Skip to content

[MLPerf 6.1][GRPO] Add Qwen3.5-397B GRPO reference RCP logs for GBS 1152 (6 seeds) #911

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
pavanky merged 2 commits into from
Oct 23, 2026
Merged

[MLPerf 6.1][GRPO] Add Qwen3.5-397B GRPO reference RCP logs for GBS 1152 (6 seeds) #911

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
3f7fc2a
Add qwen35_397b_grpo reference RCP logs for GBS 1152 (6 seeds)
syangx39 Oct 20, 2026
c4f21cd
Address review comments on llm_post_training/README.md
syangx39 Oct 20, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 5 additions & 2 deletions llm_post_training/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -396,6 +396,7 @@ The target-0.69 RCP configurations are:
| 256 | 16 x 16 | `1.0e-6` | `0.125` | 18 | 30 |
| 512 | 32 x 16 | `1.4142135624e-6` | `0.08838834765` | 10 | 20 |
| 1024 | 64 x 16 | `2.0e-6` | `0.0625` | 7 | 10 |
| 1152 | 72 x 16 | `2.1213203436e-6` | `0.05892556510` | 6 | 10 |

The learning-rate and clipping schedules represented by this table are:

Expand All @@ -415,7 +416,7 @@ controls as named command-line arguments with type checks:

| Control | Command-line argument | Qualified value or rule |
|---|---|---|
| Global batch size | `--gbs` | One of 256, 512, or 1024 for the checked-in RCPs |
| Global batch size | `--gbs` | One of 256, 512, 1024, or 1152 for the checked-in RCPs |
| Learning rate and minimum LR | `--lr` | Exact GBS-dependent value in the table above |
| Maximum gradient norm | `--clip` | Exact GBS-dependent value in the table above |
| First validation step | `--val-start` | Exact GBS-dependent value in the table above |
Expand Down Expand Up @@ -748,8 +749,9 @@ The checked-in RCP logs provide six independent seeds at each qualified GBS:
| 256 | `rcp_logs/256/seed_*.out` | 6 / 6 | Step 18 | 0.703125 - 0.75390625 |
| 512 | `rcp_logs/512/seed_*.out` | 6 / 6 | Step 10 | 0.6953125 - 0.73046875 |
| 1024 | `rcp_logs/1024/seed_*.out` | 6 / 6 | Step 7 | 0.70703125 - 0.75000000 |
| 1152 | `rcp_logs/1152/seed_*.out` | 6 / 6 | Step 6 | 0.72908366 - 0.78087652 |

All 18 runs reach the target at the first scheduled validation. This is stable
All 24 runs reach the target at the first scheduled validation. This is stable
observed convergence at the configured evaluation schedule; it does not imply
zero latent crossing-step variance before the first observation.

Expand All @@ -765,6 +767,7 @@ qualified time-to-target run.
| 256 | No | 18 | Every step | No | 30 |
| 512 | No | 10 | Every step | No | 20 |
| 1024 | No | 7 | Every step | No | 10 |
| 1152 | No | 6 | Every step | No | 10 |

