_exporter = export_model.ModelExporter( frame_features=FLAGS.frame_features, model=model, reader=reader) Trainer(cluster, task, FLAGS.train_dir, model, reader, model_exporter, FLAGS.log_device_placement, FLAGS.max_steps, FLAGS.export_model_steps).run(start_new_model=FLAGS.start_new_model) elif task.type == "ps": ParameterServer(cluster, task).run() else: raise ValueError("%s: Invalid task_type: %s." % (task_as_string(task), task.type))
Example #12
Source File: train.py From g-tensorflow-models with Apache License 2.0 | 5 votes |
def train(loss, init_fn, hparams): """Wraps slim.learning.train to run a training loop. Args: loss: a loss tensor init_fn: A callable to be executed after all other initialization is done. hparams: a model hyper parameters """ optimizer = create_optimizer(hparams) if FLAGS.sync_replicas: replica_id = tf.constant(FLAGS.task, tf.int32, shape=()) optimizer = tf.LegacySyncReplicasOptimizer( opt=optimizer, replicas_to_aggregate=FLAGS.replicas_to_aggregate, replica_id=replica_id, total_num_replicas=FLAGS.total_num_replicas) sync_optimizer = optimizer startup_delay_steps = 0 else: startup_delay_steps = 0 sync_optimizer = None train_op = slim.learning.create_train_op( loss, optimizer, summarize_gradients=True, clip_gradient_norm=FLAGS.clip_gradient_norm) slim.learning.train( train_op=train_op, logdir=FLAGS.train_log_dir, graph=loss.graph, master=FLAGS.master, is_chief=(FLAGS.task == 0), number_of_steps=FLAGS.max_number_of_steps, save_summaries_secs=FLAGS.save_summaries_secs, save_interval_secs=FLAGS.save_interval_secs, startup_delay_steps=startup_delay_steps, sync_optimizer=sync_optimizer, init_fn=init_fn)
Example #13
Source File: train.py From models with Apache License 2.0 | 5 votes |
def train(loss, init_fn, hparams): """Wraps slim.learning.train to run a training loop. Args: loss: a loss tensor init_fn: A callable to be executed after all other initialization is done. hparams: a model hyper parameters """ optimizer = create_optimizer(hparams) if FLAGS.sync_replicas: replica_id = tf.constant(FLAGS.task, tf.int32, shape=()) optimizer = tf.LegacySyncReplicasOptimizer( opt=optimizer, replicas_to_aggregate=FLAGS.replicas_to_aggregate, replica_id=replica_id, total_num_replicas=FLAGS.total_num_replicas) sync_optimizer = optimizer startup_delay_steps = 0 else: startup_delay_steps = 0 sync_optimizer = None train_op = slim.learning.create_train_op( loss, optimizer, summarize_gradients=True, clip_gradient_norm=FLAGS.clip_gradient_norm) slim.learning.train( train_op=train_op, logdir=FLAGS.train_log_dir, graph=loss.graph, master=FLAGS.master, is_chief=(FLAGS.task == 0), number_of_steps=FLAGS.max_number_of_steps, save_summaries_secs=FLAGS.save_summaries_secs, save_interval_secs=FLAGS.save_interval_secs, startup_delay_steps=startup_delay_steps, sync_optimizer=sync_optimizer, init_fn=init_fn)
Example #14
Source File: train.py From multilabel-image-classification-tensorflow with MIT License | 5 votes |
def train(loss, init_fn, hparams): """Wraps slim.learning.train to run a training loop. Args: loss: a loss tensor init_fn: A callable to be executed after all other initialization is done. hparams: a model hyper parameters """ optimizer = create_optimizer(hparams) if FLAGS.sync_replicas: replica_id = tf.constant(FLAGS.task, tf.int32, shape=()) optimizer = tf.LegacySyncReplicasOptimizer( opt=optimizer, replicas_to_aggregate=FLAGS.replicas_to_aggregate, replica_id=replica_id, total_num_replicas=FLAGS.total_num_replicas) sync_optimizer = optimizer startup_delay_steps = 0 else: startup_delay_steps = 0 sync_optimizer = None train_op = slim.learning.create_train_op( loss, optimizer, summarize_gradients=True, clip_gradient_norm=FLAGS.clip_gradient_norm) slim.learning.train( train_op=train_op, logdir=FLAGS.train_log_dir, graph=loss.graph, master=FLAGS.master, is_chief=(FLAGS.task == 0), number_of_steps=FLAGS.max_number_of_steps, save_summaries_secs=FLAGS.save_summaries_secs, save_interval_secs=FLAGS.save_interval_secs, startup_delay_steps=startup_delay_steps, sync_optimizer=sync_optimizer, init_fn=init_fn)
Example #15
Source File: split_video.py From Y8M with Apache License 2.0 | 5 votes |
def frame_example_2_np(seq_example_bytes, max_quantized_value=2, min_quantized_value=-2): feature_names=['rgb','audio'] feature_sizes = [1024, 128] with tf.Graph().as_default(): contexts, features = tf.parse_single_sequence_example( seq_example_bytes, context_features={"video_id": tf.FixedLenFeature( [], tf.string), "labels": tf.VarLenFeature(tf.int64)}, sequence_features={ feature_name : tf.FixedLenSequenceFeature([], dtype=tf.string) for feature_name in feature_names }) decoded_features = { name: tf.reshape( tf.cast(tf.decode_raw(features[name], tf.uint8), tf.float32), [-1, size]) for name, size in zip(feature_names, feature_sizes) } feature_matrices = { name: utils.Dequantize(decoded_features[name], max_quantized_value, min_quantized_value) for name in feature_names} with tf.Session() as sess: vid = sess.run(contexts['video_id']) labs = sess.run(contexts['labels'].values) rgb = sess.run(feature_matrices['rgb']) audio = sess.run(feature_matrices['audio']) return vid, labs, rgb, audio #%% Split frame level file into three video level files: all, 1st half, 2nd half.
