architecture:
    backbone_dtype: int4
    force_embedding_gradients: false
    gradient_checkpointing: false
    intermediate_dropout: 0.0
    pretrained: true
    pretrained_weights: ''
augmentation:
    random_parent_probability: 0.05
    skip_parent_probability: 0.05
    token_mask_probability: 0.05
dataset:
    add_eos_token_to_answer: true
    add_eos_token_to_prompt: true
    add_eos_token_to_system: true
    answer_column: response
    chatbot_author: H2O.ai
    chatbot_name: h2oGPT
    data_sample: 1.0
    data_sample_choice:
    - Train
    - Validation
    limit_chained_samples: false
    mask_prompt_labels: true
    parent_id_column: None
    personalize: false
    prompt_column:
    - prompt
    system_column: None
    text_answer_separator: <|answer|>
    text_prompt_start: <|prompt|>
    text_system_start: <|system|>
    train_dataframe: /workspace/data/user/ultimus_2K/ultimus_2K.csv
    validation_dataframe: None
    validation_size: 0.1
    validation_strategy: automatic
environment:
    compile_model: false
    deepspeed_reduce_bucket_size: 1000000
    deepspeed_stage3_param_persistence_threshold: 1000000
    deepspeed_stage3_prefetch_bucket_size: 1000000
    find_unused_parameters: false
    gpus:
    - '0'
    - '1'
    - '2'
    - '3'
    huggingface_branch: main
    mixed_precision: true
    number_of_workers: 8
    seed: -1
    trust_remote_code: true
    use_deepspeed: false
experiment_name: rampant-kagu.1.1.1.1
llm_backbone: mistralai/Mistral-7B-v0.1
logging:
    logger: None
    neptune_project: ''
output_directory: /workspace/output/user/rampant-kagu.1.1.1.1/
prediction:
    batch_size_inference: 4
    do_sample: true
    max_length_inference: 256
    metric: BLEU
    metric_gpt_model: gpt-4
    min_length_inference: 2
    num_beams: 2
    num_history: 4
    repetition_penalty: 1.2
    stop_tokens: ''
    temperature: 0.35
    top_k: 0
    top_p: 1.0
problem_type: text_causal_language_modeling
tokenizer:
    add_prefix_space: false
    add_prompt_answer_tokens: false
    max_length: 704
    max_length_answer: 352
    max_length_prompt: 256
    padding_quantile: 1.0
    use_fast: true
training:
    batch_size: 4
    differential_learning_rate: 1.0e-05
    differential_learning_rate_layers: []
    drop_last_batch: true
    epochs: 4
    evaluate_before_training: false
    evaluation_epochs: 1.0
    grad_accumulation: 1
    gradient_clip: 1.0
    learning_rate: 1.0e-05
    lora: true
    lora_alpha: 16
    lora_dropout: 0.05
    lora_r: 8
    lora_target_modules: ''
    loss_function: TokenAveragedCrossEntropy
    optimizer: AdamW
    save_best_checkpoint: false
    schedule: Cosine
    train_validation_data: false
    warmup_epochs: 1.0
    weight_decay: 1.0e-05