chaiml-ir-v3-3-dpo-qcld-_9704

developer_uid: chai_backend_admin

submission_id: chaiml-ir-v3-3-dpo-qcld-_9704_v1

model_name: training123

model_group: ChaiML/ir-v3-3-dpo_qcld-

status: torndown

timestamp: 2025-11-17T13:31:26+00:00

num_battles: 5625

num_wins: 2885

celo_rating: 1296.58

family_friendly_score: 0.5167999999999999

family_friendly_standard_error: 0.00706707520831638

submission_type: basic

model_repo: ChaiML/ir-v3-3-dpo_qcld-cosine

model_architecture: MistralForCausalLM

model_num_parameters: 24096691200.0

best_of: 8

max_input_tokens: 2048

max_output_tokens: 64

reward_model: default

latencies: [{'batch_size': 1, 'throughput': 0.24799477094865044, 'latency_mean': 4.032217963933944, 'latency_p50': 2.9693864583969116, 'latency_p90': 5.982918047904968}, {'batch_size': 2, 'throughput': 0.5654834525540596, 'latency_mean': 3.5300800120830536, 'latency_p50': 3.521952271461487, 'latency_p90': 3.8171244621276856}, {'batch_size': 3, 'throughput': 0.6930226528339567, 'latency_mean': 4.317442132234573, 'latency_p50': 4.316251873970032, 'latency_p90': 4.750913858413695}, {'batch_size': 4, 'throughput': 0.7857305945039799, 'latency_mean': 5.073708711862564, 'latency_p50': 5.0544809103012085, 'latency_p90': 5.6909730434417725}, {'batch_size': 5, 'throughput': 0.8466587933374995, 'latency_mean': 5.870944427251816, 'latency_p50': 5.836947083473206, 'latency_p90': 6.620693922042847}]

gpu_counts: {'NVIDIA L40S': 1}

display_name: training123

ineligible_reason: num_battles<10000

is_internal_developer: True

language_model: ChaiML/ir-v3-3-dpo_qcld-cosine

model_size: 24B

ranking_group: single

throughput_3p7s: 0.27

us_pacific_date: 2025-11-14

win_ratio: 0.5128888888888888

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': ['You:', '</s>', '\n', 'User:'], 'max_input_tokens': 2048, 'best_of': 8, 'max_output_tokens': 64}

formatter: {'memory_template': "{bot_name}'s Persona: {memory}\n####\n", 'prompt_template': '{prompt}\n<START>\n', 'bot_template': '{bot_name}: {message}\n', 'user_template': '{user_name}: {message}\n', 'response_template': '{bot_name}:', 'truncate_by_message': True}