Example #16
Source File: deep_edge_trainer.py From asymproj_edge_dnn with Apache License 2.0 | 5 votes |
def estimate_AUC(sess, nn, embeddings, positives, negatives): """Measures the AUC-ROC given positive and negative edges. Args: sess: TensorFlow session that contains parameters of edge neural net `nn`. nn: Edge Neural Network. It is expected that it contains members: embeddings_a: (B, D) float tensor for inputting embedding vectors, where `B` is batch size and `D` is the input node embedding dimensionality. embeddings_b: (B, D) float tensor for inputting embedding vectors. Every row corresponds to `embeddings_a` batch_size: int tensor, which will be fed `B`. output: Output float tensor with shape (B). Every entry `i` contains the edge score between `embeddings_a[i]` and `embeddings_b[i]`. embeddings: (|V|, D) float32 numpy.array containing all node embeddings. positives: (P, 2) int32 numpy.array containing positive edges, where `positives[j]` contains (node ID 1, node ID 2) with both IDs in range [0, |V|-1]. negatives: (N, 2) int32 numpy.array containing negative edges. Returns: roc-auc score for ranking `positives` above `negatives`. """ all_pairs = numpy.concatenate([positives, negatives], 0) all_scores = sess.run( nn.output, feed_dict={ nn.embeddings_a: embeddings[all_pairs[:, 0]], nn.embeddings_b: embeddings[all_pairs[:, 1]], nn.batch_size: len(all_pairs), }) pos_scores = all_scores[:len(positives)] neg_scores = all_scores[-len(negatives):] return metrics.roc_auc_score( [1] * len(pos_scores) + [0] * len(neg_scores), numpy.concatenate([pos_scores, neg_scores], 0))
Example #17
Source File: model.py From ad-versarial with MIT License | 5 votes |
def __init__(self, target, target_height=0, target_threshold=4): self.target_txt = target self.target_height = target_height self.sim_threshold = target_threshold use_gpu = FLAGS.use_gpu >= 0 self.char_map = read_all_chars() params = read_tesseract_params(use_gpu=use_gpu) model = MyVGSLImageModel(use_gpu=use_gpu) self.img_var = tf.placeholder(dtype=tf.float32, shape=(None, None, 4)) self.h_orig_var = tf.placeholder(dtype=tf.int64, shape=[1]) self.w_orig_var = tf.placeholder(dtype=tf.int64, shape=[1]) self.h_resized_var = tf.placeholder(dtype=tf.int64, shape=[1]) self.w_resized_var = tf.placeholder(dtype=tf.int64, shape=[1]) self.resized_dims_var = tf.cast( tf.concat([self.h_resized_var, self.w_resized_var], axis=0), tf.int32) img_preproc = self.img_var img_preproc = remove_alpha(img_preproc) img_preproc = preprocess_tf(img_preproc, self.h_orig_var[0], self.w_orig_var[0]) img_large = tf.image.resize_images(img_preproc, self.resized_dims_var, method=tf.image.ResizeMethod.BILINEAR) img_large = tf.image.rgb_to_grayscale(img_large) logits, _ = model(img_large, self.h_resized_var, self.w_resized_var) self.text_output = ctc_decode(logits, model.ctc_width) init_ops = init(params, use_gpu=use_gpu, skip=0) self.sess = tf.Session() self.sess.run(init_ops)
Example #18
Source File: evaluate.py From stereo-magnification with Apache License 2.0 | 5 votes |
def evaluate_one(result_root, model_name, data_split, example): """Compare one example on one model, returning ssim and PSNR scores.""" example_dir = os.path.join(result_root, model_name, data_split, example) tgt_file = tf.gfile.Glob(example_dir + '/tgt_image_*')[0] tgt_image = tf.convert_to_tensor(load_image(tgt_file), dtype=tf.float32) pred_file = tf.gfile.Glob(example_dir + '/output_image_*')[0] pred_image = tf.convert_to_tensor(load_image(pred_file), dtype=tf.float32) ssim = tf.image.ssim(pred_image, tgt_image, max_val=255.0) psnr = tf.image.psnr(pred_image, tgt_image, max_val=255.0) with tf.Session() as sess: return sess.run(ssim).item(), sess.run(psnr).item()
Example #19
Source File: train.py From youtube8mchallenge with Apache License 2.0 | 5 votes |
