print for debugging

utils/pipelines.py

@@ -175,6 +175,7 @@ def generate_ppl_scores_approx(sequences_batch, model, tokenizer, mask_percentag
     ppl_scores : list of float
         A list of approximate perplexity scores for each input sequence.
     """
+    print(sequences_batch)
     device = model.device
     mask_token_id = tokenizer.mask_token_id
     if mask_token_id is None:
@@ -264,6 +265,7 @@ def generate_ppl_scores_approx(sequences_batch, model, tokenizer, mask_percentag
             avg_neg_log_prob = -log_prob_sums[i] / token_counts[i]
             ppl_scores.append(float(torch.exp(avg_neg_log_prob).item()))
     
+    print(ppl_scores)
     return ppl_scores
 
 def full_embedding_pipeline(fasta_files, model, tokenizer, batch_size, max_duration):