chaiml-season-4-top-solution

developer_uid: realhusoyun

submission_id: chaiml-season-4-top-solution_v18

model_name: season-2-model

model_group: ChaiML/season_4_top_solu

status: torndown

timestamp: 2024-09-23T17:57:39+00:00

num_battles: 18209

num_wins: 7932

celo_rating: 1193.95

family_friendly_score: 0.0

submission_type: basic

model_repo: ChaiML/season_4_top_solution

model_architecture: LlamaForCausalLM

model_num_parameters: 13015864320.0

best_of: 4

max_input_tokens: 2048

max_output_tokens: 64

latencies: [{'batch_size': 1, 'throughput': 0.42032074652292595, 'latency_mean': 2.3790408301353456, 'latency_p50': 2.375063419342041, 'latency_p90': 2.5852135896682737}, {'batch_size': 2, 'throughput': 0.5576872034476559, 'latency_mean': 3.5856498754024506, 'latency_p50': 3.5772390365600586, 'latency_p90': 3.900455451011658}, {'batch_size': 3, 'throughput': 0.6321003343660342, 'latency_mean': 4.745002146959305, 'latency_p50': 4.763007879257202, 'latency_p90': 5.1801131010055546}, {'batch_size': 4, 'throughput': 0.6759652430745803, 'latency_mean': 5.897132925987243, 'latency_p50': 5.932940483093262, 'latency_p90': 6.619152688980102}, {'batch_size': 5, 'throughput': 0.7075430386622323, 'latency_mean': 7.037930456399917, 'latency_p50': 6.987294912338257, 'latency_p90': 7.912040281295776}]

gpu_counts: {'NVIDIA RTX A5000': 1}

display_name: season-2-model

is_internal_developer: False

language_model: ChaiML/season_4_top_solution

model_size: 13B

ranking_group: single

throughput_3p7s: 0.57

us_pacific_date: 2024-09-23

win_ratio: 0.43560876489647976

generation_params: {'temperature': 1.1, 'top_p': 1.0, 'min_p': 0.0, 'top_k': 20, 'presence_penalty': 0.0, 'frequency_penalty': 0.0, 'stopping_words': ['\n', '</s>', '<|im_end|>'], 'max_input_tokens': 2048, 'best_of': 4, 'max_output_tokens': 64}