def run(self): """Starts the parameter server.""" logging.info("%s: Starting parameter server within cluster %s.", task_as_string(self.task), self.cluster.as_dict()) server = start_server(self.cluster, self.task) server.join()
Example #20
Source File: train_distill.py From youtube8mchallenge with Apache License 2.0 | 5 votes |
def run(self): """Starts the parameter server.""" logging.info("%s: Starting parameter server within cluster %s.", task_as_string(self.task), self.cluster.as_dict()) server = start_server(self.cluster, self.task) server.join()
Example #21
Source File: inference.py From youtube8mchallenge with Apache License 2.0 | 5 votes |
def main(unused_argv): logging.set_verbosity(tf.logging.INFO) if FLAGS.input_model_tgz: if FLAGS.train_dir: raise ValueError("You cannot supply --train_dir if supplying " "--input_model_tgz") # Untar. if not os.path.exists(FLAGS.untar_model_dir): os.makedirs(FLAGS.untar_model_dir) tarfile.open(FLAGS.input_model_tgz).extractall(FLAGS.untar_model_dir) FLAGS.train_dir = FLAGS.untar_model_dir flags_dict_file = os.path.join(FLAGS.train_dir, "model_flags.json") if not os.path.exists(flags_dict_file): raise IOError("Cannot find %s. Did you run eval.py?" % flags_dict_file) flags_dict = json.loads(open(flags_dict_file).read()) # convert feature_names and feature_sizes to lists of values feature_names, feature_sizes = utils.GetListOfFeatureNamesAndSizes( flags_dict["feature_names"], flags_dict["feature_sizes"]) if flags_dict["frame_features"]: reader = readers.YT8MFrameFeatureReader(feature_names=feature_names, feature_sizes=feature_sizes) else: reader = readers.YT8MAggregatedFeatureReader(feature_names=feature_names, feature_sizes=feature_sizes) if FLAGS.output_file is "": raise ValueError("'output_file' was not specified. " "Unable to continue with inference.") if FLAGS.input_data_pattern is "": raise ValueError("'input_data_pattern' was not specified. " "Unable to continue with inference.") inference(reader, FLAGS.train_dir, FLAGS.input_data_pattern, FLAGS.output_file, FLAGS.batch_size, FLAGS.top_k)
Example #22
Source File: inference_gpu.py From youtube8mchallenge with Apache License 2.0 | 5 votes |
def main(unused_argv): logging.set_verbosity(tf.logging.INFO) if FLAGS.input_model_tgz: if FLAGS.train_dir: raise ValueError("You cannot supply --train_dir if supplying " "--input_model_tgz") # Untar. if not os.path.exists(FLAGS.untar_model_dir): os.makedirs(FLAGS.untar_model_dir) tarfile.open(FLAGS.input_model_tgz).extractall(FLAGS.untar_model_dir) FLAGS.train_dir = FLAGS.untar_model_dir flags_dict_file = os.path.join(FLAGS.train_dir, "model_flags.json") if not os.path.exists(flags_dict_file): raise IOError("Cannot find %s. Did you run eval.py?" % flags_dict_file) flags_dict = json.loads(open(flags_dict_file).read()) # convert feature_names and feature_sizes to lists of values feature_names, feature_sizes = utils.GetListOfFeatureNamesAndSizes( flags_dict["feature_names"], flags_dict["feature_sizes"]) if flags_dict["frame_features"]: reader = readers.YT8MFrameFeatureReader(feature_names=feature_names, feature_sizes=feature_sizes) else: reader = readers.YT8MAggregatedFeatureReader(feature_names=feature_names, feature_sizes=feature_sizes) if FLAGS.output_file is "": raise ValueError("'output_file' was not specified. " "Unable to continue with inference.") if FLAGS.input_data_pattern is "": raise ValueError("'input_data_pattern' was not specified. " "Unable to continue with inference.") inference(reader, FLAGS.train_dir, FLAGS.input_data_pattern, FLAGS.output_file, FLAGS.batch_size, FLAGS.top_k)
Example #23
Source File: train.py From youtube-8m with Apache License 2.0 | 5 votes |
