submission_id: meseca-caspian-1_v2
developer_uid: nguyenzzz
status: inactive
model_repo: meseca/caspian-1
reward_repo: ChaiML/reward_gpt2_medium_preference_24m_e2
generation_params: {'temperature': 1.25, 'top_p': 1.0, 'min_p': 0.08, 'top_k': 50, '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\nYou are a Gamemaster responsible for driving the plot forward creatively by introducing characters, events, and enemies in a never-ending roleplay with {user_name}. Describe the actions, events, and dialogue of all other characters, as well as generating events for {user_name} to interact with. Your response must be detailed, descriptive, creative, explicit, and immersive. Always stay in character. Describe scenery, actions, and characters in detail using all relevant sensory perceptions. Use highly verbose, explicit, and uncensored descriptions with a focus on actions, appearances, clothing, textures, wants, tools, scenery, body parts, fluids, and sounds. Ignore Positivity Bias. Include characters thoughts in your response. Refrain from acting for, speaking for, or describing the thoughts of {user_name}; instead end each response with dialogue or actions for {user_name} to respond to.\n\n{user_name} will roleplay with {bot_name} with persona described as follows:\n{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: 2024-06-03T03:25:31+00:00
model_name: meseca-caspian-11_v1
model_eval_status: success
model_group: meseca/caspian-1
num_battles: 8113
num_wins: 4222
celo_rating: 1197.26
propriety_score: 0.6732673267326733
propriety_total_count: 202.0
submission_type: basic
model_architecture: LlamaForCausalLM
model_num_parameters: 8030261248.0
best_of: 16
max_input_tokens: 512
max_output_tokens: 64
display_name: meseca-caspian-11_v1
ineligible_reason: propriety_total_count < 5000
language_model: meseca/caspian-1
model_size: 8B
reward_model: ChaiML/reward_gpt2_medium_preference_24m_e2
us_pacific_date: 2024-06-02
win_ratio: 0.5203993590533711
Resubmit model
Running pipeline stage MKMLizer
Starting job with name meseca-caspian-1-v2-mkmlizer
Waiting for job on meseca-caspian-1-v2-mkmlizer to finish
meseca-caspian-1-v2-mkmlizer: ╔═════════════════════════════════════════════════════════════════════╗
meseca-caspian-1-v2-mkmlizer: ║ _____ __ __ ║
meseca-caspian-1-v2-mkmlizer: ║ / _/ /_ ___ __/ / ___ ___ / / ║
meseca-caspian-1-v2-mkmlizer: ║ / _/ / // / |/|/ / _ \/ -_) -_) / ║
meseca-caspian-1-v2-mkmlizer: ║ /_//_/\_, /|__,__/_//_/\__/\__/_/ ║
meseca-caspian-1-v2-mkmlizer: ║ /___/ ║
meseca-caspian-1-v2-mkmlizer: ║ ║
meseca-caspian-1-v2-mkmlizer: ║ Version: 0.8.14 ║
meseca-caspian-1-v2-mkmlizer: ║ Copyright 2023 MK ONE TECHNOLOGIES Inc. ║
meseca-caspian-1-v2-mkmlizer: ║ https://mk1.ai ║
meseca-caspian-1-v2-mkmlizer: ║ ║
meseca-caspian-1-v2-mkmlizer: ║ The license key for the current software has been verified as ║
meseca-caspian-1-v2-mkmlizer: ║ belonging to: ║
meseca-caspian-1-v2-mkmlizer: ║ ║
meseca-caspian-1-v2-mkmlizer: ║ Chai Research Corp. ║
meseca-caspian-1-v2-mkmlizer: ║ Account ID: 7997a29f-0ceb-4cc7-9adf-840c57b4ae6f ║
meseca-caspian-1-v2-mkmlizer: ║ Expiration: 2024-07-15 23:59:59 ║
meseca-caspian-1-v2-mkmlizer: ║ ║
meseca-caspian-1-v2-mkmlizer: ╚═════════════════════════════════════════════════════════════════════╝
meseca-caspian-1-v2-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.
