Cannot cast array error when running start_mirt_pipeline.py

[brunojm]

I'm getting:

$ python start_mirt_pipeline.py --generate --train -n 2 --visualize Generating Responses Generated responses for 500 students and 10 problems Training MIRT models Starting main.{'training_set_size': 1.0, 'regularization': 1e-05, 'emit_features': False, 'resume_from_file': '', 'max_pass_lbfgs': 5, 'sampling_num_steps': 200, 'workers': 1, 'num_epochs': 2, 'data_format': 'simple', 'sampling_epsilon': 0.2, 'num_replicas': 1, 'file': '/Users/bruno/Development/MachineLearning/guacamole/sample_data/models/train.responses', 'time': False, 'output': '/Users/bruno/Development/MachineLearning/guacamole/sample_data/models/1_no_time_2014-12-13 08:59:30.905104/', 'num_abilities': 1, 'max_time_taken': 1000.0} loading dataTraining dataset, 436 students 10 exercises epoch 0, Traceback (most recent call last): File "start_mirt_pipeline.py", line 263, in main() File "start_mirt_pipeline.py", line 173, in main run_with_arguments(arguments) File "start_mirt_pipeline.py", line 237, in run_with_arguments generate_model_with_parameters(arguments) File "start_mirt_pipeline.py", line 205, in generate_model_with_parameters mirt_train_EM.run_programmatically(mirt_train_params) File "/Users/bruno/Development/MachineLearning/guacamole/mirt/mirt_train_EM.py", line 166, in run_programmatically run(options) File "/Users/bruno/Development/MachineLearning/guacamole/mirt/mirt_train_EM.py", line 249, in run model.run_em_step(epoch) File "/Users/bruno/Development/MachineLearning/guacamole/mirt/mirt_util.py", line 516, in run_em_step results = self.get_sampling_results() File "/Users/bruno/Development/MachineLearning/guacamole/mirt/mirt_util.py", line 496, in get_sampling_results for ind in range(len(self.user_states))] File "/Users/bruno/Development/MachineLearning/guacamole/mirt/mirt_util.py", line 260, in sample_abilities_diffusion_wrapper np.random.seed([time.time() * 1e9]) File "mtrand.pyx", line 652, in mtrand.RandomState.seed (numpy/random/mtrand/mtrand.c:7775) TypeError: Cannot cast array from dtype('float64') to dtype('int64') according to the rule 'safe'

I'm running in a MacOS 10.9.5, my pip freeze for the env for this project:

affinity==0.1.0 matplotlib==1.4.2 mock==1.0.1 nose==1.3.4 numpy==1.9.1 pyparsing==2.0.3 python-dateutil==2.3 pytz==2014.10 scipy==0.14.0 six==1.8.0 wsgiref==0.1.2

[himanshunimje]

@brunojm Did you get an answer to this? I am also stuck at the same error. "TypeError: Cannot cast array from dtype('float64') to dtype('int64') according to the rule 'safe'"

[imoonkey]

Same problem here. wondering what to do.

Update: It is not working on my mac (OS X Yosemite 10.10.5) or my Ubuntu 14.04.1 LTS server.

[imoonkey]

This is because numpy.random.seed()'s input seed must be convertable to 32 bit unsigned integers. Change the code in mirt/mirt_util.py starting from line 257 to the following code fixes this

    if len(id) > 0:
        np.random.seed([id[0], int(time.time() * 1e9) % 4294967296])
    else:
        np.random.seed([int(time.time() * 1e9) % 4294967296])

[ppkn]

This worked for me as well. I would create a pull request but I'm not sure if there is a less hacky way of fixing it.