def main(unused_argv): # Load the environment. env = json.loads(os.environ.get("TF_CONFIG", "{}")) # Load the cluster data from the environment. cluster_data = env.get("cluster", None) cluster = tf.train.ClusterSpec(cluster_data) if cluster_data else None # Load the task data from the environment. task_data = env.get("task", None) or {"type": "master", "index": 0} task = type("TaskSpec", (object,), task_data) # Logging the version. logging.set_verbosity(tf.logging.INFO) logging.info("%s: Tensorflow version: %s.", task_as_string(task), tf.__version__) # Dispatch to a master, a worker, or a parameter server. if not cluster or task.type == "master" or task.type == "worker": model = find_class_by_name(FLAGS.model, [frame_level_models, video_level_models])() reader = get_reader() model_exporter = export_model.ModelExporter( frame_features=FLAGS.frame_features, model=model, reader=reader) Trainer(cluster, task, FLAGS.train_dir, model, reader, model_exporter, FLAGS.log_device_placement, FLAGS.max_steps, FLAGS.export_model_steps).run(start_new_model=FLAGS.start_new_model) elif task.type == "ps": ParameterServer(cluster, task).run() else: raise ValueError("%s: Invalid task_type: %s." % (task_as_string(task), task.type))
Example #24
Source File: train.py From youtube-8m with Apache License 2.0 | 5 votes |
def run(self): """Starts the parameter server.""" logging.info("%s: Starting parameter server within cluster %s.", task_as_string(self.task), self.cluster.as_dict()) server = start_server(self.cluster, self.task) server.join()
Example #25
Source File: inference.py From youtube-8m with Apache License 2.0 | 5 votes |
def main(unused_argv): logging.set_verbosity(tf.logging.INFO) if FLAGS.input_model_tgz: if FLAGS.train_dir: raise ValueError("You cannot supply --train_dir if supplying " "--input_model_tgz") # Untar. if not os.path.exists(FLAGS.untar_model_dir): os.makedirs(FLAGS.untar_model_dir) tarfile.open(FLAGS.input_model_tgz).extractall(FLAGS.untar_model_dir) FLAGS.train_dir = FLAGS.untar_model_dir flags_dict_file = os.path.join(FLAGS.train_dir, "model_flags.json") if not file_io.file_exists(flags_dict_file): raise IOError("Cannot find %s. Did you run eval.py?" % flags_dict_file) flags_dict = json.loads(file_io.FileIO(flags_dict_file, "r").read()) # convert feature_names and feature_sizes to lists of values feature_names, feature_sizes = utils.GetListOfFeatureNamesAndSizes( flags_dict["feature_names"], flags_dict["feature_sizes"]) if flags_dict["frame_features"]: reader = readers.YT8MFrameFeatureReader(feature_names=feature_names, feature_sizes=feature_sizes) else: reader = readers.YT8MAggregatedFeatureReader(feature_names=feature_names, feature_sizes=feature_sizes) if not FLAGS.output_file: raise ValueError("'output_file' was not specified. " "Unable to continue with inference.") if not FLAGS.input_data_pattern: raise ValueError("'input_data_pattern' was not specified. " "Unable to continue with inference.") inference(reader, FLAGS.train_dir, FLAGS.input_data_pattern, FLAGS.output_file, FLAGS.batch_size, FLAGS.top_k)
Example #26
Source File: train.py From Gun-Detector with Apache License 2.0 | 5 votes |
def train(loss, init_fn, hparams): """Wraps slim.learning.train to run a training loop. Args: loss: a loss tensor init_fn: A callable to be executed after all other initialization is done. hparams: a model hyper parameters """ optimizer = create_optimizer(hparams) if FLAGS.sync_replicas: replica_id = tf.constant(FLAGS.task, tf.int32, shape=()) optimizer = tf.LegacySyncReplicasOptimizer( opt=optimizer, replicas_to_aggregate=FLAGS.replicas_to_aggregate, replica_id=replica_id, total_num_replicas=FLAGS.total_num_replicas) sync_optimizer = optimizer startup_delay_steps = 0 else: startup_delay_steps = 0 sync_optimizer = None train_op = slim.learning.create_train_op( loss, optimizer, summarize_gradients=True, clip_gradient_norm=FLAGS.clip_gradient_norm) slim.learning.train( train_op=train_op, logdir=FLAGS.train_log_dir, graph=loss.graph, master=FLAGS.master, is_chief=(FLAGS.task == 0), number_of_steps=FLAGS.max_number_of_steps, save_summaries_secs=FLAGS.save_summaries_secs, save_interval_secs=FLAGS.save_interval_secs, startup_delay_steps=startup_delay_steps, sync_optimizer=sync_optimizer, init_fn=init_fn)