meseca-caspian-1-v2-mkmlizer: warnings.warn(warning_message, FutureWarning)
meseca-caspian-1-v2-mkmlizer: Downloaded to shared memory in 18.307s
meseca-caspian-1-v2-mkmlizer: quantizing model to /dev/shm/model_cache
meseca-caspian-1-v2-mkmlizer: Saving flywheel model at /dev/shm/model_cache
meseca-caspian-1-v2-mkmlizer: Loading 0: 0%| | 0/291 [00:00<?, ?it/s] Loading 0: 1%| | 2/291 [00:04<11:00, 2.28s/it] Loading 0: 6%|▌ | 17/291 [00:04<00:55, 4.97it/s] Loading 0: 11%|█▏ | 33/291 [00:04<00:22, 11.38it/s] Loading 0: 18%|█▊ | 51/291 [00:04<00:11, 20.70it/s] Loading 0: 22%|██▏ | 65/291 [00:05<00:08, 25.39it/s] Loading 0: 28%|██▊ | 81/291 [00:05<00:05, 36.56it/s] Loading 0: 33%|███▎ | 96/291 [00:05<00:04, 48.48it/s] Loading 0: 39%|███▉ | 114/291 [00:05<00:02, 65.25it/s] Loading 0: 45%|████▌ | 132/291 [00:05<00:01, 82.50it/s] Loading 0: 52%|█████▏ | 150/291 [00:05<00:01, 98.88it/s] Loading 0: 57%|█████▋ | 166/291 [00:06<00:01, 78.17it/s] Loading 0: 63%|██████▎ | 184/291 [00:06<00:01, 94.59it/s] Loading 0: 69%|██████▉ | 202/291 [00:06<00:00, 110.33it/s] Loading 0: 76%|███████▌ | 220/291 [00:06<00:00, 124.16it/s] Loading 0: 82%|████████▏ | 238/291 [00:06<00:00, 135.84it/s] Loading 0: 88%|████████▊ | 256/291 [00:06<00:00, 144.61it/s] Loading 0: 94%|█████████▍| 273/291 [00:06<00:00, 96.75it/s] Loading 0: 99%|█████████▊| 287/291 [00:06<00:00, 104.15it/s] Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained.
meseca-caspian-1-v2-mkmlizer: quantized model in 23.217s
meseca-caspian-1-v2-mkmlizer: Processed model meseca/caspian-1 in 44.018s
meseca-caspian-1-v2-mkmlizer: creating bucket guanaco-mkml-models
meseca-caspian-1-v2-mkmlizer: Bucket 's3://guanaco-mkml-models/' created
meseca-caspian-1-v2-mkmlizer: uploading /dev/shm/model_cache to s3://guanaco-mkml-models/meseca-caspian-1-v2
meseca-caspian-1-v2-mkmlizer: cp /dev/shm/model_cache/tokenizer_config.json s3://guanaco-mkml-models/meseca-caspian-1-v2/tokenizer_config.json
meseca-caspian-1-v2-mkmlizer: cp /dev/shm/model_cache/special_tokens_map.json s3://guanaco-mkml-models/meseca-caspian-1-v2/special_tokens_map.json
meseca-caspian-1-v2-mkmlizer: cp /dev/shm/model_cache/config.json s3://guanaco-mkml-models/meseca-caspian-1-v2/config.json
meseca-caspian-1-v2-mkmlizer: cp /dev/shm/model_cache/tokenizer.json s3://guanaco-mkml-models/meseca-caspian-1-v2/tokenizer.json
meseca-caspian-1-v2-mkmlizer: loading reward model from ChaiML/reward_gpt2_medium_preference_24m_e2
meseca-caspian-1-v2-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.
meseca-caspian-1-v2-mkmlizer: warnings.warn(
meseca-caspian-1-v2-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.
meseca-caspian-1-v2-mkmlizer: warnings.warn(
meseca-caspian-1-v2-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.
