---
license: apache-2.0
library_name: transformers
base_model:
- nbeerbower/Xiaolong-Qwen3-14B
datasets:
- jondurbin/gutenberg-dpo-v0.1
- nbeerbower/gutenberg2-dpo
- nbeerbower/gutenberg-moderne-dpo
- nbeerbower/synthetic-fiction-dpo
- nbeerbower/Arkhaios-DPO
- nbeerbower/Purpura-DPO
- nbeerbower/Schule-DPO
---

![image/png](https://huggingface.co/nbeerbower/Mistral-Nemo-Gutenberg-Encore-12B/resolve/main/encore_cover.png?download=true)

# Qwen3-Gutenberg-Encore-14B

[nbeerbower/Xiaolong-Qwen3-14B](https://huggingface.co/nbeerbower/Xiaolong-Qwen3-14B) finetuned on: 
* [jondurbin/gutenberg-dpo-v0.1](https://huggingface.co/datasets/jondurbin/gutenberg-dpo-v0.1)
* [nbeerbower/gutenberg2-dpo](https://huggingface.co/datasets/nbeerbower/gutenberg2-dpo)
* [nbeerbower/gutenberg-moderne-dpo](https://huggingface.co/datasets/nbeerbower/gutenberg-moderne-dpo)
* [nbeerbower/synthetic-fiction-dpo](https://huggingface.co/datasets/nbeerbower/synthetic-fiction-dpo)
* [nbeerbower/Arkhaios-DPO](https://huggingface.co/datasets/nbeerbower/Arkhaios-DPO)
* [nbeerbower/Purpura-DPO](https://huggingface.co/datasets/nbeerbower/Purpura-DPO)
* [nbeerbower/Schule-DPO](https://huggingface.co/datasets/nbeerbower/Schule-DPO)

## Method

[ORPO tuned](https://mlabonne.github.io/blog/posts/2024-04-19_Fine_tune_Llama_3_with_ORPO.html) with 1x RTX A6000 for 3 epochs.

### QLoRA config
```
# QLoRA config
bnb_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_compute_dtype=torch_dtype,
    bnb_4bit_use_double_quant=True,
)
# LoRA config
peft_config = LoraConfig(
    r=64,
    lora_alpha=128,
    lora_dropout=0.05,
    bias="none",
    task_type="CAUSAL_LM",
    target_modules=['up_proj', 'down_proj', 'gate_proj', 'k_proj', 'q_proj', 'v_proj', 'o_proj']
)
```

### ORPO config
```
orpo_args = ORPOConfig(
    learning_rate=8e-6,
    lr_scheduler_type="cosine",
    warmup_ratio=0.05,
    max_length=4096,
    max_prompt_length=1024,
    max_completion_length=4096,
    beta=0.1,
    per_device_train_batch_size=1,
    per_device_eval_batch_size=1,
    gradient_accumulation_steps=64,
    optim="paged_adamw_8bit",
    num_train_epochs=3,
    max_grad_norm=0.5,
    bf16=True,
)
```