Use log probs for paraformer (#120)

* Use log probs for paraformer

* Fix

@@ -23,8 +23,7 @@ class OfflineParaformerDecoder {
   /** Run beam search given the output from the paraformer model.
   /** Run beam search given the output from the paraformer model.
    *
    *
    * @param log_probs A 3-D tensor of shape (N, T, vocab_size)
    * @param log_probs A 3-D tensor of shape (N, T, vocab_size)
-   * @param token_num A 2-D tensor of shape (N, T). Its dtype is int64_t.
-   *                  log_probs[i].argmax(axis=-1) equals to token_num[i]
+   * @param token_num A 1-D tensor of shape (N). token_num equals to T.
    *
    *
    * @return Return a vector of size `N` containing the decoded results.
    * @return Return a vector of size `N` containing the decoded results.
    */
    */

@@ -4,28 +4,33 @@
 
 
 #include "sherpa-onnx/csrc/offline-paraformer-greedy-search-decoder.h"
 #include "sherpa-onnx/csrc/offline-paraformer-greedy-search-decoder.h"
 
 
+#include <algorithm>
 #include <vector>
 #include <vector>
 
 
 namespace sherpa_onnx {
 namespace sherpa_onnx {
 
 
 std::vector<OfflineParaformerDecoderResult>
 std::vector<OfflineParaformerDecoderResult>
-OfflineParaformerGreedySearchDecoder::Decode(Ort::Value /*log_probs*/,
-                                             Ort::Value token_num) {
-  std::vector<int64_t> shape = token_num.GetTensorTypeAndShapeInfo().GetShape();
+OfflineParaformerGreedySearchDecoder::Decode(Ort::Value log_probs,
+                                             Ort::Value /*token_num*/) {
+  std::vector<int64_t> shape = log_probs.GetTensorTypeAndShapeInfo().GetShape();
   int32_t batch_size = shape[0];
   int32_t batch_size = shape[0];
   int32_t num_tokens = shape[1];
   int32_t num_tokens = shape[1];
+  int32_t vocab_size = shape[2];
 
 
   std::vector<OfflineParaformerDecoderResult> results(batch_size);
   std::vector<OfflineParaformerDecoderResult> results(batch_size);
 
 
-  const int64_t *p = token_num.GetTensorData<int64_t>();
   for (int32_t i = 0; i != batch_size; ++i) {
   for (int32_t i = 0; i != batch_size; ++i) {
+    const float *p =
+        log_probs.GetTensorData<float>() + i * num_tokens * vocab_size;
     for (int32_t k = 0; k != num_tokens; ++k) {
     for (int32_t k = 0; k != num_tokens; ++k) {
-      if (p[k] == eos_id_) break;
+      auto max_idx = static_cast<int64_t>(
+          std::distance(p, std::max_element(p, p + vocab_size)));
+      if (max_idx == eos_id_) break;
 
 
-      results[i].tokens.push_back(p[k]);
-    }
+      results[i].tokens.push_back(max_idx);
 
 
-    p += num_tokens;
+      p += vocab_size;
+    }
   }
   }
 
 
   return results;
   return results;

@@ -17,7 +17,7 @@ class OfflineParaformerGreedySearchDecoder : public OfflineParaformerDecoder {
       : eos_id_(eos_id) {}
       : eos_id_(eos_id) {}
 
 
   std::vector<OfflineParaformerDecoderResult> Decode(
   std::vector<OfflineParaformerDecoderResult> Decode(
-      Ort::Value /*log_probs*/, Ort::Value token_num) override;
+      Ort::Value log_probs, Ort::Value /*token_num*/) override;
 
 
  private:
  private:
   int32_t eos_id_;
   int32_t eos_id_;