janelu9 commented on issue #8065: Help? when I go on training a language model, errors occured URL: https://github.com/apache/incubator-mxnet/issues/8065#issuecomment-332729914 import mxnet as mx #mx.symbol_doc.SymbolDoc.get_output_shape() def N_gamma(seq_len,batch_size,vocab_size,embed_dim,hidden_num,N): data = mx.sym.Variable('data') embed_weight=mx.sym.Variable('embed_weight') H_weight=mx.sym.Variable('H_weight') d_bias=mx.sym.Variable('d_bias') U_weight=mx.sym.Variable('U_weight') W_weight=mx.sym.Variable('W_weight') b_bias=mx.sym.Variable('b_bias') b0_bias=mx.sym.Variable('b0_bias') label = mx.sym.Variable('softmax_label') embed = mx.sym.Embedding(data=data, input_dim=vocab_size, weight=embed_weight,output_dim=embed_dim, name='embed') wordvec = mx.sym.SliceChannel(data=embed,num_outputs=seq_len+N-1,squeeze_axis=1) y_all=[] for seq_idx in range(N-1,seq_len+N-1): word=[wordvec[i] for i in xrange(seq_idx-N+1,seq_idx+1)] hidden_concat = mx.sym.Concat(*word, dim=1) tanh_hidden=mx.sym.FullyConnected(data=hidden_concat,num_hidden=hidden_num,weight=H_weight,bias=d_bias,name='tanh_hidden') tanh_act=mx.sym.Activation(data=tanh_hidden, act_type="tanh") tanh_part=mx.sym.FullyConnected(data=tanh_act,num_hidden=vocab_size,weight=U_weight,bias=b_bias,name='tanh_part') fc_part=mx.sym.FullyConnected(data=hidden_concat,num_hidden=vocab_size,weight=W_weight,bias=b0_bias,name='fc_part') y=tanh_part+fc_part y_all.append(y) y_concat=mx.sym.Concat(*y_all,dim=0) label = mx.sym.transpose(data=label) label = mx.sym.Reshape(data=label, target_shape=(0,)) sm = mx.sym.SoftmaxOutput(data=y_concat, label=label, name='softmax') return sm ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected]
With regards, Apache Git Services