The validation dataloader uses the entire 251-task validation JSONL at every
evaluation. The four trajectories for each task use validation temperature
Expand Down
64 changes: 64 additions & 0 deletions llm_post_training/rcp_logs/GBS1152/seed_1.out
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,64 @@
:::MLLOG {"namespace": "", "time_ms": 1792395580842, "event_type": "POINT_IN_TIME", "key": "cache_clear", "value": true, "metadata": {"file": "<string>", "lineno": 3}}
:::MLLOG {"namespace": "", "time_ms": 1792395596065, "event_type": "INTERVAL_START", "key": "init_start", "value": null, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597621, "event_type": "POINT_IN_TIME", "key": "submission_benchmark", "value": "qwen35_397b_grpo", "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 193}}
:::MLLOG {"namespace": "", "time_ms": 1792395597621, "event_type": "POINT_IN_TIME", "key": "submission_org", "value": "Google", "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 193}}
:::MLLOG {"namespace": "", "time_ms": 1792395597621, "event_type": "POINT_IN_TIME", "key": "submission_division", "value": "closed", "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 193}}
:::MLLOG {"namespace": "", "time_ms": 1792395597622, "event_type": "POINT_IN_TIME", "key": "submission_status", "value": "onprem", "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 193}}
:::MLLOG {"namespace": "", "time_ms": 1792395597622, "event_type": "POINT_IN_TIME", "key": "submission_platform", "value": "64xa4x", "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 193}}
:::MLLOG {"namespace": "", "time_ms": 1792395597622, "event_type": "POINT_IN_TIME", "key": "seed", "value": 1885626193, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597623, "event_type": "POINT_IN_TIME", "key": "max_steps", "value": 10, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597623, "event_type": "POINT_IN_TIME", "key": "global_batch_size", "value": 1152, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597623, "event_type": "POINT_IN_TIME", "key": "micro_batch_size", "value": 1, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597623, "event_type": "POINT_IN_TIME", "key": "max_sequence_length", "value": 65536, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597623, "event_type": "POINT_IN_TIME", "key": "train_samples", "value": 11200, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597623, "event_type": "POINT_IN_TIME", "key": "eval_samples", "value": 251, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597623, "event_type": "POINT_IN_TIME", "key": "init_checkpoint_step", "value": 0, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597624, "event_type": "POINT_IN_TIME", "key": "opt_name", "value": "adamw", "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597624, "event_type": "POINT_IN_TIME", "key": "opt_base_learning_rate", "value": 2.121320343559642e-06, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597624, "event_type": "POINT_IN_TIME", "key": "opt_end_learning_rate", "value": 2.121320343559642e-06, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597624, "event_type": "POINT_IN_TIME", "key": "opt_adamw_beta_1", "value": 0.9, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597624, "event_type": "POINT_IN_TIME", "key": "opt_adamw_beta_2", "value": 0.999, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597624, "event_type": "POINT_IN_TIME", "key": "opt_adamw_epsilon", "value": 1e-08, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597624, "event_type": "POINT_IN_TIME", "key": "opt_adamw_weight_decay", "value": 0.0, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597625, "event_type": "POINT_IN_TIME", "key": "opt_gradient_clip_norm", "value": 0.05892556509887896, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597625, "event_type": "POINT_IN_TIME", "key": "opt_learning_rate_warmup_steps", "value": 0, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597625, "event_type": "POINT_IN_TIME", "key": "opt_learning_rate_decay_steps", "value": 100000, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597625, "event_type": "POINT_IN_TIME", "key": "opt_learning_rate_decay_schedule", "value": "constant", "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597625, "event_type": "POINT_IN_TIME", "key": "tensor_parallelism", "value": 4, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597625, "event_type": "POINT_IN_TIME", "key": "pipeline_parallelism", "value": 2, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597625, "event_type": "POINT_IN_TIME", "key": "context_parallelism", "value": 1, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597625, "event_type": "POINT_IN_TIME", "key": "expert_parallelism", "value": 32, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597626, "event_type": "POINT_IN_TIME", "key": "generation_backend", "value": "vllm", "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597626, "event_type": "POINT_IN_TIME", "key": "generation_tensor_parallelism", "value": 8, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597626, "event_type": "POINT_IN_TIME", "key": "generation_pipeline_parallelism", "value": 1, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597626, "event_type": "POINT_IN_TIME", "key": "generation_expert_parallelism", "value": 8, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597626, "event_type": "POINT_IN_TIME", "key": "generation_training_rollout_temperature", "value": 1.0, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597626, "event_type": "POINT_IN_TIME", "key": "generation_training_rollout_top_p", "value": 1.0, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597626, "event_type": "POINT_IN_TIME", "key": "generation_validation_rollout_temperature", "value": 0.1, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597626, "event_type": "POINT_IN_TIME", "key": "generation_validation_rollout_top_p", "value": 0.95, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597627, "event_type": "POINT_IN_TIME", "key": "num_prompts_per_step", "value": 72, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597627, "event_type": "POINT_IN_TIME", "key": "num_generations_per_prompt", "value": 16, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597627, "event_type": "POINT_IN_TIME", "key": "truncated_importance_sampling_ratio_min", "value": 0.999, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597627, "event_type": "POINT_IN_TIME", "key": "truncated_importance_sampling_ratio", "value": 1.002, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597627, "event_type": "POINT_IN_TIME", "key": "truncated_importance_sampling_type", "value": "seq-mask-tis", "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395597627, "event_type": "POINT_IN_TIME", "key": "target_accuracy", "value": 0.69, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395939770, "event_type": "POINT_IN_TIME", "key": "gradient_accumulation_steps", "value": 144, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125}}
