{
    "project_name": "azma-hermes-pro-llama-3-8b-030524",
    "new_model_id": "azma-hermes-pro-llama-3-8b-030524",
    "base_model_id": "NousResearch/Hermes-2-Pro-Llama-3-8B",
    "adapter_id": "azma-hermes-pro-llama-3-8b-030524-adapter",
    "chat_template": "chatml",
    "lora_config": {
        "lora_rank": 16,
        "lora_alpha": 32,
        "lora_dropout": 0.1,
        "target_modules": [
            "k_proj",
            "v_proj",
            "o_proj",
            "q_proj",
            "up_proj",
            "gate_proj",
            "down_proj"
        ],
        "modules_to_save": [
            "lm_head"
        ],
        "task_type": "CAUSAL_LM"
    },
    "load_config": {
        "use_flash_attention": true,
        "load_in_4bit": false,
        "cache_dir": "./",
        "torch_dtype": "auto",
        "device_map": "auto",
        "pad_to_multiples": false
    },
    "dataset_config": {
        "dataset_id": [
            "Azma-AI/azma-mermaid-dataset-single-turn-chatml",
            "Azma-AI/azma-dataset-v2-mermaid-without-thoughts-final-chatml-8192-seq-len"
        ],
        "max_seq_length": 8192,
        "shuffle": true,
        "data_collator": "default",
        "template_version": 1.0,
        "dataset_text_field": "text",
        "add_custom_tokens": {
            "bos_token": "<|begin_of_text|>",
            "eos_token": "<|im_end|>",
            "pad_token": "<|end_of_text|>"
        }
    },
    "dpo_args": {
        "beta": 0.6,
        "loss_types": "sigmoid"
    },
    "training_args": {
        "seed": 42,
        "max_steps": -1,
        "weight_decay": 0.01,
        "num_train_epochs": 1,
        "learning_rate": 1e-05,
        "output_dir": "./results",
        "optim": "paged_adamw_32bit",
        "gradient_accumulation_steps": 2,
        "per_device_train_batch_size": 8,
        "per_device_eval_batch_size": 2,
        "group_by_length": false,
        "resume_from_checkpoint": true,
        "gradient_checkpointing": true,
        "gradient_checkpointing_kwargs": null,
        "do_eval": false,
        "eval_steps": 100,
        "evaluation_strategy": "steps",
        "save_steps": 100,
        "save_total_limit": 2,
        "save_strategy": "steps",
        "logging_steps": 5,
        "log_level": "info",
        "logging_strategy": "steps",
        "push_to_hub": true,
        "hub_model_id": "Azma-AI/azma-hermes-pro-llama-3-8b-030524-adapter",
        "hub_private_repo": true,
        "hub_strategy": "checkpoint",
        "report_to": "wandb",
        "warmup_steps": 5,
        "neftune_noise_alpha": 5,
        "lr_scheduler_type": "cosine",
        "auto_find_batch_size": true,
        "load_best_model_at_end": true,
        "deepspeed": null,
        "bf16": true,
        "fp16": false
    },
    "merge_final_model": true,
    "push_to_organization": true
}