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 chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer
Waiting for job on chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer to finish
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: ╔═════════════════════════════════════════════════════════════════════╗
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: ║                                                                     ║
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: ║   ██████   ██████  █████   ████  ████                               ║
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: ║  ░░██████ ██████  ░░███   ███░  ░░███                               ║
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: ║   ░███░█████░███   ░███  ███     ░███                               ║
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: ║   ░███░░███ ░███   ░███████      ░███                               ║
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: ║   ░███ ░░░  ░███   ░███░░███     ░███                               ║
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: ║   ░███      ░███   ░███ ░░███    ░███                               ║
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: ║   █████     █████  █████ ░░████  █████                              ║
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: ║  ░░░░░     ░░░░░  ░░░░░   ░░░░  ░░░░░                               ║
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: ║                                                                     ║
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: ║  Version: 0.30.2                                                    ║
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: ║  Features: FLYWHEEL, CUDA                                           ║
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: ║  Copyright 2023-2025 MK ONE TECHNOLOGIES Inc.                       ║
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: ║  https://mk1.ai                                                     ║
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: ║                                                                     ║
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: ║  The license key for the current software has been verified as      ║
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: ║  belonging to:                                                      ║
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: ║                                                                     ║
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: ║  Chai Research Corp.                                                ║
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: ║  Account ID: 7997a29f-0ceb-4cc7-9adf-840c57b4ae6f                   ║
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: ║  Expiration: 2028-03-31 23:59:59                                    ║
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: ║                                                                     ║
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: ╚═════════════════════════════════════════════════════════════════════╝
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: Downloaded to shared memory in 59.491s
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: Checking if ChaiML/ir-v3-3-dpo_qcld-cosine already exists in ChaiML
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: quantizing model to /dev/shm/model_cache, profile:s0, folder:/tmp/tmp3g9yu4yr, device:0
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: Saving flywheel model at /dev/shm/model_cache
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: quantized model in 42.111s
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: Processed model ChaiML/ir-v3-3-dpo_qcld-cosine in 101.603s
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: creating bucket guanaco-mkml-models
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: Bucket 's3://guanaco-mkml-models/' created
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: uploading /dev/shm/model_cache to s3://guanaco-mkml-models/chaiml-ir-v3-3-dpo-qcld-9704-v1/nvidia
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: cp /dev/shm/model_cache/special_tokens_map.json s3://guanaco-mkml-models/chaiml-ir-v3-3-dpo-qcld-9704-v1/nvidia/special_tokens_map.json
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: cp /dev/shm/model_cache/config.json s3://guanaco-mkml-models/chaiml-ir-v3-3-dpo-qcld-9704-v1/nvidia/config.json
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: cp /dev/shm/model_cache/tokenizer_config.json s3://guanaco-mkml-models/chaiml-ir-v3-3-dpo-qcld-9704-v1/nvidia/tokenizer_config.json
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: cp /dev/shm/model_cache/tokenizer.json s3://guanaco-mkml-models/chaiml-ir-v3-3-dpo-qcld-9704-v1/nvidia/tokenizer.json
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: cp /dev/shm/model_cache/flywheel_model.0.safetensors s3://guanaco-mkml-models/chaiml-ir-v3-3-dpo-qcld-9704-v1/nvidia/flywheel_model.0.safetensors
