jic062-dpo-v1-nemo

submission_id: jic062-dpo-v1-nemo_v1

developer_uid: chace9580

alignment_samples: 12690

alignment_score: -0.002036499677690831

best_of: 8

celo_rating: 1233.74

display_name: jic062-dpo-v1-nemo_v1

formatter: {'memory_template': '[INST]system\n{memory}[/INST]\n', 'prompt_template': '[INST]user\n{prompt}[/INST]\n', 'bot_template': '[INST]assistant\n{bot_name}: {message}[/INST]\n', 'user_template': '[INST]user\n{user_name}: {message}[/INST]\n', 'response_template': '[INST]assistant\n{bot_name}:', 'truncate_by_message': False}

generation_params: {'temperature': 0.75, 'top_p': 1.0, 'min_p': 0.1, 'top_k': 40, 'presence_penalty': 0.0, 'frequency_penalty': 0.0, 'stopping_words': ['\n', '\n\n', '\nYou:', '[/INST]', '<|im_end|>', '</s>'], 'max_input_tokens': 512, 'best_of': 8, 'max_output_tokens': 64}

gpu_counts: {'NVIDIA RTX A5000': 1}

is_internal_developer: False

language_model: jic062/dpo-v1-Nemo

latencies: [{'batch_size': 1, 'throughput': 0.6915066666196024, 'latency_mean': 1.4460530912876128, 'latency_p50': 1.4527934789657593, 'latency_p90': 1.619873547554016}, {'batch_size': 3, 'throughput': 1.319402021955325, 'latency_mean': 2.2671183025836945, 'latency_p50': 2.2687665224075317, 'latency_p90': 2.5185172080993654}, {'batch_size': 5, 'throughput': 1.5457996024291834, 'latency_mean': 3.2163057589530943, 'latency_p50': 3.2157175540924072, 'latency_p90': 3.5777307987213134}, {'batch_size': 6, 'throughput': 1.6062908913627592, 'latency_mean': 3.709387127161026, 'latency_p50': 3.7431907653808594, 'latency_p90': 4.167865204811096}, {'batch_size': 8, 'throughput': 1.6010188704436157, 'latency_mean': 4.966292465925217, 'latency_p50': 5.0156779289245605, 'latency_p90': 5.597649788856506}, {'batch_size': 10, 'throughput': 1.538713987563022, 'latency_mean': 6.4431646847724915, 'latency_p50': 6.467792987823486, 'latency_p90': 7.278361487388611}]

