rm.json

{
    "train_dataset_settings": {
        "sources": [
                {
                "name": "train_preferences",
                  "records_path": "train_preferences.jsonl",
                  "num_samples": 10
              }
        ],
        "chat_settings":{
            "prompt_template": {
                "role_tag_mapping": {
                    "bot": "assistant",
                    "user": "user",
                    "system": "system"
                },
                "prefix_template": "<|start_header_id|>{role}<|end_header_id|>\n\n",
                "suffix_template": "<|eot_id|>"
            },
            "max_tokens_count": 2000
        },
        "add_labels": false,
        "dataset_type": "pair_preferences"
      },
      "val_dataset_settings": {
          "sources": [
              {
                "name": "val_preferences",
                  "records_path": "val_preferences.jsonl",
                  "num_samples": 5
              }
          ],
          "chat_settings":{
              "prompt_template": {
                "role_tag_mapping": {
                    "bot": "assistant",
                    "user": "user",
                    "system": "system"
                },
                "prefix_template": "<|start_header_id|>{role}<|end_header_id|>\n\n",
                "suffix_template": "<|eot_id|>"
            },
              "max_tokens_count": 2000
          },
          "add_labels": false,
          "dataset_type": "pair_preferences"
      },
      "cherry_pick_settings": {
        "dataset_settings":{
            "sources": [
                {
                    "name": "val_preferences",
                    "records_path": "val_preferences.jsonl",
                    "num_samples": 3
                }
            ],
            "chat_settings":{
                "prompt_template": {
                    "role_tag_mapping": {
                        "bot": "assistant",
                        "user": "user",
                        "system": "system"
                    },
                    "prefix_template": "<|start_header_id|>{role}<|end_header_id|>\n\n",
                    "suffix_template": "<|eot_id|>"
                },
                "max_tokens_count": 2000
            },
            "add_labels": false,
            "dataset_type": "pair_preferences"
          },
          "metric_settings": []
      },
      "model_settings": {
          "model_path": "yujiepan/llama-3-tiny-random",
          "model_type": "seq_cls",
          "transformers_settings": {},
          "model_kwargs": {
                "num_labels": 1
          }
      },
      "tokenizer_settings": {},
      "special_tokens_settings": {
        "bos_token": "<|begin_of_text|>",
        "eos_token": "<|end_of_text|>"
        },
      "trainer_settings": {
        "evaluation_strategy": "steps",
        "per_device_train_batch_size": 1,
        "per_device_eval_batch_size": 1,
        "gradient_accumulation_steps": 16,
        "load_best_model_at_end": false,
        "eval_steps": 32,
        "save_steps": 32,
        "logging_steps": 1,
        "learning_rate": 1e-6,
        "num_train_epochs": 2,
        "lr_scheduler_type": "cosine",
        "warmup_ratio": 0.3,
        "fp16": false,
        "bf16": true,
        "adam_beta1": 0.9,
        "adam_beta2": 0.95,
        "adam_epsilon": 1e-12,
        "max_grad_norm": 2.0,
        "weight_decay": 0.0,
        "optim": "adamw_torch",
        "save_total_limit": 16,
        "dataloader_num_workers": 18,
        "gradient_checkpointing": true,
        "save_only_model": true
      },
      "logging_settings": {
            "project_name": "alignment",
            "run_name": "rm",
            "entity": "turbo-alignment"
      },
      "seed": 0,
      "log_path": "train_output"
}