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: cp /dev/shm/model_cache/flywheel_model.1.safetensors s3://guanaco-mkml-models/chaiml-ir-v3-3-dpo-qcld-9704-v1/nvidia/flywheel_model.1.safetensors
chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer: 
Loading 0:   0%|          | 0/363 [00:00<?, ?it/s]
Loading 0:   1%|▏         | 5/363 [00:00<00:16, 21.77it/s]
Loading 0:   3%|▎         | 12/363 [00:00<00:08, 39.80it/s]
Loading 0:   5%|▍         | 17/363 [00:00<00:09, 35.51it/s]
Loading 0:   6%|▌         | 22/363 [00:00<00:09, 36.30it/s]
Loading 0:   7%|▋         | 26/363 [00:00<00:09, 37.30it/s]
Loading 0:   9%|▉         | 32/363 [00:00<00:07, 42.68it/s]
Loading 0:  10%|█         | 37/363 [00:01<00:12, 26.44it/s]
Loading 0:  11%|█▏        | 41/363 [00:01<00:12, 24.82it/s]
Loading 0:  13%|█▎        | 48/363 [00:01<00:09, 32.71it/s]
Loading 0:  15%|█▍        | 53/363 [00:01<00:09, 32.53it/s]
Loading 0:  16%|█▌        | 58/363 [00:01<00:09, 33.27it/s]
Loading 0:  17%|█▋        | 62/363 [00:01<00:09, 33.34it/s]
Loading 0:  18%|█▊        | 67/363 [00:01<00:07, 37.09it/s]
Loading 0:  20%|█▉        | 72/363 [00:02<00:10, 28.67it/s]
Loading 0:  21%|██        | 76/363 [00:02<00:11, 24.42it/s]
Loading 0:  22%|██▏       | 79/363 [00:02<00:14, 20.16it/s]
Loading 0:  23%|██▎       | 82/363 [00:03<00:16, 16.99it/s]
Loading 0:  23%|██▎       | 85/363 [00:03<00:16, 17.34it/s]
Loading 0:  24%|██▍       | 87/363 [00:03<00:19, 13.91it/s]
Loading 0:  25%|██▌       | 91/363 [00:03<00:15, 17.76it/s]
Loading 0:  26%|██▌       | 94/363 [00:03<00:13, 19.31it/s]
Loading 0:  27%|██▋       | 99/363 [00:03<00:10, 24.93it/s]
Loading 0:  28%|██▊       | 102/363 [00:03<00:10, 24.34it/s]
Loading 0:  29%|██▉       | 107/363 [00:04<00:10, 25.13it/s]
Loading 0:  31%|███       | 112/363 [00:04<00:08, 28.51it/s]
Loading 0:  32%|███▏      | 116/363 [00:04<00:08, 29.76it/s]
Loading 0:  33%|███▎      | 120/363 [00:04<00:07, 32.05it/s]
Loading 0:  34%|███▍      | 124/363 [00:04<00:07, 31.00it/s]
Loading 0:  36%|███▌      | 129/363 [00:04<00:06, 34.73it/s]
Loading 0:  37%|███▋      | 133/363 [00:04<00:07, 32.56it/s]
Loading 0:  38%|███▊      | 138/363 [00:04<00:06, 35.84it/s]
Loading 0:  39%|███▉      | 142/363 [00:05<00:06, 33.72it/s]
Loading 0:  41%|████      | 149/363 [00:05<00:05, 40.39it/s]
Loading 0:  42%|████▏     | 154/363 [00:05<00:07, 27.86it/s]
Loading 0:  44%|████▎     | 158/363 [00:05<00:07, 25.89it/s]
Loading 0:  45%|████▌     | 165/363 [00:05<00:05, 33.18it/s]
Loading 0:  47%|████▋     | 169/363 [00:05<00:06, 31.72it/s]
Loading 0:  48%|████▊     | 174/363 [00:06<00:05, 34.95it/s]
Loading 0:  49%|████▉     | 178/363 [00:06<00:05, 32.94it/s]
Loading 0:  50%|█████     | 182/363 [00:06<00:05, 33.80it/s]
Loading 0:  52%|█████▏    | 187/363 [00:06<00:05, 31.65it/s]
Loading 0:  53%|█████▎    | 191/363 [00:06<00:05, 30.29it/s]
Loading 0:  54%|█████▎    | 195/363 [00:06<00:06, 27.16it/s]
Loading 0:  55%|█████▌    | 201/363 [00:19<02:13,  1.21it/s]
Loading 0:  56%|█████▌    | 203/363 [00:19<01:54,  1.40it/s]
Loading 0:  57%|█████▋    | 208/363 [00:20<01:13,  2.12it/s]
Loading 0:  58%|█████▊    | 212/363 [00:20<00:53,  2.82it/s]
Loading 0:  60%|██████    | 218/363 [00:20<00:32,  4.40it/s]
Loading 0:  61%|██████    | 222/363 [00:20<00:24,  5.75it/s]
Loading 0:  62%|██████▏   | 226/363 [00:20<00:19,  7.13it/s]
Loading 0:  63%|██████▎   | 230/363 [00:20<00:15,  8.78it/s]
Loading 0:  65%|██████▌   | 237/363 [00:21<00:09, 13.58it/s]
Loading 0:  66%|██████▋   | 241/363 [00:21<00:07, 15.55it/s]
Loading 0:  68%|██████▊   | 246/363 [00:21<00:05, 19.53it/s]
Loading 0:  69%|██████▉   | 250/363 [00:21<00:05, 21.13it/s]
Loading 0:  70%|███████   | 255/363 [00:21<00:04, 25.38it/s]
Loading 0:  71%|███████▏  | 259/363 [00:21<00:04, 25.90it/s]
Loading 0:  73%|███████▎  | 266/363 [00:21<00:02, 32.98it/s]
Loading 0:  75%|███████▍  | 271/363 [00:22<00:03, 24.18it/s]
Loading 0:  76%|███████▌  | 275/363 [00:22<00:03, 23.24it/s]
Loading 0:  78%|███████▊  | 282/363 [00:22<00:02, 29.84it/s]
Loading 0:  79%|███████▉  | 286/363 [00:22<00:02, 29.73it/s]
Loading 0:  80%|████████  | 291/363 [00:22<00:02, 32.86it/s]
Loading 0:  81%|████████▏ | 295/363 [00:22<00:02, 31.44it/s]
Loading 0:  82%|████████▏ | 299/363 [00:22<00:01, 32.60it/s]
Loading 0:  84%|████████▎ | 304/363 [00:23<00:01, 29.99it/s]
Loading 0:  85%|████████▍ | 308/363 [00:23<00:01, 29.28it/s]
Loading 0:  86%|████████▌ | 312/363 [00:23<00:01, 26.46it/s]
Loading 0:  88%|████████▊ | 318/363 [00:23<00:01, 32.21it/s]
Loading 0:  89%|████████▊ | 322/363 [00:23<00:01, 31.46it/s]
Loading 0:  90%|█████████ | 327/363 [00:23<00:01, 34.73it/s]
Loading 0:  91%|█████████ | 331/363 [00:23<00:00, 32.70it/s]
Loading 0:  92%|█████████▏| 335/363 [00:24<00:00, 33.46it/s]
Loading 0:  93%|█████████▎| 339/363 [00:24<00:00, 34.08it/s]
Loading 0:  94%|█████████▍| 343/363 [00:24<00:00, 20.42it/s]
Loading 0:  96%|█████████▌| 348/363 [00:24<00:00, 22.02it/s]
Loading 0:  98%|█████████▊| 355/363 [00:24<00:00, 29.69it/s]
Loading 0:  99%|█████████▉| 359/363 [00:25<00:00, 29.90it/s]
                                                            