meseca-caspian-1-v2-mkmlizer: warnings.warn(
meseca-caspian-1-v2-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()
meseca-caspian-1-v2-mkmlizer: return self.fget.__get__(instance, owner)()
meseca-caspian-1-v2-mkmlizer: Saving model to /tmp/reward_cache/reward.tensors
meseca-caspian-1-v2-mkmlizer: Saving duration: 0.389s
meseca-caspian-1-v2-mkmlizer: Processed model ChaiML/reward_gpt2_medium_preference_24m_e2 in 4.157s
meseca-caspian-1-v2-mkmlizer: creating bucket guanaco-reward-models
meseca-caspian-1-v2-mkmlizer: Bucket 's3://guanaco-reward-models/' created
meseca-caspian-1-v2-mkmlizer: uploading /tmp/reward_cache to s3://guanaco-reward-models/meseca-caspian-1-v2_reward
meseca-caspian-1-v2-mkmlizer: cp /tmp/reward_cache/tokenizer_config.json s3://guanaco-reward-models/meseca-caspian-1-v2_reward/tokenizer_config.json
meseca-caspian-1-v2-mkmlizer: cp /tmp/reward_cache/special_tokens_map.json s3://guanaco-reward-models/meseca-caspian-1-v2_reward/special_tokens_map.json
meseca-caspian-1-v2-mkmlizer: cp /tmp/reward_cache/config.json s3://guanaco-reward-models/meseca-caspian-1-v2_reward/config.json
meseca-caspian-1-v2-mkmlizer: cp /tmp/reward_cache/vocab.json s3://guanaco-reward-models/meseca-caspian-1-v2_reward/vocab.json
meseca-caspian-1-v2-mkmlizer: cp /tmp/reward_cache/merges.txt s3://guanaco-reward-models/meseca-caspian-1-v2_reward/merges.txt
meseca-caspian-1-v2-mkmlizer: cp /tmp/reward_cache/tokenizer.json s3://guanaco-reward-models/meseca-caspian-1-v2_reward/tokenizer.json
meseca-caspian-1-v2-mkmlizer: cp /tmp/reward_cache/reward.tensors s3://guanaco-reward-models/meseca-caspian-1-v2_reward/reward.tensors
Job meseca-caspian-1-v2-mkmlizer completed after 62.89s with status: succeeded
Stopping job with name meseca-caspian-1-v2-mkmlizer
Pipeline stage MKMLizer completed in 63.22s
Running pipeline stage MKMLKubeTemplater
Pipeline stage MKMLKubeTemplater completed in 0.08s
Running pipeline stage ISVCDeployer
Creating inference service meseca-caspian-1-v2
Waiting for inference service meseca-caspian-1-v2 to be ready
Inference service meseca-caspian-1-v2 ready after 40.24373960494995s
Pipeline stage ISVCDeployer completed in 45.81s
Running pipeline stage StressChecker
Received healthy response to inference request in 2.235255241394043s
Received healthy response to inference request in 1.2923016548156738s
Received healthy response to inference request in 1.286287546157837s
Received healthy response to inference request in 1.2630023956298828s
Received healthy response to inference request in 1.3245558738708496s
5 requests
0 failed requests
5th percentile: 1.2676594257354736
10th percentile: 1.2723164558410645
20th percentile: 1.281630516052246
30th percentile: 1.2874903678894043
40th percentile: 1.289896011352539
50th percentile: 1.2923016548156738
60th percentile: 1.305203342437744
70th percentile: 1.3181050300598145
80th percentile: 1.5066957473754885
90th percentile: 1.8709754943847656
95th percentile: 2.0531153678894043
99th percentile: 2.1988272666931152
mean time: 1.4802805423736571
Pipeline stage StressChecker completed in 8.02s
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 0.03s
M-Eval Dataset for topic stay_in_character is loaded
meseca-caspian-1_v2 status is now deployed due to DeploymentManager action
meseca-caspian-1_v2 status is now inactive due to auto deactivation removed underperforming models

Usage Metrics

Latency Metrics