Changing name to be avg_token_to_token...

triton-inference-server · Oct 12, 2023 · dbe845b · dbe845b
1 parent 4f111a4
commit dbe845b
Show file tree

Hide file tree

Showing 3 changed files with 8 additions and 6 deletions.
diff --git a/model_analyzer/perf_analyzer/perf_analyzer.py b/model_analyzer/perf_analyzer/perf_analyzer.py
@@ -37,7 +37,9 @@
 from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
 from model_analyzer.record.record import Record
 from model_analyzer.record.types.avg_first_token_latency import AvgFirstTokenLatency
-from model_analyzer.record.types.avg_token_latency import AvgTokenLatency
+from model_analyzer.record.types.avg_token_to_token_latency import (
+    AvgTokenToTokenLatency,
+)
 from model_analyzer.record.types.gpu_free_memory import GPUFreeMemory
 from model_analyzer.record.types.gpu_power_usage import GPUPowerUsage
 from model_analyzer.record.types.gpu_used_memory import GPUUsedMemory
@@ -95,8 +97,8 @@ class PerfAnalyzer:
     ]
 
     llm_metric_table = [
-        ["avg_first_latency",           "Avg first token latency",     AvgFirstTokenLatency,     "1000"],
-        ["avg_avg_latency",           "Avg token latency",           AvgTokenLatency,          "1000"]
+        ["avg_first_latency",          None,                    AvgFirstTokenLatency,     "1000"],
+        ["avg_token_to_token_latency", None,                    AvgTokenToTokenLatency,   "1000"]
     ]
     # yapf: enable
 

diff --git a/...nalyzer/record/types/avg_token_latency.py → ...ecord/types/avg_token_to_token_latency.py b/...nalyzer/record/types/avg_token_latency.py → ...ecord/types/avg_token_to_token_latency.py
@@ -20,12 +20,12 @@
 
 
 @total_ordering
-class AvgTokenLatency(DecreasingRecord):
+class AvgTokenToTokenLatency(DecreasingRecord):
     """
     A record for perf_analyzer avg token-to-token latency metric
     """
 
-    tag = "avg_token_latency"
+    tag = "avg_token_to_token_latency"
 
     def __init__(self, value, timestamp=0):
         """

diff --git a/tests/test_record_types.py b/tests/test_record_types.py
@@ -60,7 +60,7 @@ def setUp(self):
                 "perf_server_compute_input",
                 "gpu_power_usage",
                 "avg_first_token_latency",
-                "avg_token_latency",
+                "avg_token_to_token_latency",
             ]
         }
         self.more_is_better_types = {