max_input_tokens: 512

max_output_tokens: 64

model_architecture: MistralForCausalLM

model_group: jic062/dpo-v1-Nemo

model_name: jic062-dpo-v1-nemo_v1

model_num_parameters: 12772070400.0

model_repo: jic062/dpo-v1-Nemo

model_size: 13B

num_battles: 12690

num_wins: 6162

propriety_score: 0.749770009199632

propriety_total_count: 1087.0

ranking_group: single

status: inactive

submission_type: basic

throughput_3p7s: 1.62

timestamp: 2024-09-09T15:34:12+00:00

us_pacific_date: 2024-09-09

win_ratio: 0.48557919621749407

Download Preference Data

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 jic062-dpo-v1-nemo-v1-mkmlizer
Waiting for job on jic062-dpo-v1-nemo-v1-mkmlizer to finish
Connection pool is full, discarding connection: %s. Connection pool size: %s
Connection pool is full, discarding connection: %s. Connection pool size: %s
Connection pool is full, discarding connection: %s. Connection pool size: %s
jic062-dpo-v1-nemo-v1-mkmlizer: ╔═════════════════════════════════════════════════════════════════════╗
jic062-dpo-v1-nemo-v1-mkmlizer: ║     _____            __           __                                ║
jic062-dpo-v1-nemo-v1-mkmlizer: ║    / _/ /_ ___    __/ /  ___ ___ / /                                ║
jic062-dpo-v1-nemo-v1-mkmlizer: ║   / _/ / // / |/|/ / _ \/ -_) -_) /                                 ║
jic062-dpo-v1-nemo-v1-mkmlizer: ║  /_//_/\_, /|__,__/_//_/\__/\__/_/                                  ║
jic062-dpo-v1-nemo-v1-mkmlizer: ║       /___/                                                         ║
jic062-dpo-v1-nemo-v1-mkmlizer: ║                                                                     ║
jic062-dpo-v1-nemo-v1-mkmlizer: ║  Version: 0.10.1                                                    ║
jic062-dpo-v1-nemo-v1-mkmlizer: ║  Copyright 2023 MK ONE TECHNOLOGIES Inc.                            ║
jic062-dpo-v1-nemo-v1-mkmlizer: ║  https://mk1.ai                                                     ║
jic062-dpo-v1-nemo-v1-mkmlizer: ║                                                                     ║
jic062-dpo-v1-nemo-v1-mkmlizer: ║  The license key for the current software has been verified as      ║
jic062-dpo-v1-nemo-v1-mkmlizer: ║  belonging to:                                                      ║
jic062-dpo-v1-nemo-v1-mkmlizer: ║                                                                     ║
jic062-dpo-v1-nemo-v1-mkmlizer: ║  Chai Research Corp.                                                ║
jic062-dpo-v1-nemo-v1-mkmlizer: ║  Account ID: 7997a29f-0ceb-4cc7-9adf-840c57b4ae6f                   ║
jic062-dpo-v1-nemo-v1-mkmlizer: ║  Expiration: 2024-10-15 23:59:59                                    ║
jic062-dpo-v1-nemo-v1-mkmlizer: ║                                                                     ║
jic062-dpo-v1-nemo-v1-mkmlizer: ╚═════════════════════════════════════════════════════════════════════╝
