architecture:
    backbone_dtype: float32
    force_embedding_gradients: true
    gradient_checkpointing: true
    intermediate_dropout: 0.0
    pretrained: true
    pretrained_weights: ''
augmentation:
    neftune_noise_alpha: 0.0
    random_parent_probability: 0.0
    skip_parent_probability: 0.0
    token_mask_probability: 0.0
dataset:
    add_eos_token_to_answer: true
    add_eos_token_to_prompt: true
    add_eos_token_to_system: true
    answer_column: Assistentin
    chatbot_author: tmp-networks.de
    chatbot_name: EvaGPT-German
    data_sample: 1.0
    data_sample_choice:
    - Train
    - Validation
    limit_chained_samples: false
    mask_prompt_labels: true
    parent_id_column: parent_id
    personalize: true
    prompt_column:
    - Benutzer
    - Kontext
    system_column: system
    text_answer_separator: <|Assistentin|>
    text_prompt_start: <|Benutzer|>
    text_system_start: <|System|>
    train_dataframe: /media/mtsmash/gpt/h2o-llmstudio/data/user/dataset.13-sep/dataset.csv
    validation_dataframe: /media/mtsmash/gpt/h2o-llmstudio/data/user/dataset.13-sep/dataset.csv
    validation_size: 0.01
    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: 96
    seed: -1
    trust_remote_code: true
    use_deepspeed: false
experiment_name: EvaGPT-German-446M-Mini.v0.1.1
llm_backbone: /media/mtsmash/gpt/h2o-llmstudio/output/user/EvaGPT-German-446M-Mini.v0.1/
logging:
    logger: None
    neptune_project: ''
output_directory: /media/mtsmash/gpt/h2o-llmstudio/output/user/EvaGPT-German-446M-Mini.v0.1.1/
prediction:
    batch_size_inference: 0
    do_sample: true
    max_length_inference: 1312
    metric: Perplexity
    metric_gpt_model: gpt-3.5-turbo-0301
    metric_gpt_template: general
    min_length_inference: 2
    num_beams: 3
    num_history: 4
    repetition_penalty: 1.1
    stop_tokens: ''
    temperature: 0.7
    top_k: 40
    top_p: 1.0
problem_type: text_causal_language_modeling
tokenizer:
    add_prefix_space: false
    add_prompt_answer_tokens: true
    max_length: 4096
    max_length_answer: 3072
    max_length_prompt: 1024
    padding_quantile: 1.0
    use_fast: true
training:
    batch_size: 1
    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: 8
    gradient_clip: 1.0
    learning_rate: 0.0002
    lora: true
    lora_alpha: 46
    lora_dropout: 0.05
    lora_r: 26
    lora_target_modules: ''
    loss_function: TokenAveragedCrossEntropy
    optimizer: AdamW
    save_best_checkpoint: false
    schedule: Cosine
    train_validation_data: true
    use_flash_attention_2: false
    warmup_epochs: 0.0
    weight_decay: 0.01