Some errors when using multi-gpus

[lcf000000]

trafficstars

Training is too slow, and I want using a 4xTian xp to speed up it. when I set opts.gpus=[1,2,3,4] then I get some errors as:

`Warning: Function labindex has the same name as a MATLAB builtin. We suggest you rename the function to avoid a potential name conflict.

In path at 109 In addpath at 86 In cnn_train_dag at 11 In trainig_siameseFC at 97 In run_increasingData at 6 Warning: Function numlabs has the same name as a MATLAB builtin. We suggest you rename the function to avoid a potential name conflict. In path at 109 In addpath at 86 In cnn_train_dag at 11 In trainig_siameseFC at 97 In run_increasingData at 6 Starting parallel pool (parpool) using the 'local' profile ... connected to 1 workers. cnn_train_dag: resetting GPU Lab 1:
ans = Error using cnn_train_dag>(spmd) (line 492) The parallel pool that SPMD was using has been shut down.

Error in cnn_train_dag>prepareGPUs (line 492) spmd

Error in cnn_train_dag (line 92) prepareGPUs(opts, epoch == start+1) ;

Error in trainig_siameseFC (line 97) [net, stats] = cnn_train_dag(net, imdb, batch_fn, opts.train);

Error in run_increasingData (line 6) trainig_siameseFC(imdb_data, 86);

The client lost connection to worker 1. This might be due to network problems, or the interactive communicating job might have errored. `

[jvlmdr]

Sorry, we haven't tested support for training with multiple GPUs. These look like errors coming from matconvnet and/or the parallel toolbox.