From 094a306ea2867b246814ab67728327523aea4724 Mon Sep 17 00:00:00 2001 From: Srivatsan Krishnan Date: Fri, 16 Jan 2026 17:41:24 -0800 Subject: [PATCH 1/5] fix bug in report generation log name --- .../megatron_bridge/report_generation_strategy.py | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/src/cloudai/workloads/megatron_bridge/report_generation_strategy.py b/src/cloudai/workloads/megatron_bridge/report_generation_strategy.py index 9aa3c2d34..48768fbb1 100644 --- a/src/cloudai/workloads/megatron_bridge/report_generation_strategy.py +++ b/src/cloudai/workloads/megatron_bridge/report_generation_strategy.py @@ -31,12 +31,12 @@ class MegatronBridgeReportGenerationStrategy(ReportGenerationStrategy): metrics: ClassVar[list[str]] = ["default", "step-time", "tflops-per-gpu"] def get_log_file(self) -> Path | None: - log = self.test_run.output_path / "megatron_bridge_launcher.log" + log = self.test_run.output_path / "cloudai_megatron_bridge_launcher.log" return log if log.is_file() else None @property def results_file(self) -> Path: - return self.get_log_file() or (self.test_run.output_path / "megatron_bridge_launcher.log") + return self.get_log_file() or (self.test_run.output_path / "cloudai_megatron_bridge_launcher.log") def can_handle_directory(self) -> bool: return self.get_log_file() is not None @@ -75,8 +75,8 @@ def generate_report(self) -> None: log_file, step_times_s, gpu_tflops = self._get_extracted_data() if not log_file: logging.error( - "No Megatron-Bridge launcher log file found: %s", - self.test_run.output_path / "megatron_bridge_launcher.log", + "No Megatron-Bridge launcher log file found in: %s", + self.test_run.output_path, ) return @@ -130,8 +130,8 @@ def get_metric(self, metric: str) -> float: log_file, step_times_s, gpu_tflops = self._get_extracted_data() if not log_file: logging.error( - "No Megatron-Bridge launcher log file found: %s", - self.test_run.output_path / "megatron_bridge_launcher.log", + "No Megatron-Bridge launcher log file found in: %s", + self.test_run.output_path, ) return METRIC_ERROR if not step_times_s: From c2d3ec35fae5678f31bf269d12aabfdb9803ab09 Mon Sep 17 00:00:00 2001 From: Srivatsan Krishnan Date: Fri, 16 Jan 2026 17:44:27 -0800 Subject: [PATCH 2/5] fix unit test --- .../test_megatron_bridge_report_generation_strategy.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tests/report_generation_strategy/test_megatron_bridge_report_generation_strategy.py b/tests/report_generation_strategy/test_megatron_bridge_report_generation_strategy.py index 78c3a3055..13e3be34e 100644 --- a/tests/report_generation_strategy/test_megatron_bridge_report_generation_strategy.py +++ b/tests/report_generation_strategy/test_megatron_bridge_report_generation_strategy.py @@ -38,7 +38,7 @@ def mb_tr(tmp_path: Path) -> TestRun: "", ] ) - (tr.output_path / "megatron_bridge_launcher.log").write_text(log_content) + (tr.output_path / "cloudai_megatron_bridge_launcher.log").write_text(log_content) return tr From 07d13e1b8f6162c8498c01d922b706f4af1b4d7d Mon Sep 17 00:00:00 2001 From: Srivatsan Krishnan Date: Fri, 16 Jan 2026 17:41:24 -0800 Subject: [PATCH 3/5] fix bug in report generation log name --- .../megatron_bridge/report_generation_strategy.py | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/src/cloudai/workloads/megatron_bridge/report_generation_strategy.py b/src/cloudai/workloads/megatron_bridge/report_generation_strategy.py index 9aa3c2d34..48768fbb1 100644 --- a/src/cloudai/workloads/megatron_bridge/report_generation_strategy.py +++ b/src/cloudai/workloads/megatron_bridge/report_generation_strategy.py @@ -31,12 +31,12 @@ class MegatronBridgeReportGenerationStrategy(ReportGenerationStrategy): metrics: ClassVar[list[str]] = ["default", "step-time", "tflops-per-gpu"] def