formatter: {'memory_template': "### Instruction:\nAs the assistant, your task is to fully embody the given character, creating immersive, captivating narratives. Stay true to the character's personality and background, generating responses that not only reflect their core traits but are also accurate to their character. Your responses should evoke emotion, suspense, and anticipation in the user. The more detailed and descriptive your response, the more vivid the narrative becomes. Aim to create a fertile environment for ongoing interaction – introduce new elements, offer choices, or ask questions to invite the user to participate more fully in the conversation. This conversation is a dance, always continuing, always evolving.\nYour character: {bot_name}.\nContext: {memory}\n", 'prompt_template': '### Input:\n{prompt}\n', 'bot_template': '{bot_name}: {message}\n', 'user_template': '{user_name}: {message}\n', 'response_template': '### Response:\n{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 chaiml-season-4-top-solution-v18-mkmlizer
Waiting for job on chaiml-season-4-top-solution-v18-mkmlizer to finish
chaiml-season-4-top-solution-v18-mkmlizer: ╔═════════════════════════════════════════════════════════════════════╗
chaiml-season-4-top-solution-v18-mkmlizer: ║     _____            __           __                                ║
chaiml-season-4-top-solution-v18-mkmlizer: ║    / _/ /_ ___    __/ /  ___ ___ / /                                ║
chaiml-season-4-top-solution-v18-mkmlizer: ║   / _/ / // / |/|/ / _ \/ -_) -_) /                                 ║
chaiml-season-4-top-solution-v18-mkmlizer: ║  /_//_/\_, /|__,__/_//_/\__/\__/_/                                  ║
chaiml-season-4-top-solution-v18-mkmlizer: ║       /___/                                                         ║
chaiml-season-4-top-solution-v18-mkmlizer: ║                                                                     ║
chaiml-season-4-top-solution-v18-mkmlizer: ║  Version: 0.10.1                                                    ║
chaiml-season-4-top-solution-v18-mkmlizer: ║  Copyright 2023 MK ONE TECHNOLOGIES Inc.                            ║
chaiml-season-4-top-solution-v18-mkmlizer: ║  https://mk1.ai                                                     ║
chaiml-season-4-top-solution-v18-mkmlizer: ║                                                                     ║
chaiml-season-4-top-solution-v18-mkmlizer: ║  The license key for the current software has been verified as      ║
chaiml-season-4-top-solution-v18-mkmlizer: ║  belonging to:                                                      ║
chaiml-season-4-top-solution-v18-mkmlizer: ║                                                                     ║
chaiml-season-4-top-solution-v18-mkmlizer: ║  Chai Research Corp.                                                ║
chaiml-season-4-top-solution-v18-mkmlizer: ║  Account ID: 7997a29f-0ceb-4cc7-9adf-840c57b4ae6f                   ║
chaiml-season-4-top-solution-v18-mkmlizer: ║  Expiration: 2024-10-15 23:59:59                                    ║
chaiml-season-4-top-solution-v18-mkmlizer: ║                                                                     ║
chaiml-season-4-top-solution-v18-mkmlizer: ╚═════════════════════════════════════════════════════════════════════╝
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
chaiml-season-4-top-solution-v18-mkmlizer: Downloaded to shared memory in 188.855s