Failed to get response for submission blend_jerun_2024-08-22: ('http://chaiml-elo-alignment-run-3-v36-predictor.tenant-chaiml-guanaco.k.chaiverse.com/v1/models/GPT-J-6B-lit-v2:predict', 'read tcp 127.0.0.1:47498->127.0.0.1:8080: read: connection reset by peer\n')
jic062-dpo-v1-nemo-v1-mkmlizer: Downloaded to shared memory in 49.604s
jic062-dpo-v1-nemo-v1-mkmlizer: quantizing model to /dev/shm/model_cache, profile:s0, folder:/tmp/tmpwo8rhmot, device:0
jic062-dpo-v1-nemo-v1-mkmlizer: Saving flywheel model at /dev/shm/model_cache
Failed to get response for submission chaiml-lexical-nemo-v4-1k1e5_v1: ('http://chaiml-llama-8b-pairwis-8189-v19-predictor.tenant-chaiml-guanaco.k.chaiverse.com/v1/models/GPT-J-6B-lit-v2:predict', 'read tcp 127.0.0.1:47472->127.0.0.1:8080: read: connection reset by peer\n')
jic062-dpo-v1-nemo-v1-mkmlizer: quantized model in 36.256s
jic062-dpo-v1-nemo-v1-mkmlizer: Processed model jic062/dpo-v1-Nemo in 85.859s
jic062-dpo-v1-nemo-v1-mkmlizer: creating bucket guanaco-mkml-models
jic062-dpo-v1-nemo-v1-mkmlizer: Bucket 's3://guanaco-mkml-models/' created
jic062-dpo-v1-nemo-v1-mkmlizer: uploading /dev/shm/model_cache to s3://guanaco-mkml-models/jic062-dpo-v1-nemo-v1
jic062-dpo-v1-nemo-v1-mkmlizer: cp /dev/shm/model_cache/config.json s3://guanaco-mkml-models/jic062-dpo-v1-nemo-v1/config.json
jic062-dpo-v1-nemo-v1-mkmlizer: cp /dev/shm/model_cache/special_tokens_map.json s3://guanaco-mkml-models/jic062-dpo-v1-nemo-v1/special_tokens_map.json
jic062-dpo-v1-nemo-v1-mkmlizer: cp /dev/shm/model_cache/tokenizer_config.json s3://guanaco-mkml-models/jic062-dpo-v1-nemo-v1/tokenizer_config.json
jic062-dpo-v1-nemo-v1-mkmlizer: cp /dev/shm/model_cache/tokenizer.json s3://guanaco-mkml-models/jic062-dpo-v1-nemo-v1/tokenizer.json
jic062-dpo-v1-nemo-v1-mkmlizer: cp /dev/shm/model_cache/flywheel_model.0.safetensors s3://guanaco-mkml-models/jic062-dpo-v1-nemo-v1/flywheel_model.0.safetensors
jic062-dpo-v1-nemo-v1-mkmlizer: 
Loading 0:   0%|          | 0/363 [00:00<?, ?it/s]
Loading 0:   1%|▏         | 5/363 [00:00<00:11, 31.73it/s]
Loading 0:   4%|▎         | 13/363 [00:00<00:06, 51.11it/s]
Loading 0:   5%|▌         | 19/363 [00:00<00:07, 46.64it/s]
Loading 0:   7%|▋         | 24/363 [00:00<00:08, 42.31it/s]
Loading 0:   8%|▊         | 30/363 [00:00<00:07, 46.94it/s]
Loading 0:  10%|▉         | 35/363 [00:00<00:07, 45.85it/s]
Loading 0:  11%|█         | 40/363 [00:00<00:06, 46.83it/s]
Loading 0:  13%|█▎        | 46/363 [00:01<00:07, 44.40it/s]
Loading 0:  14%|█▍        | 51/363 [00:01<00:07, 43.21it/s]
Loading 0:  16%|█▋        | 59/363 [00:01<00:05, 52.52it/s]
Loading 0:  18%|█▊        | 65/363 [00:01<00:09, 32.67it/s]
Loading 0:  20%|█▉        | 72/363 [00:01<00:07, 38.95it/s]
Loading 0:  21%|██        | 77/363 [00:01<00:07, 40.84it/s]
Loading 0:  23%|██▎       | 82/363 [00:02<00:08, 34.78it/s]
Loading 0:  25%|██▍       | 90/363 [00:02<00:06, 42.58it/s]
Loading 0:  26%|██▌       | 95/363 [00:02<00:06, 43.85it/s]
Loading 0:  28%|██▊       | 100/363 [00:02<00:06, 37.69it/s]
