Suppress matplot debug logs

lib/backend.py

@@ -26,7 +26,8 @@ from lib.data_types import (
     LogAction,
     ApiPayload_T,
     JsonDataException,
-    RequestMetrics
+    RequestMetrics,
+    BenchmarkResult
 )
 
 VERSION = "0.1.0"
@@ -332,18 +333,23 @@ class Backend:
 
             for run in range(1, self.benchmark_handler.benchmark_runs + 1):
                 start = time.time()
-                tasks = []
-                total_workload = 0
+                benchmark_requests = []
 
-                for _ in range(concurrent_requests):
+                for i in range(concurrent_requests):
                     payload = self.benchmark_handler.make_benchmark_payload()
-                    total_workload += payload.count_workload()
-                    tasks.append(
-                        self.__call_api(handler=self.benchmark_handler, payload=payload)
+                    workload = payload.count_workload()
+                    task = self.__call_api(handler=self.benchmark_handler, payload=payload)
+                    benchmark_requests.append(
+                        BenchmarkResult(request_idx=i, workload=workload, task=task)
                     )
 
-                responses = await gather(*tasks)
+                responses = await gather(*[br.task for br in benchmark_requests])
+                for br, response in zip(benchmark_requests, responses):
+                    br.response = response
+
+                total_workload = sum(br.workload for br in benchmark_requests if br.is_successful)
                 time_elapsed = time.time() - start
+                successful_responses = sum([1 for br in benchmark_requests if br.is_successful])
 
                 throughput = total_workload / time_elapsed
                 sum_throughput += throughput
@@ -357,7 +363,7 @@ class Backend:
                             f"Run: {run}, concurrent_requests: {concurrent_requests}",
                             f"Total workload: {total_workload}, time_elapsed: {time_elapsed}s",
                             f"Throughput: {throughput} workload/s",
-                            f"Successful responses: {len([r for r in responses if r.status == 200])}",
+                            f"Successful responses: {successful_responses}/{concurrent_requests}",
                             "#" * 60,
                         ]
                     )

lib/data_types.py

@@ -3,7 +3,7 @@ import logging
 from dataclasses import dataclass, field
 from enum import Enum
 from abc import ABC, abstractmethod
-from typing import Dict, Any, Union, Tuple, Optional, Set, TypeVar, Generic, Type
+from typing import Dict, Any, Union, Tuple, Optional, Set, TypeVar, Generic, Type, Awaitable
 from aiohttp import web, ClientResponse
 import inspect
 
@@ -206,6 +206,17 @@ class RequestMetrics:
     status: str
     success: bool = False
 
+@dataclass
+class BenchmarkResult:
+    request_idx: int
+    workload: float
+    task: Awaitable[ClientResponse]
+    response: Optional[ClientResponse] = None
+
+    @property
+    def is_successful(self) -> bool:
+        return self.response is not None and self.response.status == 200
+
 @dataclass
 class ModelMetrics:
     """Model specific metrics"""

lib/test_utils.py

@@ -292,12 +292,12 @@ def test_load_cmd(
     args = arg_parser.parse_args()
     if hasattr(args, "comfy_model"):
         os.environ["COMFY_MODEL"] = args.comfy_model
-    server_url = dict(
-        prod="https://run.vast.ai",
-        alpha="https://run-alpha.vast.ai",
-        candidate="https://run-candidate.vast.ai",
-        local="http://localhost:8080",
-    )[args.instance]
+    server_url = {
+        "prod": "https://run.vast.ai",
+        "alpha": "https://run-alpha.vast.ai",
+        "candidate": "https://run-candidate.vast.ai",
+        "local": "http://localhost:8080",
+    }.get(args.instance, "http://localhost:8080")
     run_test(
         num_requests=args.num_requests,
         requests_per_second=args.requests_per_second,