more lenient

Author: dbogunowicz

src/deepsparse/transformers/pipelines/text_generation.py

@@ -834,7 +834,7 @@ def engine_forward(
                     generated_tokens.append(token)
                     generated_logits.append(logits)
 
-                    if session.total_num_processed_tokens >= session.capacity:
+                    if session.total_num_processed_tokens > session.capacity:
                         # if the kv cache is full, stop generation
                         finished_reason.append(FinishReason.CAPACITY)
                         break