Autograder [Thu Apr 16 15:42:58 2020]: Received job 11785-s20_hw4p1_9_clairew1@andrew.cmu.edu:1730 Autograder [Thu Apr 16 15:43:37 2020]: Success: Autodriver returned normally Autograder [Thu Apr 16 15:43:37 2020]: Here is the output from the autograder: --- Autodriver: Job exited with status 0 mkdir -p handin tar xf handin.tar -C handin tar xf autograde.tar AUTOLAB=1 /usr/local/depot/anaconda3/bin/python3 autograde/runner.py --module-path=./handin/ Your mean NLL for generated sequences: 4.660201072692871 FYour mean NLL for predicting a single word: 5.652750015258789 F =================================== FAILURES =================================== _______________________________ test_generation ________________________________ def test_generation(): inp = np.load(fixture_path('generation.npy')) forward = 10 n = inp.shape[0] t = inp.shape[1] pred = np.load(handin_path('generated_logits.npy')) assert pred.shape[0] == n assert pred.shape[1] == forward ninp = np.concatenate((inp, pred), axis=1) vocab = np.load(fixture_path('vocab.npy')) model = Model(vocab.shape[0], args=prediction_args) model.load_state_dict(load_from_numpy(read_chunks(fixture_path('model-00000099.tar.npy.{}')))) model.eval() state = model.zero_state(n) x = Variable(torch.from_numpy(ninp.T)).long() logits, rnn_hs = model(x, state) logits = logits.cpu().data.numpy() logits = log_softmax(logits, -1) logits = logits[t - 1:-1, :, :] logits = np.transpose(logits, (1, 0, 2)) mg = np.meshgrid(np.arange(n), np.arange(forward), indexing='ij') nlls = logits[mg[0], mg[1], pred] nll = -np.mean(nlls) print("Your mean NLL for generated sequences: {}".format(nll)) > assert nll < 3. E assert 4.6602011 < 3.0 autograde/tests/test_generation.py:245: AssertionError _______________________________ test_prediction ________________________________ def test_prediction(): fixture = np.load(fixture_path('prediction.npz')) inp = fixture['inp'] targ = fixture['out'] out = np.load(handin_path('predictions.npy')) assert out.shape[0] == targ.shape[0] vocab = np.load(fixture_path('vocab.npy')) assert out.shape[1] == vocab.shape[0] out = log_softmax(out, 1) nlls = out[np.arange(out.shape[0]), targ] nll = -np.mean(nlls) print("Your mean NLL for predicting a single word: {}".format(nll)) > assert nll < 5.4 E assert 5.65275 < 5.4 autograde/tests/test_prediction.py:31: AssertionError Run time: 22.351541757583618 {"scores": {"Generation": 0.0, "Prediction": 0.0}}