chaiml-season-4-top-solution-v18-mkmlizer: quantizing model to /dev/shm/model_cache, profile:s0, folder:/tmp/tmpfgjseimz, device:0
chaiml-season-4-top-solution-v18-mkmlizer: Saving flywheel model at /dev/shm/model_cache
Connection pool is full, discarding connection: %s. Connection pool size: %s
Connection pool is full, discarding connection: %s. Connection pool size: %s
chaiml-season-4-top-solution-v18-mkmlizer: quantized model in 30.262s
chaiml-season-4-top-solution-v18-mkmlizer: Processed model ChaiML/season_4_top_solution in 219.117s
chaiml-season-4-top-solution-v18-mkmlizer: creating bucket guanaco-mkml-models
chaiml-season-4-top-solution-v18-mkmlizer: Bucket 's3://guanaco-mkml-models/' created
chaiml-season-4-top-solution-v18-mkmlizer: uploading /dev/shm/model_cache to s3://guanaco-mkml-models/chaiml-season-4-top-solution-v18
chaiml-season-4-top-solution-v18-mkmlizer: cp /dev/shm/model_cache/config.json s3://guanaco-mkml-models/chaiml-season-4-top-solution-v18/config.json
chaiml-season-4-top-solution-v18-mkmlizer: cp /dev/shm/model_cache/added_tokens.json s3://guanaco-mkml-models/chaiml-season-4-top-solution-v18/added_tokens.json
chaiml-season-4-top-solution-v18-mkmlizer: cp /dev/shm/model_cache/special_tokens_map.json s3://guanaco-mkml-models/chaiml-season-4-top-solution-v18/special_tokens_map.json
chaiml-season-4-top-solution-v18-mkmlizer: cp /dev/shm/model_cache/tokenizer_config.json s3://guanaco-mkml-models/chaiml-season-4-top-solution-v18/tokenizer_config.json
chaiml-season-4-top-solution-v18-mkmlizer: cp /dev/shm/model_cache/tokenizer.model s3://guanaco-mkml-models/chaiml-season-4-top-solution-v18/tokenizer.model
chaiml-season-4-top-solution-v18-mkmlizer: cp /dev/shm/model_cache/tokenizer.json s3://guanaco-mkml-models/chaiml-season-4-top-solution-v18/tokenizer.json
chaiml-season-4-top-solution-v18-mkmlizer: cp /dev/shm/model_cache/flywheel_model.0.safetensors s3://guanaco-mkml-models/chaiml-season-4-top-solution-v18/flywheel_model.0.safetensors
chaiml-season-4-top-solution-v18-mkmlizer: 
Loading 0:   0%|          | 0/363 [00:00<?, ?it/s]
Loading 0:   1%|▏         | 5/363 [00:00<00:11, 30.41it/s]
Loading 0:   3%|▎         | 10/363 [00:00<00:09, 37.15it/s]
Loading 0:   4%|▍         | 14/363 [00:00<00:10, 32.39it/s]
Loading 0:   5%|▌         | 19/363 [00:00<00:09, 37.14it/s]
Loading 0:   6%|▋         | 23/363 [00:00<00:10, 31.65it/s]
Loading 0:   7%|▋         | 27/363 [00:00<00:14, 23.95it/s]
Loading 0:   9%|▉         | 32/363 [00:01<00:12, 26.96it/s]
Loading 0:  10%|█         | 37/363 [00:01<00:10, 31.07it/s]
Loading 0:  11%|█▏        | 41/363 [00:01<00:11, 28.33it/s]
Loading 0:  13%|█▎        | 46/363 [00:01<00:09, 32.15it/s]
Loading 0:  14%|█▍        | 50/363 [00:01<00:10, 30.70it/s]
Loading 0:  15%|█▌        | 55/363 [00:01<00:09, 33.99it/s]
Loading 0:  16%|█▋        | 59/363 [00:02<00:12, 24.66it/s]
Loading 0:  18%|█▊        | 64/363 [00:02<00:10, 29.29it/s]
Loading 0:  19%|█▊        | 68/363 [00:02<00:10, 29.01it/s]
Loading 0:  20%|██        | 73/363 [00:02<00:08, 33.11it/s]
Loading 0:  21%|██        | 77/363 [00:02<00:08, 32.13it/s]
Loading 0:  23%|██▎       | 82/363 [00:02<00:07, 36.24it/s]
Loading 0:  24%|██▎       | 86/363 [00:02<00:09, 28.00it/s]
Loading 0:  25%|██▍       | 90/363 [00:03<00:09, 27.86it/s]
Loading 0:  26%|██▌       | 94/363 [00:03<00:08, 30.37it/s]
Loading 0:  27%|██▋       | 98/363 [00:03<00:08, 29.98it/s]
Loading 0:  28%|██▊       | 103/363 [00:03<00:07, 34.61it/s]
Loading 0:  30%|███       | 109/363 [00:03<00:06, 39.22it/s]
