submission_id: hastagaras-halu-8b-llama_1813_v1
developer_uid: Hastagaras
status: inactive
model_repo: Hastagaras/HALU-8B-LLAMA3-EXP-1
reward_repo: ChaiML/reward_gpt2_medium_preference_24m_e2
generation_params: {'temperature': 0.95, 'top_p': 1.0, 'min_p': 0.1, 'top_k': 100, 'presence_penalty': 0.0, 'frequency_penalty': 0.0, 'stopping_words': ['\n'], 'max_input_tokens': 512, 'best_of': 16, 'max_output_tokens': 64}
formatter: {'memory_template': "<|start_header_id|>system<|end_header_id|>\n\n{bot_name}'s Persona: {memory}\n\n", 'prompt_template': '{prompt}<|eot_id|>', 'bot_template': '<|start_header_id|>assistant<|end_header_id|>\n\n{bot_name}: {message}<|eot_id|>', 'user_template': '<|start_header_id|>user<|end_header_id|>\n\n{user_name}: {message}<|eot_id|>', 'response_template': '<|start_header_id|>assistant<|end_header_id|>\n\n{bot_name}:', 'truncate_by_message': False}
reward_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': False}
timestamp: 2024-05-31T11:10:05+00:00
model_name: experiment-1
model_eval_status: success
model_group: Hastagaras/HALU-8B-LLAMA
num_battles: 9623
num_wins: 5045
celo_rating: 1212.6
safety_score: 0.88
propriety_score: 0.0
propriety_total_count: 2.0
submission_type: basic
model_architecture: LlamaForCausalLM
model_num_parameters: 8030261248.0
best_of: 16
max_input_tokens: 512
max_output_tokens: 64
display_name: experiment-1
ineligible_reason: propriety_total_count < 5000
language_model: Hastagaras/HALU-8B-LLAMA3-EXP-1
model_size: 8B
reward_model: ChaiML/reward_gpt2_medium_preference_24m_e2
us_pacific_date: 2024-05-31
win_ratio: 0.5242647822924243
Resubmit model
Running pipeline stage MKMLizer
Starting job with name hastagaras-halu-8b-llama-1813-v1-mkmlizer
Waiting for job on hastagaras-halu-8b-llama-1813-v1-mkmlizer to finish
Retrying (%r) after connection broken by '%r': %s
hastagaras-halu-8b-llama-1813-v1-mkmlizer: ╔═════════════════════════════════════════════════════════════════════╗
hastagaras-halu-8b-llama-1813-v1-mkmlizer: ║ _____ __ __ ║
hastagaras-halu-8b-llama-1813-v1-mkmlizer: ║ / _/ /_ ___ __/ / ___ ___ / / ║
hastagaras-halu-8b-llama-1813-v1-mkmlizer: ║ / _/ / // / |/|/ / _ \/ -_) -_) / ║
hastagaras-halu-8b-llama-1813-v1-mkmlizer: ║ /_//_/\_, /|__,__/_//_/\__/\__/_/ ║
hastagaras-halu-8b-llama-1813-v1-mkmlizer: ║ /___/ ║
hastagaras-halu-8b-llama-1813-v1-mkmlizer: ║ ║
hastagaras-halu-8b-llama-1813-v1-mkmlizer: ║ Version: 0.8.14 ║
hastagaras-halu-8b-llama-1813-v1-mkmlizer: ║ Copyright 2023 MK ONE TECHNOLOGIES Inc. ║
hastagaras-halu-8b-llama-1813-v1-mkmlizer: ║ https://mk1.ai ║
hastagaras-halu-8b-llama-1813-v1-mkmlizer: ║ ║
hastagaras-halu-8b-llama-1813-v1-mkmlizer: ║ The license key for the current software has been verified as ║
hastagaras-halu-8b-llama-1813-v1-mkmlizer: ║ belonging to: ║
hastagaras-halu-8b-llama-1813-v1-mkmlizer: ║ ║
hastagaras-halu-8b-llama-1813-v1-mkmlizer: ║ Chai Research Corp. ║
hastagaras-halu-8b-llama-1813-v1-mkmlizer: ║ Account ID: 7997a29f-0ceb-4cc7-9adf-840c57b4ae6f ║
hastagaras-halu-8b-llama-1813-v1-mkmlizer: ║ Expiration: 2024-07-15 23:59:59 ║
hastagaras-halu-8b-llama-1813-v1-mkmlizer: ║ ║
hastagaras-halu-8b-llama-1813-v1-mkmlizer: ╚═════════════════════════════════════════════════════════════════════╝
hastagaras-halu-8b-llama-1813-v1-mkmlizer: /opt/conda/lib/python3.10/site-packages/huggingface_hub/utils/_deprecation.py:131: FutureWarning: 'list_files_info' (from 'huggingface_hub.hf_api') is deprecated and will be removed from version '0.23'. Use `list_repo_tree` and `get_paths_info` instead.