Loading 0:  30%|██▉       | 108/363 [00:02<00:05, 46.66it/s]
Loading 0:  31%|███▏      | 114/363 [00:02<00:05, 41.60it/s]
Loading 0:  33%|███▎      | 119/363 [00:02<00:05, 40.71it/s]
Loading 0:  34%|███▍      | 125/363 [00:02<00:05, 44.93it/s]
Loading 0:  36%|███▌      | 130/363 [00:03<00:05, 43.05it/s]
Loading 0:  37%|███▋      | 135/363 [00:03<00:05, 41.26it/s]
Loading 0:  39%|███▊      | 140/363 [00:03<00:05, 41.74it/s]
Loading 0:  40%|███▉      | 145/363 [00:03<00:08, 26.84it/s]
Loading 0:  41%|████      | 149/363 [00:03<00:07, 27.34it/s]
Loading 0:  43%|████▎     | 156/363 [00:03<00:05, 35.10it/s]
Loading 0:  44%|████▍     | 161/363 [00:04<00:05, 36.65it/s]
Loading 0:  46%|████▌     | 166/363 [00:04<00:05, 38.23it/s]
Loading 0:  47%|████▋     | 171/363 [00:04<00:04, 40.24it/s]
Loading 0:  48%|████▊     | 176/363 [00:04<00:05, 35.21it/s]
Loading 0:  50%|█████     | 183/363 [00:04<00:04, 42.81it/s]
Loading 0:  52%|█████▏    | 188/363 [00:04<00:04, 42.83it/s]
Loading 0:  53%|█████▎    | 193/363 [00:04<00:03, 42.83it/s]
Loading 0:  55%|█████▍    | 199/363 [00:04<00:03, 41.65it/s]
Loading 0:  56%|█████▌    | 204/363 [00:05<00:03, 40.70it/s]
Loading 0:  58%|█████▊    | 210/363 [00:05<00:03, 44.89it/s]
Loading 0:  59%|█████▉    | 215/363 [00:05<00:03, 44.07it/s]
Loading 0:  61%|██████    | 220/363 [00:05<00:03, 45.52it/s]
Loading 0:  62%|██████▏   | 225/363 [00:05<00:04, 29.02it/s]
Loading 0:  63%|██████▎   | 230/363 [00:05<00:04, 31.33it/s]
Loading 0:  65%|██████▌   | 237/363 [00:05<00:03, 37.23it/s]
Loading 0:  67%|██████▋   | 242/363 [00:06<00:03, 38.54it/s]
Loading 0:  68%|██████▊   | 247/363 [00:06<00:02, 40.00it/s]
Loading 0:  69%|██████▉   | 252/363 [00:06<00:02, 42.22it/s]
Loading 0:  71%|███████   | 257/363 [00:06<00:03, 35.27it/s]
Loading 0:  73%|███████▎  | 264/363 [00:06<00:02, 42.65it/s]
Loading 0:  74%|███████▍  | 269/363 [00:06<00:02, 43.23it/s]
Loading 0:  75%|███████▌  | 274/363 [00:06<00:02, 43.95it/s]
Loading 0:  77%|███████▋  | 279/363 [00:06<00:01, 44.87it/s]
Loading 0:  78%|███████▊  | 284/363 [00:07<00:02, 37.79it/s]
Loading 0:  80%|████████  | 291/363 [00:07<00:01, 45.25it/s]
Loading 0:  82%|████████▏ | 296/363 [00:07<00:01, 44.45it/s]
Loading 0:  83%|████████▎ | 302/363 [00:07<00:01, 48.19it/s]
Loading 0:  85%|████████▍ | 308/363 [00:14<00:20,  2.67it/s]
Loading 0:  86%|████████▌ | 312/363 [00:14<00:15,  3.39it/s]
Loading 0:  88%|████████▊ | 320/363 [00:14<00:07,  5.44it/s]
Loading 0:  90%|████████▉ | 326/363 [00:14<00:05,  7.34it/s]
Loading 0:  91%|█████████ | 331/363 [00:14<00:03,  9.34it/s]
Loading 0:  93%|█████████▎| 338/363 [00:14<00:01, 13.22it/s]
Loading 0:  95%|█████████▍| 344/363 [00:15<00:01, 16.49it/s]
Loading 0:  96%|█████████▌| 349/363 [00:15<00:00, 19.57it/s]
Loading 0:  98%|█████████▊| 356/363 [00:15<00:00, 25.43it/s]
Loading 0: 100%|█████████▉| 362/363 [00:15<00:00, 28.75it/s]
                                                            