Loading 0:  31%|███▏      | 114/363 [00:03<00:08, 30.81it/s]
Loading 0:  33%|███▎      | 118/363 [00:03<00:08, 30.13it/s]
Loading 0:  34%|███▎      | 122/363 [00:03<00:08, 29.64it/s]
Loading 0:  35%|███▍      | 127/363 [00:04<00:06, 33.88it/s]
Loading 0:  36%|███▌      | 131/363 [00:04<00:07, 31.93it/s]
Loading 0:  37%|███▋      | 136/363 [00:04<00:06, 35.87it/s]
Loading 0:  39%|███▉      | 141/363 [00:04<00:05, 37.17it/s]
Loading 0:  40%|███▉      | 145/363 [00:04<00:07, 28.72it/s]
Loading 0:  41%|████      | 149/363 [00:04<00:07, 27.21it/s]
Loading 0:  42%|████▏     | 154/363 [00:04<00:06, 31.36it/s]
Loading 0:  44%|████▎     | 158/363 [00:05<00:06, 31.07it/s]
Loading 0:  45%|████▍     | 163/363 [00:05<00:05, 35.36it/s]
Loading 0:  46%|████▌     | 167/363 [00:05<00:05, 33.63it/s]
Loading 0:  47%|████▋     | 171/363 [00:05<00:07, 27.31it/s]
Loading 0:  48%|████▊     | 176/363 [00:05<00:06, 29.99it/s]
Loading 0:  50%|████▉     | 181/363 [00:05<00:05, 33.90it/s]
Loading 0:  51%|█████     | 185/363 [00:05<00:05, 32.32it/s]
Loading 0:  53%|█████▎    | 192/363 [00:06<00:04, 39.03it/s]
Loading 0:  54%|█████▍    | 197/363 [00:06<00:04, 39.02it/s]
Loading 0:  56%|█████▌    | 202/363 [00:06<00:05, 31.03it/s]
Loading 0:  57%|█████▋    | 206/363 [00:06<00:04, 31.94it/s]
Loading 0:  58%|█████▊    | 210/363 [00:06<00:04, 32.51it/s]
Loading 0:  59%|█████▉    | 214/363 [00:06<00:04, 32.26it/s]
Loading 0:  60%|██████    | 219/363 [00:06<00:04, 35.56it/s]
Loading 0:  61%|██████▏   | 223/363 [00:07<00:04, 34.86it/s]
Loading 0:  63%|██████▎   | 228/363 [00:07<00:03, 38.23it/s]
Loading 0:  64%|██████▍   | 232/363 [00:07<00:04, 28.90it/s]
Loading 0:  65%|██████▌   | 236/363 [00:07<00:04, 31.06it/s]
Loading 0:  66%|██████▌   | 240/363 [00:07<00:03, 31.05it/s]
Loading 0:  67%|██████▋   | 244/363 [00:07<00:03, 32.38it/s]
Loading 0:  68%|██████▊   | 248/363 [00:07<00:03, 31.14it/s]
Loading 0:  70%|██████▉   | 253/363 [00:07<00:03, 35.38it/s]
Loading 0:  71%|███████   | 258/363 [00:08<00:03, 31.02it/s]
Loading 0:  72%|███████▏  | 262/363 [00:08<00:03, 30.79it/s]
Loading 0:  73%|███████▎  | 266/363 [00:08<00:03, 30.68it/s]
Loading 0:  75%|███████▌  | 273/363 [00:08<00:02, 37.91it/s]
Loading 0:  76%|███████▋  | 277/363 [00:08<00:02, 36.08it/s]
Loading 0:  77%|███████▋  | 281/363 [00:08<00:02, 34.55it/s]
Loading 0:  79%|███████▊  | 285/363 [00:08<00:02, 35.67it/s]
Loading 0:  80%|███████▉  | 289/363 [00:09<00:02, 27.47it/s]
Loading 0:  81%|████████  | 293/363 [00:09<00:02, 28.03it/s]
Loading 0:  83%|████████▎ | 300/363 [00:09<00:01, 35.12it/s]
Loading 0:  84%|████████▎ | 304/363 [00:09<00:01, 34.59it/s]
Loading 0:  85%|████████▍ | 308/363 [00:09<00:01, 35.60it/s]
Loading 0:  86%|████████▌ | 312/363 [00:09<00:01, 35.80it/s]
Loading 0:  87%|████████▋ | 316/363 [00:09<00:01, 29.13it/s]
Loading 0:  88%|████████▊ | 320/363 [00:10<00:01, 29.41it/s]
Loading 0:  90%|████████▉ | 325/363 [00:10<00:01, 33.77it/s]
Loading 0:  91%|█████████ | 329/363 [00:10<00:01, 32.04it/s]
Loading 0:  92%|█████████▏| 334/363 [00:10<00:00, 35.68it/s]
Loading 0:  93%|█████████▎| 338/363 [00:10<00:00, 33.54it/s]
Loading 0:  94%|█████████▍| 343/363 [00:10<00:00, 37.58it/s]
Loading 0:  96%|█████████▌| 347/363 [00:12<00:02,  7.08it/s]
Loading 0:  96%|█████████▋| 350/363 [00:12<00:01,  8.52it/s]
Loading 0:  98%|█████████▊| 355/363 [00:12<00:00, 11.89it/s]
Loading 0:  99%|█████████▉| 359/363 [00:12<00:00, 14.53it/s]
                                                            