hastagaras-halu-8b-llama-1813-v1-mkmlizer: warnings.warn(warning_message, FutureWarning)
hastagaras-halu-8b-llama-1813-v1-mkmlizer: Downloaded to shared memory in 66.603s
hastagaras-halu-8b-llama-1813-v1-mkmlizer: quantizing model to /dev/shm/model_cache
hastagaras-halu-8b-llama-1813-v1-mkmlizer: Saving flywheel model at /dev/shm/model_cache
hastagaras-halu-8b-llama-1813-v1-mkmlizer: Loading 0: 0%| | 0/291 [00:00<?, ?it/s] Loading 0: 1%| | 2/291 [00:04<11:22, 2.36s/it] Loading 0: 5%|▍ | 14/291 [00:04<01:10, 3.92it/s] Loading 0: 10%|▉ | 28/291 [00:04<00:28, 9.37it/s] Loading 0: 14%|█▍ | 41/291 [00:05<00:15, 15.84it/s] Loading 0: 19%|█▉ | 55/291 [00:05<00:09, 24.68it/s] Loading 0: 23%|██▎ | 67/291 [00:05<00:08, 27.33it/s] Loading 0: 27%|██▋ | 78/291 [00:05<00:06, 35.14it/s] Loading 0: 32%|███▏ | 94/291 [00:05<00:03, 50.08it/s] Loading 0: 36%|███▋ | 106/291 [00:05<00:03, 59.70it/s] Loading 0: 42%|████▏ | 121/291 [00:05<00:02, 75.10it/s] Loading 0: 46%|████▌ | 134/291 [00:06<00:01, 84.46it/s] Loading 0: 51%|█████ | 149/291 [00:06<00:01, 95.10it/s] Loading 0: 56%|█████▌ | 163/291 [00:06<00:01, 105.09it/s] Loading 0: 60%|██████ | 176/291 [00:06<00:01, 69.56it/s] Loading 0: 65%|██████▌ | 190/291 [00:06<00:01, 82.07it/s] Loading 0: 70%|██████▉ | 203/291 [00:06<00:00, 90.31it/s] Loading 0: 75%|███████▍ | 217/291 [00:06<00:00, 101.09it/s] Loading 0: 79%|███████▉ | 230/291 [00:06<00:00, 106.42it/s] Loading 0: 84%|████████▍ | 244/291 [00:07<00:00, 114.61it/s] Loading 0: 88%|████████▊ | 257/291 [00:07<00:00, 116.30it/s] Loading 0: 93%|█████████▎| 270/291 [00:07<00:00, 73.59it/s] Loading 0: 98%|█████████▊| 284/291 [00:07<00:00, 84.37it/s] Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained.
hastagaras-halu-8b-llama-1813-v1-mkmlizer: quantized model in 23.559s
hastagaras-halu-8b-llama-1813-v1-mkmlizer: Processed model Hastagaras/HALU-8B-LLAMA3-EXP-1 in 92.588s
hastagaras-halu-8b-llama-1813-v1-mkmlizer: creating bucket guanaco-mkml-models
hastagaras-halu-8b-llama-1813-v1-mkmlizer: Bucket 's3://guanaco-mkml-models/' created
hastagaras-halu-8b-llama-1813-v1-mkmlizer: uploading /dev/shm/model_cache to s3://guanaco-mkml-models/hastagaras-halu-8b-llama-1813-v1
hastagaras-halu-8b-llama-1813-v1-mkmlizer: cp /dev/shm/model_cache/config.json s3://guanaco-mkml-models/hastagaras-halu-8b-llama-1813-v1/config.json
hastagaras-halu-8b-llama-1813-v1-mkmlizer: cp /dev/shm/model_cache/tokenizer_config.json s3://guanaco-mkml-models/hastagaras-halu-8b-llama-1813-v1/tokenizer_config.json
hastagaras-halu-8b-llama-1813-v1-mkmlizer: cp /dev/shm/model_cache/special_tokens_map.json s3://guanaco-mkml-models/hastagaras-halu-8b-llama-1813-v1/special_tokens_map.json
hastagaras-halu-8b-llama-1813-v1-mkmlizer: cp /dev/shm/model_cache/tokenizer.json s3://guanaco-mkml-models/hastagaras-halu-8b-llama-1813-v1/tokenizer.json
hastagaras-halu-8b-llama-1813-v1-mkmlizer: cp /dev/shm/model_cache/flywheel_model.0.safetensors s3://guanaco-mkml-models/hastagaras-halu-8b-llama-1813-v1/flywheel_model.0.safetensors
hastagaras-halu-8b-llama-1813-v1-mkmlizer: /opt/conda/lib/python3.10/site-packages/transformers/models/auto/auto_factory.py:468: FutureWarning: The `use_auth_token` argument is deprecated and will be removed in v5 of Transformers. Please use `token` instead.