Example #27
Source File: train.py From DOTA_models with Apache License 2.0 | 5 votes |
def train(loss, init_fn, hparams): """Wraps slim.learning.train to run a training loop. Args: loss: a loss tensor init_fn: A callable to be executed after all other initialization is done. hparams: a model hyper parameters """ optimizer = create_optimizer(hparams) if FLAGS.sync_replicas: replica_id = tf.constant(FLAGS.task, tf.int32, shape=()) optimizer = tf.LegacySyncReplicasOptimizer( opt=optimizer, replicas_to_aggregate=FLAGS.replicas_to_aggregate, replica_id=replica_id, total_num_replicas=FLAGS.total_num_replicas) sync_optimizer = optimizer startup_delay_steps = 0 else: startup_delay_steps = 0 sync_optimizer = None train_op = slim.learning.create_train_op( loss, optimizer, summarize_gradients=True, clip_gradient_norm=FLAGS.clip_gradient_norm) slim.learning.train( train_op=train_op, logdir=FLAGS.train_log_dir, graph=loss.graph, master=FLAGS.master, is_chief=(FLAGS.task == 0), number_of_steps=FLAGS.max_number_of_steps, save_summaries_secs=FLAGS.save_summaries_secs, save_interval_secs=FLAGS.save_interval_secs, startup_delay_steps=startup_delay_steps, sync_optimizer=sync_optimizer, init_fn=init_fn)
Example #28
Source File: train.py From youtube-8m with Apache License 2.0 | 5 votes |
def main(unused_argv): # Load the environment. env = json.loads(os.environ.get("TF_CONFIG", "{}")) # Load the cluster data from the environment. cluster_data = env.get("cluster", None) cluster = tf.train.ClusterSpec(cluster_data) if cluster_data else None # Load the task data from the environment. task_data = env.get("task", None) or {"type": "master", "index": 0} task = type("TaskSpec", (object,), task_data) # Logging the version. logging.set_verbosity(tf.logging.INFO) logging.info("%s: Tensorflow version: %s.", task_as_string(task), tf.__version__) # Dispatch to a master, a worker, or a parameter server. if not cluster or task.type == "master" or task.type == "worker": Trainer(cluster, task, FLAGS.train_dir, FLAGS.log_device_placement).run( start_new_model=FLAGS.start_new_model) elif task.type == "ps": ParameterServer(cluster, task).run() else: raise ValueError("%s: Invalid task_type: %s." % (task_as_string(task), task.type))
Example #29
Source File: train.py From youtube-8m with Apache License 2.0 | 5 votes |
def optional_assign_weights(sess, weights_input, weights_assignment): if weights_input is not None: weights, length = get_video_weights_array() _ = sess.run(weights_assignment, feed_dict={weights_input: weights}) print "Assigned weights from %s" % FLAGS.sample_freq_file else: print "Collection weights_input not found"
Example #30
Source File: train-with-predictions.py From youtube-8m with Apache License 2.0 | 5 votes |
def main(unused_argv): # Load the environment. env = json.loads(os.environ.get("TF_CONFIG", "{}")) # Load the cluster data from the environment. cluster_data = env.get("cluster", None) cluster = tf.train.ClusterSpec(cluster_data) if cluster_data else None # Load the task data from the environment. task_data = env.get("task", None) or {"type": "master", "index": 0} task = type("TaskSpec", (object,), task_data) # Logging the version. logging.set_verbosity(tf.logging.INFO) logging.info("%s: Tensorflow version: %s.", task_as_string(task), tf.__version__) # Dispatch to a master, a worker, or a parameter server. if not cluster or task.type == "master" or task.type == "worker": Trainer(cluster, task, FLAGS.train_dir, FLAGS.log_device_placement).run( start_new_model=FLAGS.start_new_model) elif task.type == "ps": ParameterServer(cluster, task).run() else: raise ValueError("%s: Invalid task_type: %s." % (task_as_string(task), task.type))