Job chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer completed after 156.4s with status: succeeded
Stopping job with name chaiml-ir-v3-3-dpo-qcld-9704-v1-mkmlizer
Pipeline stage MKMLizer completed in 156.94s
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 chaiml-ir-v3-3-dpo-qcld-9704-v1
Waiting for inference service chaiml-ir-v3-3-dpo-qcld-9704-v1 to be ready
Inference service chaiml-ir-v3-3-dpo-qcld-9704-v1 ready after 151.01715445518494s
Pipeline stage MKMLDeployer completed in 152.25s
run pipeline stage %s
Running pipeline stage StressChecker
Received healthy response to inference request in 2.668973922729492s
Received healthy response to inference request in 2.452151298522949s
Received healthy response to inference request in 2.841921806335449s
Received healthy response to inference request in 2.235736846923828s
Received healthy response to inference request in 2.225076913833618s
5 requests
0 failed requests
5th percentile: 2.22720890045166
10th percentile: 2.2293408870697022
20th percentile: 2.2336048603057863
30th percentile: 2.2790197372436523
40th percentile: 2.365585517883301
50th percentile: 2.452151298522949
60th percentile: 2.5388803482055664
70th percentile: 2.6256093978881836
80th percentile: 2.7035634994506834
90th percentile: 2.7727426528930663
95th percentile: 2.807332229614258
99th percentile: 2.835003890991211
mean time: 2.4847721576690676
Pipeline stage StressChecker completed in 13.84s
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 0.68s
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 0.69s
Shutdown handler de-registered
chaiml-ir-v3-3-dpo-qcld-_9704_v1 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 0.08s
run pipeline stage %s
Running pipeline stage MKMLProfilerDeployer
Creating inference service chaiml-ir-v3-3-dpo-qcld-9704-v1-profiler
Waiting for inference service chaiml-ir-v3-3-dpo-qcld-9704-v1-profiler to be ready
Inference service chaiml-ir-v3-3-dpo-qcld-9704-v1-profiler ready after 141.19267225265503s
Pipeline stage MKMLProfilerDeployer completed in 141.80s
run pipeline stage %s
Running pipeline stage MKMLProfilerRunner
kubectl --kubeconfig /code/guanaco/guanaco_services/resources/kchai_coreweave_us_east_04a.yaml cp /code/guanaco/guanaco_inference_services/src/inference_scripts tenant-chaiml-guanaco/chaiml-ir-v3-3-dpo-qe03127c2c1124e51a610b83c7e59a4f8-deplogpt58:/code/chaiverse_profiler_1763122692 --namespace tenant-chaiml-guanaco
kubectl --kubeconfig /code/guanaco/guanaco_services/resources/kchai_coreweave_us_east_04a.yaml exec -it chaiml-ir-v3-3-dpo-qe03127c2c1124e51a610b83c7e59a4f8-deplogpt58 --namespace tenant-chaiml-guanaco -- sh -c 'cd /code/chaiverse_profiler_1763122692 && 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 2048 --output_tokens 64 --summary /code/chaiverse_profiler_1763122692/summary.json'
%s, retrying in %s seconds...
kubectl --kubeconfig /code/guanaco/guanaco_services/resources/kchai_coreweave_us_east_04a.yaml cp /code/guanaco/guanaco_inference_services/src/inference_scripts tenant-chaiml-guanaco/chaiml-ir-v3-3-dpo-qe03127c2c1124e51a610b83c7e59a4f8-deplogpt58:/code/chaiverse_profiler_1763123888 --namespace tenant-chaiml-guanaco
kubectl --kubeconfig /code/guanaco/guanaco_services/resources/kchai_coreweave_us_east_04a.yaml exec -it chaiml-ir-v3-3-dpo-qe03127c2c1124e51a610b83c7e59a4f8-deplogpt58 --namespace tenant-chaiml-guanaco -- sh -c 'cd /code/chaiverse_profiler_1763123888 && 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 2048 --output_tokens 64 --summary /code/chaiverse_profiler_1763123888/summary.json'
kubectl --kubeconfig /code/guanaco/guanaco_services/resources/kchai_coreweave_us_east_04a.yaml exec -it chaiml-ir-v3-3-dpo-qe03127c2c1124e51a610b83c7e59a4f8-deplogpt58 --namespace tenant-chaiml-guanaco -- bash -c 'cat /code/chaiverse_profiler_1763123888/summary.json'
Pipeline stage MKMLProfilerRunner completed in 3143.86s
run pipeline stage %s
Running pipeline stage MKMLProfilerDeleter
Checking if service chaiml-ir-v3-3-dpo-qcld-9704-v1-profiler is running
Tearing down inference service chaiml-ir-v3-3-dpo-qcld-9704-v1-profiler
Service chaiml-ir-v3-3-dpo-qcld-9704-v1-profiler has been torndown
Pipeline stage MKMLProfilerDeleter completed in 0.71s
Shutdown handler de-registered
chaiml-ir-v3-3-dpo-qcld-_9704_v1 status is now inactive due to auto deactivation removed underperforming models
chaiml-ir-v3-3-dpo-qcld-_9704_v1 status is now torndown due to DeploymentManager action