升级megatron_v0.10[B

3bec6514 · xingjinliang · 3aca1415 · 3bec6514 · 3aca1415 · 3bec6514
Commit 3bec6514 authored Dec 09, 2024 by xingjinliang
20 changed files
--- a/tests/functional_tests/shell_test_utils/_run_training.sh
+++ b/tests/functional_tests/shell_test_utils/_run_training.sh
--- a/tests/functional_tests/shell_test_utils/jobwait.sh
+++ b/tests/functional_tests/shell_test_utils/jobwait.sh
--- a/tests/functional_tests/shell_test_utils/run_ci_test.sh
+++ b/tests/functional_tests/shell_test_utils/run_ci_test.sh
--- a/tests/functional_tests/test_cases/bert/bert_mr_mcore_tp2_pp2_dgx_a100_1N8G/golden_values_dev.json
+++ b/tests/functional_tests/test_cases/bert/bert_mr_mcore_tp2_pp2_dgx_a100_1N8G/golden_values_dev.json
--- a/tests/functional_tests/test_cases/bert/bert_mr_mcore_tp2_pp2_dgx_a100_1N8G/golden_values_lts.json
+++ b/tests/functional_tests/test_cases/bert/bert_mr_mcore_tp2_pp2_dgx_a100_1N8G/golden_values_lts.json
--- a/tests/functional_tests/test_cases/bert/bert_mr_mcore_tp2_pp2_dgx_a100_1N8G/model_config.yaml
+++ b/tests/functional_tests/test_cases/bert/bert_mr_mcore_tp2_pp2_dgx_a100_1N8G/model_config.yaml
--- a/tests/functional_tests/test_cases/bert/bert_mr_mcore_tp2_pp2_local_spec_dgx_a100_1N8G/golden_values_dev.json
+++ b/tests/functional_tests/test_cases/bert/bert_mr_mcore_tp2_pp2_local_spec_dgx_a100_1N8G/golden_values_dev.json
--- a/tests/functional_tests/test_cases/bert/bert_mr_mcore_tp2_pp2_local_spec_dgx_a100_1N8G/golden_values_lts.json
+++ b/tests/functional_tests/test_cases/bert/bert_mr_mcore_tp2_pp2_local_spec_dgx_a100_1N8G/golden_values_lts.json
--- a/tests/functional_tests/test_cases/bert/bert_mr_mcore_tp2_pp2_local_spec_dgx_a100_1N8G/model_config.yaml
+++ b/tests/functional_tests/test_cases/bert/bert_mr_mcore_tp2_pp2_local_spec_dgx_a100_1N8G/model_config.yaml
--- a/tests/functional_tests/test_cases/bert/bert_mr_mcore_tp2_pp2_resume_torch_dist_dgx_a100_1N8G/model_config.yaml
+++ b/tests/functional_tests/test_cases/bert/bert_mr_mcore_tp2_pp2_resume_torch_dist_dgx_a100_1N8G/model_config.yaml
--- a/tests/functional_tests/test_cases/bert/bert_mr_mcore_tp2_pp2_resume_torch_dist_local_spec_dgx_a100_1N8G/model_config.yaml
+++ b/tests/functional_tests/test_cases/bert/bert_mr_mcore_tp2_pp2_resume_torch_dist_local_spec_dgx_a100_1N8G/model_config.yaml
--- a/tests/functional_tests/test_results/bert/bert_tp1_pp2_1nodes_50steps.json
+++ b/tests/functional_tests/test_results/bert/bert_tp1_pp2_1nodes_50steps.json
--- a/tests/functional_tests/test_results/bert/bert_tp1_pp4_1nodes_50steps.json
+++ b/tests/functional_tests/test_results/bert/bert_tp1_pp4_1nodes_50steps.json
--- a/tests/functional_tests/test_cases/bert/bert_mr_tp1_pp4_vp2_dgx_a100_1N8G/model_config.yaml
+++ b/tests/functional_tests/test_cases/bert/bert_mr_tp1_pp4_vp2_dgx_a100_1N8G/model_config.yaml
--- a/tests/functional_tests/test_cases/bert/bert_mr_tp1_pp4_vp2_resume_torch_dgx_a100_1N8G/model_config.yaml
+++ b/tests/functional_tests/test_cases/bert/bert_mr_tp1_pp4_vp2_resume_torch_dgx_a100_1N8G/model_config.yaml
--- a/tests/functional_tests/test_cases/bert/bert_mr_tp2_pp2_dgx_a100_1N8G/golden_values_dev.json
+++ b/tests/functional_tests/test_cases/bert/bert_mr_tp2_pp2_dgx_a100_1N8G/golden_values_dev.json
--- a/tests/functional_tests/test_cases/bert/bert_mr_tp2_pp2_dgx_a100_1N8G/golden_values_lts.json
+++ b/tests/functional_tests/test_cases/bert/bert_mr_tp2_pp2_dgx_a100_1N8G/golden_values_lts.json
--- a/tests/functional_tests/test_cases/bert/bert_mr_tp2_pp2_dgx_a100_1N8G/model_config.yaml
+++ b/tests/functional_tests/test_cases/bert/bert_mr_tp2_pp2_dgx_a100_1N8G/model_config.yaml
--- a/tests/functional_tests/test_cases/bert/bert_mr_tp2_pp2_resume_torch_dgx_a100_1N8G/model_config.yaml
+++ b/tests/functional_tests/test_cases/bert/bert_mr_tp2_pp2_resume_torch_dgx_a100_1N8G/model_config.yaml
--- a/tests/functional_tests/test_cases/bert/bert_nightly_dgx_a100_1N8G_mcore_tp1_pp2/golden_values_dev.json
+++ b/tests/functional_tests/test_cases/bert/bert_nightly_dgx_a100_1N8G_mcore_tp1_pp2/golden_values_dev.json