diff --git a/soundbay/trainers.py b/soundbay/trainers.py
index 9e3470a..681732f 100644
--- a/soundbay/trainers.py
+++ b/soundbay/trainers.py
@@ -102,7 +102,8 @@ def train_epoch(self, epoch):
             audio, label = audio.to(self.device), label.to(self.device)
 
             if (it == 0) and (not self.debug) and ((epoch % 5) == 0):
-                self.logger.upload_artifacts(audio, label, raw_wav, meta, sample_rate=self.train_dataloader.dataset.sample_rate, flag='train')
+                self.logger.upload_artifacts(audio, label, raw_wav, meta, sample_rate=self.train_dataloader.dataset.sample_rate,
+                                             flag='train', data_sample_rate=self.train_dataloader.dataset.data_sample_rate)
 
             # estimate and calc losses
             estimated_label = self.model(audio)
@@ -140,7 +141,8 @@ def eval_epoch(self, epoch: int, datatset_name: str = None):
                 audio, label, raw_wav, meta = batch
                 audio, label = audio.to(self.device), label.to(self.device)
                 if (it == 0) and (not self.debug) and ((epoch % 5) == 0):
-                    self.logger.upload_artifacts(audio, label, raw_wav, meta, sample_rate=self.train_dataloader.dataset.sample_rate, flag=datatset_name)
+                    self.logger.upload_artifacts(audio, label, raw_wav, meta, sample_rate=self.train_dataloader.dataset.sample_rate,
+                                                 flag=datatset_name, data_sample_rate=self.train_dataloader.dataset.data_sample_rate)
 
                 # estimate and calc losses
                 estimated_label = self.model(audio)
diff --git a/soundbay/utils/logging.py b/soundbay/utils/logging.py
index 630daa8..e51d326 100644
--- a/soundbay/utils/logging.py
+++ b/soundbay/utils/logging.py
@@ -153,7 +153,7 @@ def calc_metrics(self, epoch: int, mode: str = 'train', label_names: Optional[Li
         self.label_list = []
         self.pred_proba_list = []
 
-    def upload_artifacts(self, audio: torch.Tensor, label: torch.Tensor, raw_wav: torch.Tensor, meta: dict, sample_rate: int=16000, flag: str='train'):
+    def upload_artifacts(self, audio: torch.Tensor, label: torch.Tensor, raw_wav: torch.Tensor, meta: dict, sample_rate: int=16000, flag: str='train', data_sample_rate: int = 16000):
         """upload algorithm artifacts to W&B during training session"""
         volume = 50
         matplotlib.use('Agg')
@@ -171,7 +171,7 @@ def upload_artifacts(self, audio: torch.Tensor, label: torch.Tensor, raw_wav: to
 
         artifact_wav = torch.squeeze(raw_wav).detach().cpu().numpy()
         artifact_wav = artifact_wav / np.expand_dims(np.abs(artifact_wav).max(axis=1) + 1e-8, 1) * 0.5  # gain -6dB
-        list_of_wavs_objects = [wandb.Audio(data_or_path=wav, caption=f'{flag}_label{lab}_i{ind}_{round(b_t/sample_rate,3)}sec_{f_n}', sample_rate=sample_rate) for wav, ind, lab, b_t, f_n in zip(artifact_wav,idx, label, meta['begin_time'], meta['org_file'])]
+        list_of_wavs_objects = [wandb.Audio(data_or_path=wav, caption=f'{flag}_label{lab}_i{ind}_{round(b_t/data_sample_rate,2)}sec_{f_n}', sample_rate=sample_rate) for wav, ind, lab, b_t, f_n in zip(artifact_wav,idx, label, meta['begin_time'], meta['org_file'])]
 
         # Spectrograms batch
         artifact_spec = torch.squeeze(audio).detach().cpu().numpy()
@@ -181,7 +181,7 @@ def upload_artifacts(self, audio: torch.Tensor, label: torch.Tensor, raw_wav: to
             specs.append(librosa.display.specshow(artifact_spec[artifact_id,...], ax=ax[1]))
             plt.close('all')
             del ax
-        list_of_specs_objects = [wandb.Image(data_or_path=spec, caption=f'{flag}_label{lab}_i{ind}_{round(b_t/sample_rate,2)}sec_{f_n}') for spec, ind, lab, b_t, f_n in zip(specs,idx, label, meta['begin_time'], meta['org_file'])]
+        list_of_specs_objects = [wandb.Image(data_or_path=spec, caption=f'{flag}_label{lab}_i{ind}_{round(b_t/data_sample_rate,2)}sec_{f_n}') for spec, ind, lab, b_t, f_n in zip(specs,idx, label, meta['begin_time'], meta['org_file'])]
         log_wavs = {f'First batch {flag} original wavs': list_of_wavs_objects}
         log_specs = {f'First batch {flag} augmented spectrogram\'s': list_of_specs_objects}