:::MLLOG {"namespace": "", "time_ms": 1792395939842, "event_type": "INTERVAL_END", "key": "init_stop", "value": null, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 291}}
:::MLLOG {"namespace": "", "time_ms": 1792395939842, "event_type": "INTERVAL_START", "key": "run_start", "value": null, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 291}}
:::MLLOG {"namespace": "", "time_ms": 1792395939842, "event_type": "INTERVAL_START", "key": "block_start", "value": null, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125, "samples_count": 6912, "step": 0}}
:::MLLOG {"namespace": "", "time_ms": 1792398265750, "event_type": "POINT_IN_TIME", "key": "tracked_stats", "value": {"reduced_train_loss": -0.08669434789044317, "reward": 0.3255208432674408, "grad_norm": 0.04490421712398529, "global_valid_toks": 6374981.0, "global_valid_seqs": 1135.0}, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125, "samples_count": 1152, "step": 1}}
:::MLLOG {"namespace": "", "time_ms": 1792398265761, "event_type": "POINT_IN_TIME", "key": "tracked_stats", "value": {"train_step_time": 1046.0511003360007, "policy_training_time": 441.2504875110026, "exposed_generation_time": 359.42124389399396, "weight_sync_time": 15.67865129099664, "valid_tokens_per_sec_per_gpu": 23.805978047584077}, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125, "samples_count": 1152, "step": 1}}
:::MLLOG {"namespace": "", "time_ms": 1792399300324, "event_type": "POINT_IN_TIME", "key": "tracked_stats", "value": {"reduced_train_loss": -0.07544504791439977, "reward": 0.2795138955116272, "grad_norm": 0.06886813789606094, "global_valid_toks": 6226554.0, "global_valid_seqs": 1133.0}, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125, "samples_count": 2304, "step": 2}}
:::MLLOG {"namespace": "", "time_ms": 1792399300335, "event_type": "POINT_IN_TIME", "key": "tracked_stats", "value": {"train_step_time": 934.6668656110487, "policy_training_time": 283.87288653299765, "exposed_generation_time": 549.8984528569927, "weight_sync_time": 15.811214244000439, "valid_tokens_per_sec_per_gpu": 26.022615604971634}, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125, "samples_count": 2304, "step": 2}}
:::MLLOG {"namespace": "", "time_ms": 1792400389497, "event_type": "POINT_IN_TIME", "key": "tracked_stats", "value": {"reduced_train_loss": -0.08843275631807046, "reward": 0.3272569477558136, "grad_norm": 0.03901178017258644, "global_valid_toks": 6115870.0, "global_valid_seqs": 1109.0}, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125, "samples_count": 3456, "step": 3}}
:::MLLOG {"namespace": "", "time_ms": 1792400389507, "event_type": "POINT_IN_TIME", "key": "tracked_stats", "value": {"train_step_time": 986.5537100890506, "policy_training_time": 286.8524610050008, "exposed_generation_time": 601.8947445570084, "weight_sync_time": 16.254230523998558, "valid_tokens_per_sec_per_gpu": 24.21572889867656}, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125, "samples_count": 3456, "step": 3}}
:::MLLOG {"namespace": "", "time_ms": 1792401489217, "event_type": "POINT_IN_TIME", "key": "tracked_stats", "value": {"reduced_train_loss": -0.05171314728067955, "reward": 0.3671875, "grad_norm": 0.032063666731119156, "global_valid_toks": 6746553.0, "global_valid_seqs": 1117.0}, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125, "samples_count": 4608, "step": 4}}
:::MLLOG {"namespace": "", "time_ms": 1792401489228, "event_type": "POINT_IN_TIME", "key": "tracked_stats", "value": {"train_step_time": 998.8582938328836, "policy_training_time": 287.20581423099793, "exposed_generation_time": 614.0916588618857, "weight_sync_time": 15.83781676000217, "valid_tokens_per_sec_per_gpu": 26.38384525508998}, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125, "samples_count": 4608, "step": 4}}
:::MLLOG {"namespace": "", "time_ms": 1792402672077, "event_type": "POINT_IN_TIME", "key": "tracked_stats", "value": {"reduced_train_loss": -0.043433990846097004, "reward": 0.3880208432674408, "grad_norm": 0.022654572501778603, "global_valid_toks": 6909070.0, "global_valid_seqs": 1139.0}, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125, "samples_count": 5760, "step": 5}}
:::MLLOG {"namespace": "", "time_ms": 1792402672087, "event_type": "POINT_IN_TIME", "key": "tracked_stats", "value": {"train_step_time": 1080.1048807139523, "policy_training_time": 291.4241938240011, "exposed_generation_time": 688.9525599839835, "weight_sync_time": 16.256288522999967, "valid_tokens_per_sec_per_gpu": 24.986975958909184}, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125, "samples_count": 5760, "step": 5}}
:::MLLOG {"namespace": "", "time_ms": 1792403752361, "event_type": "INTERVAL_END", "key": "block_stop", "value": null, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125, "samples_count": 6912, "step": 6}}
:::MLLOG {"namespace": "", "time_ms": 1792403752366, "event_type": "INTERVAL_START", "key": "eval_start", "value": null, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125, "samples_count": 6912, "step": 6}}
:::MLLOG {"namespace": "", "time_ms": 1792404963418, "event_type": "POINT_IN_TIME", "key": "tracked_stats", "value": {"validation_time": 1208.7332950900018}, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125, "step": 6}}
:::MLLOG {"namespace": "", "time_ms": 1792404963427, "event_type": "POINT_IN_TIME", "key": "eval_accuracy", "value": 0.7529880404472351, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125, "samples_count": 6912}}
:::MLLOG {"namespace": "", "time_ms": 1792404963447, "event_type": "INTERVAL_END", "key": "eval_stop", "value": null, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125, "samples_count": 6912, "step": 6}}
:::MLLOG {"namespace": "", "time_ms": 1792404963462, "event_type": "INTERVAL_END", "key": "run_stop", "value": null, "metadata": {"file": "/opt/nemo-rl/nemo_rl/algorithms/mlperf_grpo_logging.py", "lineno": 125, "samples_count": 6912, "status": "success"}}
Loading
Loading