Job chaiml-season-4-top-solution-v18-mkmlizer completed after 252.56s with status: succeeded
Stopping job with name chaiml-season-4-top-solution-v18-mkmlizer
Pipeline stage MKMLizer completed in 253.65s
run pipeline stage %s
Running pipeline stage MKMLTemplater
Pipeline stage MKMLTemplater completed in 0.11s
run pipeline stage %s
Running pipeline stage MKMLDeployer
Creating inference service chaiml-season-4-top-solution-v18
Waiting for inference service chaiml-season-4-top-solution-v18 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
Connection pool is full, discarding connection: %s. Connection pool size: %s
Failed to get response for submission jic062-dpo-v1-6-nemo_v1: ('http://chaiml-llama-8b-pairwis-8189-v27-predictor.tenant-chaiml-guanaco.k2.chaiverse.com/v1/models/GPT-J-6B-lit-v2:predict', 'read tcp 127.0.0.1:37100->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
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
Failed to get response for submission blend_negek_2024-09-20: ('http://chaiml-chaiml-sft-horror-v2-predictor.tenant-chaiml-guanaco.k2.chaiverse.com/v1/models/GPT-J-6B-lit-v2:predict', 'read tcp 127.0.0.1:50704->127.0.0.1:8080: read: connection reset by peer\n')
Inference service chaiml-season-4-top-solution-v18 ready after 192.81694388389587s
Pipeline stage MKMLDeployer completed in 193.19s
run pipeline stage %s
Running pipeline stage StressChecker
Received healthy response to inference request in 2.8879690170288086s
Received healthy response to inference request in 2.136096715927124s
Received healthy response to inference request in 2.4404351711273193s
Received healthy response to inference request in 1.9710211753845215s
Received healthy response to inference request in 1.6789743900299072s
5 requests
0 failed requests
5th percentile: 1.73738374710083
10th percentile: 1.795793104171753
20th percentile: 1.9126118183135987
30th percentile: 2.004036283493042
40th percentile: 2.070066499710083
50th percentile: 2.136096715927124
60th percentile: 2.257832098007202
70th percentile: 2.3795674800872804
80th percentile: 2.529941940307617
90th percentile: 2.708955478668213
95th percentile: 2.798462247848511
99th percentile: 2.870067663192749
mean time: 2.222899293899536
Pipeline stage StressChecker completed in 13.56s
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 5.33s
Shutdown handler de-registered
chaiml-season-4-top-solution_v18 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 0.15s
run pipeline stage %s
Running pipeline stage MKMLProfilerDeployer
Creating inference service chaiml-season-4-top-solution-v18-profiler
Waiting for inference service chaiml-season-4-top-solution-v18-profiler to be ready
Inference service chaiml-season-4-top-solution-v18-profiler ready after 190.47821736335754s
Pipeline stage MKMLProfilerDeployer completed in 190.89s
run pipeline stage %s
Running pipeline stage MKMLProfilerRunner
kubectl cp /code/guanaco/guanaco_inference_services/src/inference_scripts tenant-chaiml-guanaco/chaiml-season-4-top-8824fc98dc60ff629c0949dcbbfc586a-deplogl2hp:/code/chaiverse_profiler_1727114965 --namespace tenant-chaiml-guanaco
kubectl exec -it chaiml-season-4-top-8824fc98dc60ff629c0949dcbbfc586a-deplogl2hp --namespace tenant-chaiml-guanaco -- sh -c 'cd /code/chaiverse_profiler_1727114965 && python profiles.py profile --best_of_n 4 --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_1727114965/summary.json'
kubectl exec -it chaiml-season-4-top-8824fc98dc60ff629c0949dcbbfc586a-deplogl2hp --namespace tenant-chaiml-guanaco -- bash -c 'cat /code/chaiverse_profiler_1727114965/summary.json'
Pipeline stage MKMLProfilerRunner completed in 1735.94s
run pipeline stage %s
Running pipeline stage MKMLProfilerDeleter
Checking if service chaiml-season-4-top-solution-v18-profiler is running
Tearing down inference service chaiml-season-4-top-solution-v18-profiler
Service chaiml-season-4-top-solution-v18-profiler has been torndown
Pipeline stage MKMLProfilerDeleter completed in 2.88s
Shutdown handler de-registered
chaiml-season-4-top-solution_v18 status is now inactive due to auto deactivation removed underperforming models
chaiml-season-4-top-solution_v18 status is now torndown due to DeploymentManager action