search

iamjanvijay / rnnt

An implementation of RNN-Transducer loss in TF-2.0.
MIT License
45 stars 9 forks source link

Invalid argument: indices[0,0,2] = [0, 0, 2, -1] does not index into shape [10,79,18,6484] #7

Open zzpDapeng opened 3 years ago

zzpDapeng commented 3 years ago

logits shape: (10, 79, 18, 6485) labels shape (10, 17) labels_length: tf.Tensor([ 2 14 9 9 9 13 17 9 9 17], shape=(10,), dtype=int64) logit_length : tf.Tensor([20 47 36 35 41 58 64 38 45 78], shape=(10,), dtype=int64)

I got this error： 2020-12-11 10:48:09.516460: W tensorflow/core/framework/op_kernel.cc:1767] OP_REQUIRES failed at scatter_nd_op.cc:133 : Invalid argument: indices[0,0,2] = [0, 0, 2, -1] does not index into shape [10,79,18,6484] Traceback (most recent call last): File "/home/dapeng/PycharmProjects/convTT/train.py", line 68, in train(model, train_set, optimizer, train_loss, epoch) File "/home/dapeng/PycharmProjects/convTT/train.py", line 33, in train label_length=labels_length) File "/home/dapeng/anaconda3/envs/tf/lib/python3.7/site-packages/rnnt/rnnt.py", line 204, in rnnt_loss return compute_rnnt_loss_and_grad(args) File "/home/dapeng/anaconda3/envs/tf/lib/python3.7/site-packages/tensorflow/python/ops/custom_gradient.py", line 264, in call return self._d(self._f, a, k) File "/home/dapeng/anaconda3/envs/tf/lib/python3.7/site-packages/tensorflow/python/ops/custom_gradient.py", line 218, in decorated return _eager_mode_decorator(wrapped, args, kwargs) File "/home/dapeng/anaconda3/envs/tf/lib/python3.7/site-packages/tensorflow/python/ops/custom_gradient.py", line 412, in _eager_mode_decorator result, grad_fn = f(args, kwargs) File "/home/dapeng/anaconda3/envs/tf/lib/python3.7/site-packages/rnnt/rnnt.py", line 195, in compute_rnnt_loss_and_grad result = compute_rnnt_loss_and_grad_helper(kwargs) File "/home/dapeng/anaconda3/envs/tf/lib/python3.7/site-packages/rnnt/rnnt.py", line 168, in compute_rnnt_loss_and_grad_helper [batch_size, input_max_len, target_max_len, vocab_size - 1]) File "/home/dapeng/anaconda3/envs/tf/lib/python3.7/site-packages/tensorflow/python/ops/gen_array_ops.py", line 8842, in scatter_nd indices, updates, shape, name=name, ctx=_ctx) File "/home/dapeng/anaconda3/envs/tf/lib/python3.7/site-packages/tensorflow/python/ops/gen_array_ops.py", line 8885, in scatter_nd_eager_fallback attrs=_attrs, ctx=ctx, name=name) File "/home/dapeng/anaconda3/envs/tf/lib/python3.7/site-packages/tensorflow/python/eager/execute.py", line 60, in quick_execute inputs, attrs, num_outputs) tensorflow.python.framework.errors_impl.InvalidArgumentError: indices[0,0,2] = [0, 0, 2, -1] does not index into shape [10,79,18,6484] [Op:ScatterNd]

zzpDapeng commented 3 years ago

I run your Sample Train Script and found that you set lables=decoder_seqs + 1 in return tf.reduce_mean(self.loss(log_probs, decoder_seqs + 1, decoder_lens, encoder_lens)) If i change your code to return tf.reduce_mean(self.loss(log_probs, decoder_seqs, decoder_lens, encoder_lens)), it lead to the same error. I can't understand why you do that. In my dictionary, I reserve index 0 for blank, and the word indexed from 1 to 6484 (6485 at all), so my decoder_seqs represent the real index of word, why plus 1?

zzpDapeng commented 3 years ago

my labels is padded with 0 as follows: [[ 379 57 230 88 214 268 907 1061 62 142 2027 0 0 0 0 0] [ 588 459 357 547 209 102 101 86 3 96 87 89 32 0 0 0] [ 144 942 196 571 1956 219 82 908 46 363 1148 696 634 55 761 383] [ 624 316 378 716 1737 362 49 110 746 1157 346 380 570 0 0 0] [ 781 260 1311 1562 15 429 16 666 150 0 0 0 0 0 0 0] [ 809 411 310 178 541 1209 2 99 102 87 232 88 231 0 0 0] [ 704 6 3 570 167 151 3 290 101 102 87 89 2287 0 0 0] [ 3 645 146 3748 4067 4735 869 60 1407 95 1334 4735 869 0 0 0] [ 52 342 372 1020 440 441 628 440 441 628 1154 2121 378 233 766 0] [1180 478 1087 1687 2381 777 174 231 102 97 174 0 0 0 0 0]]