diff --git a/gpttype_adapter.cpp b/gpttype_adapter.cpp
index 9c52feeb4..7bba001cc 100644
--- a/gpttype_adapter.cpp
+++ b/gpttype_adapter.cpp
@@ -2083,10 +2083,12 @@ generation_outputs gpttype_generate(const generation_inputs inputs, generation_o
     }
     time2 = timer_check();
     float pt1 = (time1*1000.0/(embd_inp.size()==0?1:embd_inp.size()));
+    float ts1 = (1000.0/pt1);
     int realnpredict = kcpp_params->n_predict-stopper_unused_tokens;
     float pt2 = (time2*1000.0/(realnpredict==0?1:realnpredict));
+    float ts2 = (1000.0/pt2);
     float tokens_per_second = (realnpredict == 0 ? 0 : realnpredict / (time1 + time2));
-    printf("\nContextLimit: %d/%d, Processing:%.2fs (%.1fms/T), Generation:%.2fs (%.1fms/T), Total:%.2fs (%.1fms/T = %.2fT/s)",current_context_tokens.size(),nctx, time1, pt1, time2, pt2, (time1 + time2), (1000.0f/tokens_per_second) , tokens_per_second);
+    printf("\nContextLimit: %d/%d, Processing:%.2fs (%.1fms/T = %.2fT/s), Generation:%.2fs (%.1fms/T = %.2fT/s), Total:%.2fs (%.1fms/T = %.2fT/s)",current_context_tokens.size(),nctx, time1, pt1, ts1, time2, pt2, ts2, (time1 + time2), (1000.0f/tokens_per_second) , tokens_per_second);
     fflush(stdout);
     output.status = 1;
     generation_finished = true;
diff --git a/koboldcpp.py b/koboldcpp.py
index 8f72fbe2c..dd02ba239 100755
--- a/koboldcpp.py
+++ b/koboldcpp.py
@@ -334,9 +334,9 @@ def generate(prompt, memory="", max_length=32, max_context_length=512, temperatu
     outputs = ctypes.create_unicode_buffer(ctypes.sizeof(generation_outputs))
     inputs.prompt = prompt.encode("UTF-8")
     inputs.memory = memory.encode("UTF-8")
-    if max_length >= max_context_length:
+    if max_length >= (max_context_length-1):
         max_length = max_context_length-1
-        print("\nWARNING: You are trying to generate with max_length near or exceeding max_context_length. Most of the context will be gone and your outputs will not be very coherent.")
+        print("\nWarning: You are trying to generate with max_length near or exceeding max_context_length. Most of the context will be removed, and your outputs will not be very coherent.")
     global showmaxctxwarning
     if max_context_length > maxctx:
         if showmaxctxwarning: