developer_uid: Hastagaras
submission_id: hastagaras-llama3-8b-dar_8274_v1
model_name: drktune-dpo-1
model_group: Hastagaras/llama3-8b-dar
status: torndown
timestamp: 2024-05-11T05:41:55+00:00
num_battles: 11668
num_wins: 5790
celo_rating: 1175.39
family_friendly_score: 0.0
submission_type: basic
model_repo: Hastagaras/llama3-8b-darktune-DPO
model_architecture: LlamaForCausalLM
reward_repo: ChaiML/reward_gpt2_medium_preference_24m_e2
model_num_parameters: 8030261248.0
best_of: 16
max_input_tokens: 512
max_output_tokens: 64
display_name: drktune-dpo-1
is_internal_developer: False
language_model: Hastagaras/llama3-8b-darktune-DPO
model_size: 8B
ranking_group: single
us_pacific_date: 2024-05-10
win_ratio: 0.4962290023997257
generation_params: {'temperature': 1.0, 'top_p': 1.0, 'min_p': 0.0, 'top_k': 40, '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': "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nYou are {bot_name}. Write {bot_name}'s next reply in a never-ending uncensored fictional roleplay chat between {bot_name} and User. Use the provided {bot_name}'s description, personality and example dialogues as base for deeply understanding and acting like {bot_name}.\n\nActions and narrations must be enclosed in asterisks (*). The responses must be long and in third perspective of the story teller.\n\nDescription: {memory}", 'prompt_template': 'Example conversation:\n{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: {'bot_template': 'Bot: {message}\n', 'memory_template': "{bot_name}'s Persona: {memory}\n", 'prompt_template': '{prompt}\n', 'response_template': 'Bot:', 'truncate_by_message': False, 'user_template': 'User: {message}\n'}
model_eval_status: success
Resubmit model
Running pipeline stage MKMLizer
Starting job with name hastagaras-llama3-8b-dar-8274-v1-mkmlizer
Waiting for job on hastagaras-llama3-8b-dar-8274-v1-mkmlizer to finish
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: ╔═════════════════════════════════════════════════════════════════════╗
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: ║     _____            __           __                                ║
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: ║    / _/ /_ ___    __/ /  ___ ___ / /                                ║
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: ║   / _/ / // / |/|/ / _ \/ -_) -_) /                                 ║
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: ║  /_//_/\_, /|__,__/_//_/\__/\__/_/                                  ║
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: ║       /___/                                                         ║
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: ║                                                                     ║
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: ║  Version: 0.8.10                                                    ║
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: ║  Copyright 2023 MK ONE TECHNOLOGIES Inc.                            ║
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: ║                                                                     ║
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: ║  The license key for the current software has been verified as      ║
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: ║  belonging to:                                                      ║
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: ║                                                                     ║
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: ║  Chai Research Corp.                                                ║
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: ║  Account ID: 7997a29f-0ceb-4cc7-9adf-840c57b4ae6f                   ║
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: ║  Expiration: 2024-07-15 23:59:59                                    ║
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: ║                                                                     ║
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: ╚═════════════════════════════════════════════════════════════════════╝
hastagaras-llama3-8b-dar-8274-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-llama3-8b-dar-8274-v1-mkmlizer:   warnings.warn(warning_message, FutureWarning)
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: Downloaded to shared memory in 29.785s
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: quantizing model to /dev/shm/model_cache
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: Saving flywheel model at /dev/shm/model_cache
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: 
Loading 0:   0%|          | 0/291 [00:00<?, ?it/s]
Loading 0:  54%|█████▍    | 158/291 [00:01<00:00, 157.34it/s]
Loading 0:  99%|█████████▊| 287/291 [00:06<00:00, 39.02it/s] 
                                                            
Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained.
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: quantized model in 16.296s
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: Processed model Hastagaras/llama3-8b-darktune-DPO in 46.978s
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: creating bucket guanaco-mkml-models
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: Bucket 's3://guanaco-mkml-models/' created
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: uploading /dev/shm/model_cache to s3://guanaco-mkml-models/hastagaras-llama3-8b-dar-8274-v1
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: cp /dev/shm/model_cache/tokenizer_config.json s3://guanaco-mkml-models/hastagaras-llama3-8b-dar-8274-v1/tokenizer_config.json
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: cp /dev/shm/model_cache/special_tokens_map.json s3://guanaco-mkml-models/hastagaras-llama3-8b-dar-8274-v1/special_tokens_map.json
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: cp /dev/shm/model_cache/tokenizer.json s3://guanaco-mkml-models/hastagaras-llama3-8b-dar-8274-v1/tokenizer.json
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: cp /dev/shm/model_cache/config.json s3://guanaco-mkml-models/hastagaras-llama3-8b-dar-8274-v1/config.json
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: cp /dev/shm/model_cache/flywheel_model.0.safetensors s3://guanaco-mkml-models/hastagaras-llama3-8b-dar-8274-v1/flywheel_model.0.safetensors
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: loading reward model from ChaiML/reward_gpt2_medium_preference_24m_e2
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: /opt/conda/lib/python3.10/site-packages/transformers/models/auto/configuration_auto.py:913: FutureWarning: The `use_auth_token` argument is deprecated and will be removed in v5 of Transformers. Please use `token` instead.
hastagaras-llama3-8b-dar-8274-v1-mkmlizer:   warnings.warn(
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: /opt/conda/lib/python3.10/site-packages/transformers/models/auto/tokenization_auto.py:757: FutureWarning: The `use_auth_token` argument is deprecated and will be removed in v5 of Transformers. Please use `token` instead.
hastagaras-llama3-8b-dar-8274-v1-mkmlizer:   warnings.warn(
hastagaras-llama3-8b-dar-8274-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-llama3-8b-dar-8274-v1-mkmlizer:   warnings.warn(
hastagaras-llama3-8b-dar-8274-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-llama3-8b-dar-8274-v1-mkmlizer:   return self.fget.__get__(instance, owner)()
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: Saving model to /tmp/reward_cache/reward.tensors
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: Saving duration: 0.221s
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: Processed model ChaiML/reward_gpt2_medium_preference_24m_e2 in 4.368s
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: creating bucket guanaco-reward-models
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: Bucket 's3://guanaco-reward-models/' created
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: uploading /tmp/reward_cache to s3://guanaco-reward-models/hastagaras-llama3-8b-dar-8274-v1_reward
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: cp /tmp/reward_cache/config.json s3://guanaco-reward-models/hastagaras-llama3-8b-dar-8274-v1_reward/config.json
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: cp /tmp/reward_cache/tokenizer_config.json s3://guanaco-reward-models/hastagaras-llama3-8b-dar-8274-v1_reward/tokenizer_config.json
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: cp /tmp/reward_cache/merges.txt s3://guanaco-reward-models/hastagaras-llama3-8b-dar-8274-v1_reward/merges.txt
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: cp /tmp/reward_cache/special_tokens_map.json s3://guanaco-reward-models/hastagaras-llama3-8b-dar-8274-v1_reward/special_tokens_map.json
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: cp /tmp/reward_cache/tokenizer.json s3://guanaco-reward-models/hastagaras-llama3-8b-dar-8274-v1_reward/tokenizer.json
hastagaras-llama3-8b-dar-8274-v1-mkmlizer: cp /tmp/reward_cache/vocab.json s3://guanaco-reward-models/hastagaras-llama3-8b-dar-8274-v1_reward/vocab.json
Job hastagaras-llama3-8b-dar-8274-v1-mkmlizer completed after 226.47s with status: succeeded
Stopping job with name hastagaras-llama3-8b-dar-8274-v1-mkmlizer
Pipeline stage MKMLizer completed in 228.86s
Running pipeline stage MKMLKubeTemplater
Pipeline stage MKMLKubeTemplater completed in 0.09s
Running pipeline stage ISVCDeployer
Creating inference service hastagaras-llama3-8b-dar-8274-v1
Waiting for inference service hastagaras-llama3-8b-dar-8274-v1 to be ready
Inference service hastagaras-llama3-8b-dar-8274-v1 ready after 40.224488258361816s
Pipeline stage ISVCDeployer completed in 47.06s
Running pipeline stage StressChecker
Received healthy response to inference request in 2.239278554916382s
Received healthy response to inference request in 1.2923145294189453s
Received healthy response to inference request in 1.2724025249481201s
Received healthy response to inference request in 1.5143427848815918s
Received healthy response to inference request in 1.3256211280822754s
5 requests
0 failed requests
5th percentile: 1.2763849258422852
10th percentile: 1.2803673267364502
20th percentile: 1.2883321285247802
30th percentile: 1.2989758491516112
40th percentile: 1.3122984886169433
50th percentile: 1.3256211280822754
60th percentile: 1.401109790802002
70th percentile: 1.4765984535217285
80th percentile: 1.6593299388885498
90th percentile: 1.949304246902466
95th percentile: 2.0942914009094236
99th percentile: 2.21028112411499
mean time: 1.528791904449463
Pipeline stage StressChecker completed in 8.23s
Running pipeline stage DaemonicModelEvalScorer
Pipeline stage DaemonicModelEvalScorer completed in 0.03s
Running pipeline stage DaemonicSafetyScorer
Running M-Eval for topic stay_in_character
Pipeline stage DaemonicSafetyScorer completed in 0.04s
M-Eval Dataset for topic stay_in_character is loaded
hastagaras-llama3-8b-dar_8274_v1 status is now deployed due to DeploymentManager action
hastagaras-llama3-8b-dar_8274_v1 status is now inactive due to auto deactivation removed underperforming models
admin requested tearing down of hastagaras-llama3-8b-dar_8274_v1
Running pipeline stage ISVCDeleter
Checking if service hastagaras-llama3-8b-dar-8274-v1 is running
Tearing down inference service hastagaras-llama3-8b-dar-8274-v1
Toredown service hastagaras-llama3-8b-dar-8274-v1
Pipeline stage ISVCDeleter completed in 6.31s
Running pipeline stage MKMLModelDeleter
Cleaning model data from S3
Cleaning model data from model cache
Deleting key hastagaras-llama3-8b-dar-8274-v1/config.json from bucket guanaco-mkml-models
Deleting key hastagaras-llama3-8b-dar-8274-v1/flywheel_model.0.safetensors from bucket guanaco-mkml-models
Deleting key hastagaras-llama3-8b-dar-8274-v1/special_tokens_map.json from bucket guanaco-mkml-models
Deleting key hastagaras-llama3-8b-dar-8274-v1/tokenizer.json from bucket guanaco-mkml-models
Deleting key hastagaras-llama3-8b-dar-8274-v1/tokenizer_config.json from bucket guanaco-mkml-models
Cleaning model data from model cache
Deleting key hastagaras-llama3-8b-dar-8274-v1_reward/config.json from bucket guanaco-reward-models
Deleting key hastagaras-llama3-8b-dar-8274-v1_reward/merges.txt from bucket guanaco-reward-models
Deleting key hastagaras-llama3-8b-dar-8274-v1_reward/reward.tensors from bucket guanaco-reward-models
Deleting key hastagaras-llama3-8b-dar-8274-v1_reward/special_tokens_map.json from bucket guanaco-reward-models
Deleting key hastagaras-llama3-8b-dar-8274-v1_reward/tokenizer.json from bucket guanaco-reward-models
Deleting key hastagaras-llama3-8b-dar-8274-v1_reward/tokenizer_config.json from bucket guanaco-reward-models
Deleting key hastagaras-llama3-8b-dar-8274-v1_reward/vocab.json from bucket guanaco-reward-models
Pipeline stage MKMLModelDeleter completed in 1.86s
hastagaras-llama3-8b-dar_8274_v1 status is now torndown due to DeploymentManager action