rirv938-anthropic-grpo-4_3092

developer_uid: rirv938

submission_id: rirv938-anthropic-grpo-4_3092_v4

model_name: rirv938-anthropic-grpo-4_3092_v4

model_group: rirv938/anthropic_grpo_4

status: torndown

timestamp: 2025-05-13T17:20:28+00:00

num_battles: 7456

num_wins: 3875

celo_rating: 1303.71

family_friendly_score: 0.5426

family_friendly_standard_error: 0.007045356484948083

submission_type: basic

model_repo: rirv938/anthropic_grpo_40k_cp888_95ff_merged

model_architecture: MistralForCausalLM

model_num_parameters: 12772070400.0

best_of: 8

max_input_tokens: 768

max_output_tokens: 64

reward_model: default

latencies: [{'batch_size': 1, 'throughput': 0.6518006688042226, 'latency_mean': 1.5341128873825074, 'latency_p50': 1.5247615575790405, 'latency_p90': 1.7153270244598389}, {'batch_size': 3, 'throughput': 1.2403185060839008, 'latency_mean': 2.4088317692279815, 'latency_p50': 2.411019802093506, 'latency_p90': 2.651649260520935}, {'batch_size': 5, 'throughput': 1.5608071113687543, 'latency_mean': 3.184964956045151, 'latency_p50': 3.1780078411102295, 'latency_p90': 3.624485397338867}, {'batch_size': 6, 'throughput': 1.646018322541121, 'latency_mean': 3.617769619226456, 'latency_p50': 3.619633436203003, 'latency_p90': 4.102118372917175}, {'batch_size': 8, 'throughput': 1.7470997013590353, 'latency_mean': 4.538427077531814, 'latency_p50': 4.545286417007446, 'latency_p90': 5.142684912681579}, {'batch_size': 10, 'throughput': 1.7926658652968228, 'latency_mean': 5.519537209272385, 'latency_p50': 5.514781713485718, 'latency_p90': 6.347585415840149}]

gpu_counts: {'NVIDIA RTX A5000': 1}

display_name: rirv938-anthropic-grpo-4_3092_v4

is_internal_developer: True

language_model: rirv938/anthropic_grpo_40k_cp888_95ff_merged

model_size: 13B

ranking_group: single

throughput_3p7s: 1.67

us_pacific_date: 2025-05-13

win_ratio: 0.5197156652360515

generation_params: {'temperature': 1.0, 'top_p': 1.0, 'min_p': 0.0, 'top_k': 80, 'presence_penalty': 0.0, 'frequency_penalty': 0.0, 'stopping_words': ['</s>', '###', 'You:', '\n'], 'max_input_tokens': 768, 'best_of': 8, 'max_output_tokens': 64}

formatter: {'memory_template': '', 'prompt_template': '', 'bot_template': '{bot_name}: {message}\n', 'user_template': '{user_name}: {message}\n', 'response_template': '{bot_name}:', 'truncate_by_message': False}

Resubmit model

Shutdown handler not registered because Python interpreter is not running in the main thread
run pipeline %s
run pipeline stage %s
Running pipeline stage MKMLizer
Starting job with name rirv938-anthropic-grpo-4-3092-v4-mkmlizer
Waiting for job on rirv938-anthropic-grpo-4-3092-v4-mkmlizer to finish
rirv938-anthropic-grpo-4-3092-v4-mkmlizer: Downloaded to shared memory in 89.321s
rirv938-anthropic-grpo-4-3092-v4-mkmlizer: quantizing model to /dev/shm/model_cache, profile:s0, folder:/tmp/tmp88oir02r, device:0
rirv938-anthropic-grpo-4-3092-v4-mkmlizer: Saving flywheel model at /dev/shm/model_cache
rirv938-anthropic-grpo-4-3092-v4-mkmlizer: quantized model in 42.238s
rirv938-anthropic-grpo-4-3092-v4-mkmlizer: Processed model rirv938/anthropic_grpo_40k_cp888_95ff_merged in 131.560s
rirv938-anthropic-grpo-4-3092-v4-mkmlizer: creating bucket guanaco-mkml-models
rirv938-anthropic-grpo-4-3092-v4-mkmlizer: Bucket 's3://guanaco-mkml-models/' created
rirv938-anthropic-grpo-4-3092-v4-mkmlizer: uploading /dev/shm/model_cache to s3://guanaco-mkml-models/rirv938-anthropic-grpo-4-3092-v4
rirv938-anthropic-grpo-4-3092-v4-mkmlizer: cp /dev/shm/model_cache/special_tokens_map.json s3://guanaco-mkml-models/rirv938-anthropic-grpo-4-3092-v4/special_tokens_map.json
rirv938-anthropic-grpo-4-3092-v4-mkmlizer: cp /dev/shm/model_cache/config.json s3://guanaco-mkml-models/rirv938-anthropic-grpo-4-3092-v4/config.json
rirv938-anthropic-grpo-4-3092-v4-mkmlizer: cp /dev/shm/model_cache/tokenizer_config.json s3://guanaco-mkml-models/rirv938-anthropic-grpo-4-3092-v4/tokenizer_config.json
rirv938-anthropic-grpo-4-3092-v4-mkmlizer: cp /dev/shm/model_cache/tokenizer.json s3://guanaco-mkml-models/rirv938-anthropic-grpo-4-3092-v4/tokenizer.json
rirv938-anthropic-grpo-4-3092-v4-mkmlizer: cp /dev/shm/model_cache/flywheel_model.0.safetensors s3://guanaco-mkml-models/rirv938-anthropic-grpo-4-3092-v4/flywheel_model.0.safetensors
rirv938-anthropic-grpo-4-3092-v4-mkmlizer: 
Loading 0:   0%|          | 0/363 [00:00<?, ?it/s]
Loading 0:   1%|▏         | 5/363 [00:00<00:17, 21.02it/s]
Loading 0:   3%|▎         | 10/363 [00:00<00:13, 26.83it/s]
Loading 0:   4%|▍         | 14/363 [00:00<00:14, 23.72it/s]
Loading 0:   5%|▌         | 19/363 [00:00<00:11, 30.16it/s]
Loading 0:   6%|▋         | 23/363 [00:00<00:16, 21.12it/s]
Loading 0:   7%|▋         | 26/363 [00:01<00:17, 19.07it/s]
Loading 0:   9%|▊         | 31/363 [00:01<00:13, 24.84it/s]
Loading 0:  10%|▉         | 35/363 [00:01<00:12, 26.22it/s]
Loading 0:  11%|█         | 39/363 [00:01<00:12, 26.56it/s]
Loading 0:  12%|█▏        | 42/363 [00:01<00:13, 23.99it/s]
Loading 0:  13%|█▎        | 46/363 [00:01<00:11, 27.05it/s]
Loading 0:  14%|█▍        | 50/363 [00:02<00:13, 23.93it/s]
Loading 0:  15%|█▌        | 55/363 [00:02<00:10, 28.10it/s]
Loading 0:  17%|█▋        | 60/363 [00:02<00:10, 29.35it/s]
Loading 0:  18%|█▊        | 64/363 [00:02<00:15, 19.10it/s]
Loading 0:  19%|█▉        | 69/363 [00:02<00:12, 23.42it/s]
Loading 0:  20%|██        | 73/363 [00:03<00:12, 23.35it/s]
Loading 0:  21%|██        | 77/363 [00:03<00:12, 22.05it/s]
Loading 0:  23%|██▎       | 82/363 [00:03<00:10, 26.19it/s]
Loading 0:  24%|██▎       | 86/363 [00:03<00:11, 23.58it/s]
Loading 0:  25%|██▌       | 91/363 [00:03<00:10, 27.01it/s]
Loading 0:  26%|██▌       | 95/363 [00:03<00:11, 23.92it/s]
Loading 0:  28%|██▊       | 100/363 [00:04<00:09, 28.26it/s]
Loading 0:  29%|██▊       | 104/363 [00:04<00:13, 18.75it/s]
Loading 0:  30%|███       | 109/363 [00:04<00:10, 23.19it/s]
Loading 0:  31%|███       | 113/363 [00:04<00:11, 22.07it/s]
Loading 0:  33%|███▎      | 118/363 [00:04<00:09, 26.62it/s]
Loading 0:  34%|███▎      | 122/363 [00:05<00:10, 23.75it/s]
Loading 0:  35%|███▍      | 127/363 [00:05<00:08, 27.63it/s]
Loading 0:  36%|███▌      | 131/363 [00:05<00:09, 24.20it/s]
Loading 0:  37%|███▋      | 136/363 [00:05<00:07, 28.73it/s]
Loading 0:  39%|███▉      | 141/363 [00:05<00:07, 30.07it/s]
Loading 0:  40%|███▉      | 145/363 [00:06<00:10, 21.37it/s]
Loading 0:  41%|████      | 149/363 [00:06<00:10, 20.66it/s]
Loading 0:  42%|████▏     | 154/363 [00:06<00:08, 25.06it/s]
Loading 0:  44%|████▎     | 158/363 [00:06<00:08, 23.16it/s]
Loading 0:  45%|████▍     | 163/363 [00:06<00:07, 27.84it/s]
Loading 0:  46%|████▌     | 167/363 [00:06<00:07, 25.02it/s]
Loading 0:  47%|████▋     | 172/363 [00:06<00:06, 29.60it/s]
Loading 0:  48%|████▊     | 176/363 [00:07<00:07, 26.24it/s]
Loading 0:  50%|████▉     | 181/363 [00:07<00:06, 29.97it/s]
Loading 0:  51%|█████     | 185/363 [00:07<00:08, 20.27it/s]
Loading 0:  53%|█████▎    | 192/363 [00:07<00:06, 26.47it/s]
Loading 0:  54%|█████▍    | 196/363 [00:07<00:06, 26.28it/s]
Loading 0:  55%|█████▌    | 201/363 [00:08<00:05, 27.76it/s]
Loading 0:  56%|█████▋    | 205/363 [00:08<00:05, 27.11it/s]
Loading 0:  58%|█████▊    | 210/363 [00:08<00:05, 29.49it/s]
Loading 0:  59%|█████▉    | 214/363 [00:08<00:05, 28.41it/s]
Loading 0:  60%|██████    | 218/363 [00:08<00:05, 28.86it/s]
Loading 0:  61%|██████▏   | 223/363 [00:08<00:05, 25.42it/s]
Loading 0:  62%|██████▏   | 226/363 [00:09<00:05, 24.25it/s]
Loading 0:  63%|██████▎   | 230/363 [00:09<00:05, 23.30it/s]
Loading 0:  65%|██████▌   | 237/363 [00:09<00:04, 29.87it/s]
Loading 0:  66%|██████▋   | 241/363 [00:09<00:04, 28.87it/s]
Loading 0:  68%|██████▊   | 246/363 [00:09<00:03, 30.91it/s]
Loading 0:  69%|██████▉   | 250/363 [00:09<00:03, 29.46it/s]
Loading 0:  70%|███████   | 255/363 [00:09<00:03, 31.64it/s]
Loading 0:  71%|███████▏  | 259/363 [00:10<00:03, 30.31it/s]
Loading 0:  72%|███████▏  | 263/363 [00:10<00:04, 24.90it/s]
Loading 0:  73%|███████▎  | 266/363 [00:10<00:04, 22.13it/s]
Loading 0:  75%|███████▌  | 273/363 [00:10<00:03, 28.50it/s]
Loading 0:  76%|███████▋  | 277/363 [00:10<00:03, 27.96it/s]
Loading 0:  78%|███████▊  | 282/363 [00:10<00:02, 30.26it/s]
Loading 0:  79%|███████▉  | 286/363 [00:11<00:02, 28.41it/s]
Loading 0:  80%|████████  | 291/363 [00:11<00:02, 29.85it/s]
Loading 0:  81%|████████▏ | 295/363 [00:11<00:02, 28.58it/s]
Loading 0:  82%|████████▏ | 299/363 [00:11<00:02, 28.36it/s]
Loading 0:  84%|████████▎ | 304/363 [00:11<00:02, 23.70it/s]
Loading 0:  85%|████████▍ | 307/363 [00:12<00:02, 22.42it/s]
Loading 0:  86%|████████▌ | 311/363 [00:12<00:02, 21.34it/s]
Loading 0:  87%|████████▋ | 316/363 [00:12<00:01, 26.36it/s]
Loading 0:  88%|████████▊ | 320/363 [00:12<00:01, 23.98it/s]
Loading 0:  90%|█████████ | 327/363 [00:12<00:01, 30.38it/s]
Loading 0:  91%|█████████ | 331/363 [00:12<00:01, 29.59it/s]
Loading 0:  93%|█████████▎| 336/363 [00:12<00:00, 31.44it/s]
Loading 0:  94%|█████████▎| 340/363 [00:13<00:00, 29.90it/s]
Loading 0:  95%|█████████▍| 344/363 [00:20<00:09,  2.01it/s]
Loading 0:  96%|█████████▌| 348/363 [00:20<00:05,  2.69it/s]
Loading 0:  97%|█████████▋| 353/363 [00:20<00:02,  3.89it/s]
Loading 0:  98%|█████████▊| 357/363 [00:20<00:01,  5.02it/s]
                                                            