get_log_file(self) -> Path | None: - log = self.test_run.output_path / "megatron_bridge_launcher.log" + log = self.test_run.output_path / "cloudai_megatron_bridge_launcher.log" return log if log.is_file() else None @property def results_file(self) -> Path: - return self.get_log_file() or (self.test_run.output_path / "megatron_bridge_launcher.log") + return self.get_log_file() or (self.test_run.output_path / "cloudai_megatron_bridge_launcher.log") def can_handle_directory(self) -> bool: return self.get_log_file() is not None @@ -75,8 +75,8 @@ def generate_report(self) -> None: log_file, step_times_s, gpu_tflops = self._get_extracted_data() if not log_file: logging.error( - "No Megatron-Bridge launcher log file found: %s", - self.test_run.output_path / "megatron_bridge_launcher.log", + "No Megatron-Bridge launcher log file found in: %s", + self.test_run.output_path, ) return @@ -130,8 +130,8 @@ def get_metric(self, metric: str) -> float: log_file, step_times_s, gpu_tflops = self._get_extracted_data() if not log_file: logging.error( - "No Megatron-Bridge launcher log file found: %s", - self.test_run.output_path / "megatron_bridge_launcher.log", + "No Megatron-Bridge launcher log file found in: %s", + self.test_run.output_path, ) return METRIC_ERROR if not step_times_s: From 59de2213c608bb7d364ae481cb753a5735a3d17a Mon Sep 17 00:00:00 2001 From: Srivatsan Krishnan Date: Fri, 16 Jan 2026 17:44:27 -0800 Subject: [PATCH 4/5] fix unit test --- .../test_megatron_bridge_report_generation_strategy.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tests/report_generation_strategy/test_megatron_bridge_report_generation_strategy.py b/tests/report_generation_strategy/test_megatron_bridge_report_generation_strategy.py index 78c3a3055..13e3be34e 100644 --- a/tests/report_generation_strategy/test_megatron_bridge_report_generation_strategy.py +++ b/tests/report_generation_strategy/test_megatron_bridge_report_generation_strategy.py @@ -38,7 +38,7 @@ def mb_tr(tmp_path: Path) -> TestRun: "", ] ) - (tr.output_path / "megatron_bridge_launcher.log").write_text(log_content) + (tr.output_path / "cloudai_megatron_bridge_launcher.log").write_text(log_content) return tr From 91f7b352a7781fa5f0b147ee9699de21d1d53a37 Mon Sep 17 00:00:00 2001 From: Srivatsan Krishnan Date: Fri, 16 Jan 2026 17:51:12 -0800 Subject: [PATCH 5/5] fix copyright year --- .../workloads/megatron_bridge/report_generation_strategy.py | 2 +- .../test_megatron_bridge_report_generation_strategy.py | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/src/cloudai/workloads/megatron_bridge/report_generation_strategy.py b/src/cloudai/workloads/megatron_bridge/report_generation_strategy.py index 48768fbb1..06610f5e0 100644 --- a/src/cloudai/workloads/megatron_bridge/report_generation_strategy.py +++ b/src/cloudai/workloads/megatron_bridge/report_generation_strategy.py @@ -1,5 +1,5 @@ # SPDX-FileCopyrightText: NVIDIA CORPORATION & AFFILIATES -# Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved. # SPDX-License-Identifier: Apache-2.0 # # Licensed under the Apache License, Version 2.0 (the "License"); diff --git a/tests/report_generation_strategy/test_megatron_bridge_report_generation_strategy.py b/tests/report_generation_strategy/test_megatron_bridge_report_generation_strategy.py index 13e3be34e..a59e3a1fb 100644 --- a/tests/report_generation_strategy/test_megatron_bridge_report_generation_strategy.py +++ b/tests/report_generation_strategy/test_megatron_bridge_report_generation_strategy.py @@ -1,5 +1,5 @@ # SPDX-FileCopyrightText: NVIDIA CORPORATION & AFFILIATES -# Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved. # SPDX-License-Identifier: Apache-2.0 # # Licensed under the Apache License, Version 2.0 (the "License");