[src] Make user parameter check ERROR not ASSERT (kaldi-asr#4181)

Fixes kaldi-asr#3183

src/chain/chain-generic-numerator.cc

@@ -39,9 +39,21 @@ GenericNumeratorComputation::GenericNumeratorComputation(
     supervision_(supervision),
     nnet_output_(nnet_output),
     opts_(opts) {
-  KALDI_ASSERT(supervision.num_sequences *
-               supervision.frames_per_sequence == nnet_output.NumRows() &&
-               supervision.label_dim == nnet_output.NumCols());
+  if (supervision.num_sequences * supervision.frames_per_sequence !=
+      nnet_output.NumRows()) {
+    KALDI_ERR << "Dimension mismatch: nnet_output number of rows "
+              << nnet_output.NumRows() << " must be equal to the total number "
+              << "of frames in the supervision object: (num_sequences="
+              << supervision.num_sequences << ") * (frames_per_sequence="
+              << supervision.frames_per_sequence << ") = "
+              << supervision.num_sequences * supervision.frames_per_sequence;
+  }
+  if (supervision.label_dim != nnet_output.NumCols()) {
+    KALDI_ERR << "Dimension mismatch: nnet_output number of columns "
+              << nnet_output.NumCols() << " is not equal to the supervision "
+              << "object feature dimension label_dim=" << supervision.label_dim;
+  }
+
   NVTX_RANGE(__func__);
 
   using std::vector;
@@ -242,13 +254,13 @@ bool GenericNumeratorComputation::ForwardBackward(
   unsigned int nthreads = opts_.num_threads > 0 ? opts_.num_threads :
                               std::thread::hardware_concurrency();
   // Naive load balancing, each thread gets a chunk of the sequences to process
-  unsigned int num_sequences_per_thread = 
+  unsigned int num_sequences_per_thread =
     (num_sequences + nthreads - 1) / nthreads;
 
   // Allocate one alpha and beta matrix per thread to avoid contention
   std::vector<Matrix<BaseFloat>> alpha(nthreads);
   std::vector<Matrix<BaseFloat>> beta(nthreads);
-  
+
   // Per thread partial values and boolean
   std::vector<BaseFloat> partial_loglike_mt(nthreads, static_cast<BaseFloat>(0));
   std::vector<bool> ok_mt(nthreads, true);