Job rirv938-anthropic-grpo-4-3092-v4-mkmlizer completed after 166.8s with status: succeeded
Stopping job with name rirv938-anthropic-grpo-4-3092-v4-mkmlizer
Pipeline stage MKMLizer completed in 167.33s
run pipeline stage %s
Running pipeline stage MKMLTemplater
Pipeline stage MKMLTemplater completed in 0.17s
run pipeline stage %s
Running pipeline stage MKMLDeployer
Creating inference service rirv938-anthropic-grpo-4-3092-v4
Waiting for inference service rirv938-anthropic-grpo-4-3092-v4 to be ready
Inference service rirv938-anthropic-grpo-4-3092-v4 ready after 120.79951739311218s
Pipeline stage MKMLDeployer completed in 121.40s
run pipeline stage %s
Running pipeline stage StressChecker
Received healthy response to inference request in 2.353396415710449s
Received healthy response to inference request in 1.7207131385803223s
Received healthy response to inference request in 2.0857553482055664s
Received healthy response to inference request in 1.8465533256530762s
Received healthy response to inference request in 1.744729995727539s
5 requests
0 failed requests
5th percentile: 1.7255165100097656
10th percentile: 1.730319881439209
20th percentile: 1.7399266242980957
30th percentile: 1.7650946617126464
40th percentile: 1.8058239936828613
50th percentile: 1.8465533256530762
60th percentile: 1.9422341346740724
70th percentile: 2.0379149436950685
80th percentile: 2.139283561706543
90th percentile: 2.2463399887084963
95th percentile: 2.2998682022094727
99th percentile: 2.3426907730102537
mean time: 1.9502296447753906
Pipeline stage StressChecker completed in 11.02s
run pipeline stage %s
Running pipeline stage OfflineFamilyFriendlyTriggerPipeline
run_pipeline:run_in_cloud %s
starting trigger_guanaco_pipeline args=%s
triggered trigger_guanaco_pipeline args=%s
Pipeline stage OfflineFamilyFriendlyTriggerPipeline completed in 0.69s
run pipeline stage %s
Running pipeline stage TriggerMKMLProfilingPipeline
run_pipeline:run_in_cloud %s
starting trigger_guanaco_pipeline args=%s
triggered trigger_guanaco_pipeline args=%s
Pipeline stage TriggerMKMLProfilingPipeline completed in 0.89s
Shutdown handler de-registered
rirv938-anthropic-grpo-4_3092_v4 status is now deployed due to DeploymentManager action
Shutdown handler registered
run pipeline %s
run pipeline stage %s
Running pipeline stage MKMLProfilerDeleter
Skipping teardown as no inference service was successfully deployed
Pipeline stage MKMLProfilerDeleter completed in 0.09s
run pipeline stage %s
Running pipeline stage MKMLProfilerTemplater
Pipeline stage MKMLProfilerTemplater completed in 0.09s
run pipeline stage %s
Running pipeline stage MKMLProfilerDeployer
Creating inference service rirv938-anthropic-grpo-4-3092-v4-profiler
Waiting for inference service rirv938-anthropic-grpo-4-3092-v4-profiler to be ready
Inference service rirv938-anthropic-grpo-4-3092-v4-profiler ready after 121.55186128616333s
Pipeline stage MKMLProfilerDeployer completed in 122.19s
run pipeline stage %s
Running pipeline stage MKMLProfilerRunner
kubectl cp /code/guanaco/guanaco_inference_services/src/inference_scripts tenant-chaiml-guanaco/rirv938-anthropic-gr35bf7c663c6f9d4b8f511d6ebb310571-deplocpbps:/code/chaiverse_profiler_1747157543 --namespace tenant-chaiml-guanaco
kubectl exec -it rirv938-anthropic-gr35bf7c663c6f9d4b8f511d6ebb310571-deplocpbps --namespace tenant-chaiml-guanaco -- sh -c 'cd /code/chaiverse_profiler_1747157543 && python profiles.py profile --best_of_n 8 --auto_batch 5 --batches 1,5,10,15,20,25,30,35,40,45,50,55,60,65,70,75,80,85,90,95,100,105,110,115,120,125,130,135,140,145,150,155,160,165,170,175,180,185,190,195 --samples 200 --input_tokens 768 --output_tokens 64 --summary /code/chaiverse_profiler_1747157543/summary.json'
Shutdown handler registered
run pipeline %s
run pipeline stage %s
Running pipeline stage OfflineFamilyFriendlyScorer
Evaluating %s Family Friendly Score with %s threads
Received signal 15, running shutdown handler
Shutdown handler de-registered
Shutdown handler registered
run pipeline %s
run pipeline stage %s
Running pipeline stage OfflineFamilyFriendlyScorer
Evaluating %s Family Friendly Score with %s threads
Pipeline stage OfflineFamilyFriendlyScorer completed in 2542.82s
Shutdown handler de-registered
rirv938-anthropic-grpo-4_3092_v4 status is now inactive due to auto deactivation removed underperforming models
rirv938-anthropic-grpo-4_3092_v4 status is now torndown due to DeploymentManager action