README.md

---
library_name: transformers
pipeline_tag: image-text-to-text
tags:
- merlina
- grimoire
- image-text-to-text
- vision-language-model
- sft
datasets:
- hemlang/Hemlock2-DPO
- hemlang/hemlock-formulary-SFT
- hemlang/hemlock-codex-SFT
base_model:
- Jackrong/Qwopus3.5-9B-Coder
---

# Hemlock-Qwopus3.5-9B-Coder

## Training Configuration

| Parameter | Value |
|-----------|-------|
| Training Mode | SFT |
| Base Model | `Jackrong/Qwopus3.5-9B-Coder` |
| Learning Rate | 0.0002 |
| Epochs | 2 |
| Batch Size | 2 |
| Gradient Accumulation | 8 |
| Effective Batch Size | 16 |
| Max Sequence Length | 4096 |
| Optimizer | paged_adamw_8bit |
| LR Scheduler | cosine |
| Warmup Ratio | 0.05 |
| Weight Decay | 0.01 |
| Max Grad Norm | 1.0 |
| Seed | 42 |
| LoRA Rank (r) | 256 |
| LoRA Alpha | 256 |
| LoRA Dropout | 0.05 |
| Target Modules | k_proj, o_proj, q_proj, v_proj, down_proj, gate_proj, up_proj |
| Quantization | 4-bit (NF4) |
| GPU | NVIDIA RTX A6000 |

## Datasets

Trained on 3 concatenated datasets:

1. [`hemlang/Hemlock2-DPO`](https://huggingface.co/datasets/hemlang/Hemlock2-DPO) (split: `train`)
2. [`hemlang/hemlock-formulary-SFT`](https://huggingface.co/datasets/hemlang/hemlock-formulary-SFT) (split: `train`)
3. [`hemlang/hemlock-codex-SFT`](https://huggingface.co/datasets/hemlang/hemlock-codex-SFT) (split: `train`)

## Reproduce this training run

This model was trained with [Merlina](https://github.com/Schneewolf-Labs/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.

```json
{
  "_metadata": {
    "name": "Hemlock-Qwopus3.5-9B-Coder",
    "description": "Training configuration shared from a Merlina-trained model.",
    "tags": [],
    "schema": "merlina/training-config",
    "schema_version": 1,
    "merlina_version": "2.0.1"
  },
  "base_model": "Jackrong/Qwopus3.5-9B-Coder",
  "output_name": "Hemlock-Qwopus3.5-9B-Coder",
  "use_lora": true,
  "lora_r": 256,
  "lora_alpha": 256,
  "lora_dropout": 0.05,
  "target_modules": [
    "k_proj",
    "o_proj",
    "q_proj",
    "v_proj",
    "down_proj",
    "gate_proj",
    "up_proj"
  ],
  "modules_to_save": [],
  "lora_task_type": "CAUSAL_LM",
  "learning_rate": 0.0002,
  "num_epochs": 2,
  "batch_size": 2,
  "gradient_accumulation_steps": 8,
  "max_length": 4096,
  "max_prompt_length": 1024,
  "model_type": "auto",
  "training_mode": "sft",
  "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": 1024,
  "lora_rank": 32,
  "lora_target_modules": null,
  "lora_use_dora": false,
  "mid_training_samples": true,
  "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": null
    },
    "additional_sources": [
      {
        "source_type": "huggingface",
        "repo_id": "hemlang/hemlock-formulary-SFT",
        "split": "train",
        "file_path": null,
        "file_format": null,
        "dataset_id": null,
        "streaming": false,
        "streaming_batch_size": 10000,
        "column_mapping": {
          "instruction": "prompt",
          "output": "chosen"
        }
      },
      {
        "source_type": "huggingface",
        "repo_id": "hemlang/hemlock-codex-SFT",
        "split": "train",
        "file_path": null,
        "file_format": null,
        "dataset_id": null,
        "streaming": false,
        "streaming_batch_size": 10000,
        "column_mapping": {
          "instruction": "prompt",
          "output": "chosen"
        }
      }
    ],
    "format": {
      "format_type": "tokenizer",
      "custom_templates": null,
      "enable_thinking": true
    },
    "model_name": "Jackrong/Qwopus3.5-9B-Coder",
    "column_mapping": {
      "prompt": "prompt",
      "chosen": "chosen",
      "rejected": "rejected"
    },
    "convert_messages_format": true,
    "deduplicate": false,
    "dedupe_strategy": "prompt_chosen",
    "test_size": 0.01,
    "max_samples": null,
    "system_prompt": null,
    "system_prompt_mode": "fill_empty",
    "training_mode": "sft"
  },
  "seed": 42,
  "max_grad_norm": 1.0,
  "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": "sdpa",
  "use_liger": true,
  "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
}
```

---

![Trained with Merlina](https://raw.githubusercontent.com/Schneewolf-Labs/Merlina/refs/heads/main/frontend/madewithmerlina_smol.png)

[Merlina on GitHub](https://github.com/Schneewolf-Labs/Merlina)