Job jic062-dpo-v1-nemo-v1-mkmlizer completed after 106.45s with status: succeeded
Stopping job with name jic062-dpo-v1-nemo-v1-mkmlizer
Pipeline stage MKMLizer completed in 107.35s
run pipeline stage %s
Running pipeline stage MKMLTemplater
Pipeline stage MKMLTemplater completed in 0.23s
run pipeline stage %s
Running pipeline stage MKMLDeployer
Creating inference service jic062-dpo-v1-nemo-v1
Waiting for inference service jic062-dpo-v1-nemo-v1 to be ready
Connection pool is full, discarding connection: %s. Connection pool size: %s
Connection pool is full, discarding connection: %s. Connection pool size: %s
Connection pool is full, discarding connection: %s. Connection pool size: %s
Failed to get response for submission chaiml-elo-alignment-run-3_v36: ('http://chaiml-llama-8b-pairwis-8189-v19-predictor.tenant-chaiml-guanaco.k.chaiverse.com/v1/models/GPT-J-6B-lit-v2:predict', 'read tcp 127.0.0.1:40886->127.0.0.1:8080: read: connection reset by peer\n')
Connection pool is full, discarding connection: %s. Connection pool size: %s
Connection pool is full, discarding connection: %s. Connection pool size: %s
Connection pool is full, discarding connection: %s. Connection pool size: %s
Inference service jic062-dpo-v1-nemo-v1 ready after 150.65467429161072s
Pipeline stage MKMLDeployer completed in 151.01s
run pipeline stage %s
Running pipeline stage StressChecker
Connection pool is full, discarding connection: %s. Connection pool size: %s
Connection pool is full, discarding connection: %s. Connection pool size: %s
Connection pool is full, discarding connection: %s. Connection pool size: %s
Connection pool is full, discarding connection: %s. Connection pool size: %s
Connection pool is full, discarding connection: %s. Connection pool size: %s
Connection pool is full, discarding connection: %s. Connection pool size: %s
Connection pool is full, discarding connection: %s. Connection pool size: %s
Connection pool is full, discarding connection: %s. Connection pool size: %s
Connection pool is full, discarding connection: %s. Connection pool size: %s
Connection pool is full, discarding connection: %s. Connection pool size: %s
Connection pool is full, discarding connection: %s. Connection pool size: %s
Connection pool is full, discarding connection: %s. Connection pool size: %s
Connection pool is full, discarding connection: %s. Connection pool size: %s
Received healthy response to inference request in 2.562763214111328s
Received healthy response to inference request in 2.223562240600586s
Received healthy response to inference request in 1.9033176898956299s
Received healthy response to inference request in 3.4776761531829834s
Received healthy response to inference request in 2.517422914505005s
5 requests
0 failed requests
5th percentile: 1.967366600036621
10th percentile: 2.031415510177612
20th percentile: 2.159513330459595
30th percentile: 2.2823343753814695
40th percentile: 2.399878644943237
50th percentile: 2.517422914505005
60th percentile: 2.5355590343475343
70th percentile: 2.5536951541900637
80th percentile: 2.7457458019256595
90th percentile: 3.1117109775543215
95th percentile: 3.294693565368652
99th percentile: 3.441079635620117
mean time: 2.5369484424591064
Pipeline stage StressChecker completed in 14.81s
run pipeline stage %s
Running pipeline stage TriggerMKMLProfilingPipeline
run_pipeline:run_in_cloud %s
starting trigger_guanaco_pipeline args=%s
Pipeline stage TriggerMKMLProfilingPipeline completed in 4.42s
Shutdown handler de-registered
jic062-dpo-v1-nemo_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 0.14s
run pipeline stage %s
Running pipeline stage MKMLProfilerTemplater
Pipeline stage MKMLProfilerTemplater completed in 0.12s
run pipeline stage %s
Running pipeline stage MKMLProfilerDeployer
Creating inference service jic062-dpo-v1-nemo-v1-profiler
Waiting for inference service jic062-dpo-v1-nemo-v1-profiler to be ready
Inference service jic062-dpo-v1-nemo-v1-profiler ready after 150.37041664123535s
Pipeline stage MKMLProfilerDeployer completed in 150.77s
run pipeline stage %s
Running pipeline stage MKMLProfilerRunner
kubectl cp /code/guanaco/guanaco_inference_services/src/inference_scripts tenant-chaiml-guanaco/jic062-dpo-v1-nemo-v1-profiler-predictor-00001-deployment-s66hx:/code/chaiverse_profiler_1725896527 --namespace tenant-chaiml-guanaco
kubectl exec -it jic062-dpo-v1-nemo-v1-profiler-predictor-00001-deployment-s66hx --namespace tenant-chaiml-guanaco -- sh -c 'cd /code/chaiverse_profiler_1725896527 && 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 512 --output_tokens 64 --summary /code/chaiverse_profiler_1725896527/summary.json'
kubectl exec -it jic062-dpo-v1-nemo-v1-profiler-predictor-00001-deployment-s66hx --namespace tenant-chaiml-guanaco -- bash -c 'cat /code/chaiverse_profiler_1725896527/summary.json'
Pipeline stage MKMLProfilerRunner completed in 956.23s
run pipeline stage %s
Running pipeline stage MKMLProfilerDeleter
Checking if service jic062-dpo-v1-nemo-v1-profiler is running
Tearing down inference service jic062-dpo-v1-nemo-v1-profiler
Service jic062-dpo-v1-nemo-v1-profiler has been torndown
Pipeline stage MKMLProfilerDeleter completed in 1.79s
Shutdown handler de-registered
jic062-dpo-v1-nemo_v1 status is now inactive due to auto deactivation removed underperforming models