fix bug

app.py

@@ -132,7 +132,7 @@ with col2:
 with col3: # Prefilling
     prefilling_operation_count = prefilling_operation(model_config, inference_config)
     inference_info['inference_prefilling_time'] = prefilling_operation_count['total'] / (gpu_config['TFLOP']*10**12)
-    inference_info['inference_prefilling_throughput'] = inference_config['input_seq_length']/inference_info['inference_prefilling_time']
+    inference_info['inference_prefilling_throughput'] = inference_config['input_seq_length']*inference_config['batchsize']/inference_info['inference_prefilling_time']
     cached_parameter_count['kv_cache'] = 2 * (inference_config['batchsize'] * (model_config['hidden_size'] * model_config['num_hidden_layers'] * inference_config['input_seq_length']))
 
     operation_items = {key: "{:,}".format(int(prefilling_operation_count[key])) for key in prefilling_operation_count if key not in subtotal_operations}
@@ -157,8 +157,8 @@ with col3: # Prefilling
 with col4: # Prefilling
     generation_operation_count = generation_operation(model_config, inference_config)
     inference_info['inference_generation_time'] = generation_operation_count['total'] / (gpu_config['TFLOP']*10**12)
-    inference_info['inference_generation_throughput'] = inference_config['output_seq_length']/inference_info['inference_generation_time']
-    inference_info['inference_client_generation_throughput'] = inference_config['output_seq_length'] / (inference_info['inference_prefilling_time'] + inference_info['inference_generation_time'])
+    inference_info['inference_generation_throughput'] = inference_config['output_seq_length']*inference_config['batchsize']/inference_info['inference_generation_time']
+    inference_info['inference_client_generation_throughput'] = inference_config['output_seq_length']*inference_config['batchsize'] / (inference_info['inference_prefilling_time'] + inference_info['inference_generation_time'])
     cached_parameter_count['kv_cache'] = 2 * (inference_config['batchsize'] * (model_config['hidden_size'] * model_config['num_hidden_layers'] * (inference_config['input_seq_length']+inference_config['output_seq_length'])))
 
     operation_items = {key: "{:,}".format(int(generation_operation_count[key])) for key in generation_operation_count if key not in subtotal_operations}