---
library_name: transformers
license: apache-2.0
base_model: mistralai/Mistral-Nemo-Instruct-2407
tags:
- roleplay
- conversational
- axolotl
---

# 残响 MN 12b (第一系列)

[English](./README.md) | [简体中文](./README-cn.md)

*空气中飘浮着一缕尘埃。它仿佛来自某个逝去的时代，但你无从追溯。它落在你的舌尖，滋味奇妙。*

![image/png](https://cdn-uploads.huggingface.co/production/uploads/634262af8d8089ebaefd410e/_ovgodU331FO4YAqFGCnk.png)

「残响」是一系列专注于SFW与NSFW角色扮演及对话的微调大语言模型。

## 量化版本
GGUF:
- 待补充！

EXL3:
- 待补充！

EXL2:
- 待补充！

其他格式:
- 待补充！

## 推荐配置
对话模板: Mistral v7 Tekken  
采样器:
视效果自行调整！

## 致谢
特别鸣谢 Allura 和 ilya <3  
衷心感谢 Axolotl（训练框架）、Mistral（基础模型）、Nebius（GPU资源）的开发者们，以及我的银行（感谢它提供的信用卡）

## 其他

[<img src="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/axolotl-ai-cloud/axolotl)
<details><summary>查看 axolotl 配置</summary>

axolotl 版本: `0.8.0.dev0`
```yaml
# === 模型配置 ===
base_model: mistralai/Mistral-Nemo-Instruct-2407  # 例如 "mistralai/Mistral-Small-24B-Instruct-2501"
load_in_8bit: false
load_in_4bit: false

# === 训练设置 ===
num_epochs: 2
micro_batch_size: 16
gradient_accumulation_steps: 1
sequence_len: 8192
sample_packing: true
pad_to_sequence_len: true

# === 超参数配置 ===
optimizer: apollo_adamw
# Apollo-mini 配置:
optim_args: "proj=random,rank=1,scale=128.0,scale_type=tensor,update_proj_gap=200"
# 标准 Apollo 配置:
# optim_args: 
optim_target_modules: all_linear
learning_rate: 1e-5
lr_scheduler: rex
weight_decay: 0.01
warmup_ratio: 0.05

# === 数据配置 ===
datasets:
  - path: allura-org/inkmix-v3.0
    type: chat_template
    split: train
    field_messages: conversations
    message_field_role: from
    message_field_content: value

dataset_prepared_path: last_run_prepared
chat_template: jinja
chat_template_jinja: |
  {{- bos_token }}{%- for message in messages %}
  {%- if message['role'] == 'system' %}
  {{- '[SYSTEM_PROMPT]' + message['content'] + '[/SYSTEM_PROMPT]' }}
  {%- elif message['role'] == 'user' %}
  {{- '[INST]' + message['content'] + '[/INST]' }}
  {%- elif message['role'] == 'assistant' %}
  {{- message['content'] + eos_token }}
  {%- endif %}
  {%- endfor %}

# === 插件 ===
plugins:
  - axolotl.integrations.liger.LigerPlugin
  - axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin

# === 硬件优化 ===
gradient_checkpointing: unsloth
gradient_checkpointing_kwargs:
  use_reentrant: false
liger_rope: true
liger_rms_norm: true
liger_glu_activation: true
cut_cross_entropy: true
torch_compile: true
# 多 GPU 时启用:
# deepspeed: [DEEPSPEED_CONFIG_PATH]  # 例如 "deepspeed_configs/zero3_bf16.json"

# === Wandb 追踪 ===
wandb_project: nemo12b-inkmix-v3

# === 检查点 ===
saves_per_epoch: 2
save_total_limit: 3

# === 高级设置 ===
output_dir: offload
bf16: auto
flash_attention: true
train_on_inputs: false
group_by_length: false
logging_steps: 1
trust_remote_code: true
# nemo 原生不支持系统提示
tokens:
  - "[SYSTEM_PROMPT]"
  - "[/SYSTEM_PROMPT]"
special_tokens:
  pad_token: "<pad>"

```

</details>