Merge branch 'mlcommons:master' into master

GATEOverflow · Aug 22, 2024 · f4ba37d · f4ba37d
2 parents 66a9f10 + c7db1c3
commit f4ba37d
Show file tree

Hide file tree

Showing 2 changed files with 34 additions and 1 deletion.
diff --git a/tools/submission/generate_final_report.py b/tools/submission/generate_final_report.py
@@ -111,6 +111,7 @@ def main():
                        'Latency (ms)',
                        'Samples/s',
                        'Queries/s',
+                       'Tokens/s',
                        'millijoules',
                        'Watts',
                    ]]

diff --git a/tools/submission/submission_checker.py b/tools/submission/submission_checker.py
@@ -1592,6 +1592,38 @@ def log_result(
         if system_json.get("sw_notes"):
             notes = notes + ". " if notes else ""
             notes = notes + system_json.get("sw_notes")
+        special_unit_dict = {
+            "gptj-99": {
+                "SingleStream": "Latency (ms)",
+                "MultiStream": "Latency (ms)",
+                "Offline": "Tokens/s",
+                "Server": "Tokens/s",
+            },
+            "gptj-99.9": {
+                "SingleStream": "Latency (ms)",
+                "MultiStream": "Latency (ms)",
+                "Offline": "Tokens/s",
+                "Server": "Tokens/s",
+            },
+            "llama2-70b-99" : {
+                "SingleStream": "Latency (ms)",
+                "MultiStream": "Latency (ms)",
+                "Offline": "Tokens/s",
+                "Server": "Tokens/s",
+            },
+            "llama2-70b-99.9" : {
+                "SingleStream": "Latency (ms)",
+                "MultiStream": "Latency (ms)",
+                "Offline": "Tokens/s",
+                "Server": "Tokens/s",
+            },
+            "mixtral-8x7b" : {
+                "SingleStream": "Latency (ms)",
+                "MultiStream": "Latency (ms)",
+                "Offline": "Tokens/s",
+                "Server": "Tokens/s",
+            }
+        }
         unit_dict = {
             "SingleStream": "Latency (ms)",
             "MultiStream": "Latency (ms)",
@@ -1604,7 +1636,7 @@ def log_result(
             "Offline": "Watts",
             "Server": "Watts",
         }
-        unit = unit_dict[scenario_fixed]
+        unit = special_unit_dict.get(model_name, unit_dict)[scenario_fixed]
         power_unit = power_unit_dict[scenario_fixed]
 
         csv.write(