Hemlock-Philter-14B

Training Configuration

Parameter	Value
Training Mode	ORPO
Base Model	nbeerbower/Hemlock-Tincture-14B
Learning Rate	8e-06
Epochs	2
Batch Size	2
Gradient Accumulation	8
Effective Batch Size	16
Max Sequence Length	2048
Optimizer	paged_adamw_8bit
LR Scheduler	cosine
Warmup Ratio	0.05
Weight Decay	0.01
Max Grad Norm	0.3
Seed	42
Beta	0.1
Max Prompt Length	1024
LoRA Rank (r)	128
LoRA Alpha	128
LoRA Dropout	0.05
Target Modules	up_proj, down_proj, gate_proj, k_proj, q_proj, v_proj, o_proj
Quantization	4-bit (NF4)
GPU	NVIDIA RTX A6000

Reproduce this training run

This model was trained with Merlina. Save the JSON below to data/configs/<name>.json (or import it via the Load Configuration dialog) to reproduce the exact training setup. Credentials are not included — Merlina will use your own HF_TOKEN and WANDB_API_KEY from .env or the form.

{
  "_metadata": {
    "name": "Hemlock-Philter-14B",
    "description": "Training configuration shared from a Merlina-trained model.",
    "tags": [],
    "schema": "merlina/training-config",
    "schema_version": 1,
    "merlina_version": "2.0.3"
  },
  "base_model": "nbeerbower/Hemlock-Tincture-14B",
  "output_name": "Hemlock-Philter-14B",
  "use_lora": true,
  "lora_r": 128,
  "lora_alpha": 128,
  "lora_dropout": 0.05,
  "target_modules": [
    "up_proj",
    "down_proj",
    "gate_proj",
    "k_proj",
    "q_proj",
    "v_proj",
    "o_proj"
  ],
  "modules_to_save": [],
  "lora_task_type": "CAUSAL_LM",
  "learning_rate": 8e-06,
  "num_epochs": 2,
  "batch_size": 2,
  "gradient_accumulation_steps": 8,
  "max_length": 2048,
  "max_prompt_length": 1024,
  "model_type": "auto",
  "training_mode": "orpo",
  "beta": 0.1,
  "label_smoothing": 0.0,
  "gamma": 0.5,
  "vision_model_id": null,
  "stage": null,
  "unfreeze_vision_top_n": null,
  "image_token_id": null,
  "min_pixels": null,
  "max_pixels": null,
  "image_column": null,
  "caption_column": null,
  "instruction": null,
  "streaming": null,
  "model_name": null,
  "image_resolution": null,
  "lora_rank": null,
  "lora_target_modules": null,
  "lora_use_dora": null,
  "mid_training_samples": null,
  "dataset_jsonl_path": null,
  "dataset_name": null,
  "dataset_split": null,
  "sample_prompts": null,
  "sample_num_steps": null,
  "dataset": {
    "source": {
      "source_type": "huggingface",
      "repo_id": "hemlang/Hemlock2-DPO",
      "split": "train",
      "file_path": null,
      "file_format": null,
      "dataset_id": null,
      "streaming": false,
      "streaming_batch_size": 10000,
      "column_mapping": {}
    },
    "additional_sources": [],
    "format": {
      "format_type": "tokenizer",
      "custom_templates": null,
      "enable_thinking": false
    },
    "model_name": null,
    "column_mapping": {},
    "convert_messages_format": false,
    "deduplicate": false,
    "dedupe_strategy": "prompt_chosen",
    "test_size": 0.01,
    "max_samples": null,
    "system_prompt": null,
    "system_prompt_mode": "fill_empty",
    "training_mode": "orpo"
  },
  "seed": 42,
  "max_grad_norm": 0.3,
  "warmup_ratio": 0.05,
  "eval_steps": 0.2,
  "use_4bit": true,
  "use_wandb": true,
  "push_to_hub": true,
  "merge_lora_before_upload": true,
  "hf_hub_private": true,
  "export_gguf": false,
  "gguf_quant_types": [
    "Q4_K_M"
  ],
  "keep_gguf_fp16": false,
  "shuffle_dataset": true,
  "weight_decay": 0.01,
  "lr_scheduler_type": "cosine",
  "gradient_checkpointing": true,
  "logging_steps": 1,
  "optimizer_type": "paged_adamw_8bit",
  "adam_beta1": 0.9,
  "adam_beta2": 0.999,
  "adam_epsilon": 1e-08,
  "adafactor_relative_step": false,
  "adafactor_scale_parameter": false,
  "adafactor_warmup_init": false,
  "adafactor_decay_rate": -0.8,
  "adafactor_beta1": null,
  "adafactor_clip_threshold": 1.0,
  "attn_implementation": "auto",
  "use_liger": false,
  "torch_compile": false,
  "neftune_alpha": null,
  "eval_on_start": false,
  "gpu_ids": null,
  "multi_gpu_strategy": "auto",
  "wandb_project": null,
  "wandb_run_name": null,
  "wandb_tags": null,
  "wandb_notes": null
}

---

Merlina on GitHub