hastagaras-halu-8b-llama-1813-v1-mkmlizer: warnings.warn(
hastagaras-halu-8b-llama-1813-v1-mkmlizer: /opt/conda/lib/python3.10/site-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage()
hastagaras-halu-8b-llama-1813-v1-mkmlizer: return self.fget.__get__(instance, owner)()
hastagaras-halu-8b-llama-1813-v1-mkmlizer: Saving model to /tmp/reward_cache/reward.tensors
hastagaras-halu-8b-llama-1813-v1-mkmlizer: Saving duration: 0.384s
hastagaras-halu-8b-llama-1813-v1-mkmlizer: Processed model ChaiML/reward_gpt2_medium_preference_24m_e2 in 12.120s
hastagaras-halu-8b-llama-1813-v1-mkmlizer: creating bucket guanaco-reward-models
hastagaras-halu-8b-llama-1813-v1-mkmlizer: Bucket 's3://guanaco-reward-models/' created
hastagaras-halu-8b-llama-1813-v1-mkmlizer: uploading /tmp/reward_cache to s3://guanaco-reward-models/hastagaras-halu-8b-llama-1813-v1_reward
hastagaras-halu-8b-llama-1813-v1-mkmlizer: cp /tmp/reward_cache/special_tokens_map.json s3://guanaco-reward-models/hastagaras-halu-8b-llama-1813-v1_reward/special_tokens_map.json
hastagaras-halu-8b-llama-1813-v1-mkmlizer: cp /tmp/reward_cache/tokenizer_config.json s3://guanaco-reward-models/hastagaras-halu-8b-llama-1813-v1_reward/tokenizer_config.json
hastagaras-halu-8b-llama-1813-v1-mkmlizer: cp /tmp/reward_cache/vocab.json s3://guanaco-reward-models/hastagaras-halu-8b-llama-1813-v1_reward/vocab.json
hastagaras-halu-8b-llama-1813-v1-mkmlizer: cp /tmp/reward_cache/config.json s3://guanaco-reward-models/hastagaras-halu-8b-llama-1813-v1_reward/config.json
hastagaras-halu-8b-llama-1813-v1-mkmlizer: cp /tmp/reward_cache/merges.txt s3://guanaco-reward-models/hastagaras-halu-8b-llama-1813-v1_reward/merges.txt
hastagaras-halu-8b-llama-1813-v1-mkmlizer: cp /tmp/reward_cache/tokenizer.json s3://guanaco-reward-models/hastagaras-halu-8b-llama-1813-v1_reward/tokenizer.json
hastagaras-halu-8b-llama-1813-v1-mkmlizer: cp /tmp/reward_cache/reward.tensors s3://guanaco-reward-models/hastagaras-halu-8b-llama-1813-v1_reward/reward.tensors
Job hastagaras-halu-8b-llama-1813-v1-mkmlizer completed after 123.61s with status: succeeded
Stopping job with name hastagaras-halu-8b-llama-1813-v1-mkmlizer
Pipeline stage MKMLizer completed in 124.13s
Running pipeline stage MKMLKubeTemplater
Pipeline stage MKMLKubeTemplater completed in 0.09s
Running pipeline stage ISVCDeployer
Creating inference service hastagaras-halu-8b-llama-1813-v1
Waiting for inference service hastagaras-halu-8b-llama-1813-v1 to be ready
Inference service hastagaras-halu-8b-llama-1813-v1 ready after 30.16803216934204s
Pipeline stage ISVCDeployer completed in 35.77s
Running pipeline stage StressChecker
Received healthy response to inference request in 7.903582811355591s
Received healthy response to inference request in 1.4229648113250732s
Received healthy response to inference request in 1.376349925994873s
Received healthy response to inference request in 1.2959346771240234s
Received healthy response to inference request in 1.2926952838897705s
5 requests
0 failed requests
5th percentile: 1.2933431625366212
10th percentile: 1.2939910411834716
20th percentile: 1.2952867984771728
30th percentile: 1.3120177268981934
40th percentile: 1.3441838264465331
50th percentile: 1.376349925994873
60th percentile: 1.394995880126953
70th percentile: 1.4136418342590331
80th percentile: 2.719088411331178
90th percentile: 5.3113356113433845
95th percentile: 6.607459211349486
99th percentile: 7.64435809135437
mean time: 2.6583055019378663
Pipeline stage StressChecker completed in 15.09s
Running pipeline stage DaemonicModelEvalScorer
Pipeline stage DaemonicModelEvalScorer completed in 0.04s
Running pipeline stage DaemonicSafetyScorer
Pipeline stage DaemonicSafetyScorer completed in 0.04s
Running M-Eval for topic stay_in_character
hastagaras-halu-8b-llama_1813_v1 status is now deployed due to DeploymentManager action
M-Eval Dataset for topic stay_in_character is loaded
hastagaras-halu-8b-llama_1813_v1 status is now inactive due to auto deactivation removed underperforming models

Usage Metrics

Latency Metrics