Many to one, how to set 'max_grad_norm'

[cobnut]

Many to one, like this, but i don't know how to set 'max_grad_norm' ? what's the meaning of 'max_grad_norm'? how to set 'max_grad_norm'?

network = tl.layers.DynamicRNNLayer(network,
                        cell_fn=tf.nn.rnn_cell.BasicLSTMCell,
                        cell_init_args={'forget_bias': 0.0, 'state_is_tuple': True},
                        n_hidden=hidden_size,
                        n_layer=2,
                        dropout=0.5,
                        initializer=tf.random_uniform_initializer(-init_scale, init_scale),
                        sequence_length = tl.layers.retrieve_seq_length_op2(inputs),
                        return_last=True,
                        name='dynamic_lstm_layer')
lstm = network

[zsdonghao]

@narrator-wong you can find the example in PTB tutorials.

    grads, _ = tf.clip_by_global_norm(tf.gradients(cost, tvars),
                                      max_grad_norm)
    optimizer = tf.train.GradientDescentOptimizer(lr)
    train_op = optimizer.apply_gradients(zip(grads, tvars))

[cobnut]

@zsdonghao Thanks...
i saw this, but i mean.. for example, set max_grad_norm = 10 or 20 or 50 ...? what's the difference?
i set 'max_grad_norm' according to my num_step?
if the num_step of my LSTM is 100, and my max_grad_norm also 100?

[zsdonghao]

@narrator-wong Gitter may be the better place to discuss algorithm, we normally report bugs on Issues.