423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-08-02 03:32:50 +08:00
Code Issues Packages Projects Releases Wiki Activity

[example] add bash usage (#7794)

Browse Source
This commit is contained in:
hoshi-hiyouga 2025-04-22 00:25:51 +08:00 committed by GitHub
parent d128382d3c
commit a62cba3d05
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
13 changed files with 184 additions and 98 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

25
README.md
View File

@ -77,7 +77,7 @@ Choose your path:
- **Various models**: LLaMA, LLaVA, Mistral, Mixtral-MoE, Qwen, Qwen2-VL, DeepSeek, Yi, Gemma, ChatGLM, Phi, etc.
- **Integrated methods**: (Continuous) pre-training, (multimodal) supervised fine-tuning, reward modeling, PPO, DPO, KTO, ORPO, etc.
- **Scalable resources**: 16-bit full-tuning, freeze-tuning, LoRA and 2/3/4/5/6/8-bit QLoRA via AQLM/AWQ/GPTQ/LLM.int8/HQQ/EETQ.
- **Advanced algorithms**: [Muon](https://github.com/KellerJordan/Muon), [GaLore](https://github.com/jiaweizzhao/GaLore), [BAdam](https://github.com/Ledzy/BAdam), [APOLLO](https://github.com/zhuhanqing/APOLLO), [Adam-mini](https://github.com/zyushun/Adam-mini), DoRA, LongLoRA, LLaMA Pro, Mixture-of-Depths, LoRA+, LoftQ and PiSSA.
- **Advanced algorithms**: [GaLore](https://github.com/jiaweizzhao/GaLore), [BAdam](https://github.com/Ledzy/BAdam), [APOLLO](https://github.com/zhuhanqing/APOLLO), [Adam-mini](https://github.com/zyushun/Adam-mini), [Muon](https://github.com/KellerJordan/Muon), DoRA, LongLoRA, LLaMA Pro, Mixture-of-Depths, LoRA+, LoftQ and PiSSA.
- **Practical tricks**: [FlashAttention-2](https://github.com/Dao-AILab/flash-attention), [Unsloth](https://github.com/unslothai/unsloth), [Liger Kernel](https://github.com/linkedin/Liger-Kernel), RoPE scaling, NEFTune and rsLoRA.
- **Wide tasks**: Multi-turn dialogue, tool using, image understanding, visual grounding, video recognition, audio understanding, etc.
- **Experiment monitors**: LlamaBoard, TensorBoard, Wandb, MLflow, [SwanLab](https://github.com/SwanHubX/SwanLab), etc.
@ -107,7 +107,9 @@ Compared to ChatGLM's [P-Tuning](https://github.com/THUDM/ChatGLM2-6B/tree/main/
## Changelog
[25/04/16] We supported **[Muon](https://github.com/KellerJordan/Muon)** optimizer. See [examples](examples/README.md) for usage. Thank [Juanxi Tian](https://tianshijing.github.io)'s PR.
[25/04/21] We supported the **[Muon](https://github.com/KellerJordan/Muon)** optimizer. See [examples](examples/README.md) for usage. Thank [@tianshijing](https://github.com/tianshijing)'s PR.
[25/04/16] We supported fine-tuning the **[InternVL3](https://huggingface.co/OpenGVLab/InternVL3-8B)** model. See [PR #7258](https://github.com/hiyouga/LLaMA-Factory/pull/7258) to get started.
[25/04/14] We supported fine-tuning the **[GLM-Z1](https://huggingface.co/THUDM/GLM-Z1-9B-0414)** and **[Kimi-VL](https://huggingface.co/moonshotai/Kimi-VL-A3B-Instruct)** models.
@ -115,14 +117,14 @@ Compared to ChatGLM's [P-Tuning](https://github.com/THUDM/ChatGLM2-6B/tree/main/
[25/03/31] We supported fine-tuning the **[Qwen2.5 Omni](https://qwenlm.github.io/blog/qwen2.5-omni/)** model. See [PR #7537](https://github.com/hiyouga/LLaMA-Factory/pull/7537) to get started.
<details><summary>Full Changelog</summary>
[25/03/15] We supported **[SGLang](https://github.com/sgl-project/sglang)** as inference backend. Try `infer_backend: sglang` to accelerate inference.
[25/03/12] We supported fine-tuning the **[Gemma 3](https://huggingface.co/blog/gemma3)** model.
[25/02/24] Announcing **[EasyR1](https://github.com/hiyouga/EasyR1)**, an efficient, scalable and multi-modality RL training framework for efficient GRPO training.
<details><summary>Full Changelog</summary>
[25/02/11] We supported saving the **[Ollama](https://github.com/ollama/ollama)** modelfile when exporting the model checkpoints. See [examples](examples/README.md) for usage.
[25/02/05] We supported fine-tuning the **[Qwen2-Audio](Qwen/Qwen2-Audio-7B-Instruct)** and **[MiniCPM-o-2.6](https://huggingface.co/openbmb/MiniCPM-o-2_6)** on audio understanding tasks.
@ -245,11 +247,11 @@ Compared to ChatGLM's [P-Tuning](https://github.com/THUDM/ChatGLM2-6B/tree/main/
| [Gemma 3](https://huggingface.co/google) | 1B/4B/12B/27B | gemma3/gemma (1B) |
| [GLM-4/GLM-4-0414/GLM-Z1](https://huggingface.co/THUDM) | 9B/32B | glm4 |
| [GPT-2](https://huggingface.co/openai-community) | 0.1B/0.4B/0.8B/1.5B | - |
| [Granite 3.0-3.1](https://huggingface.co/ibm-granite) | 1B/2B/3B/8B | granite3 |
| [Granite 3.0-3.3](https://huggingface.co/ibm-granite) | 1B/2B/3B/8B | granite3 |
| [Hunyuan](https://huggingface.co/tencent/) | 7B | hunyuan |
| [Index](https://huggingface.co/IndexTeam) | 1.9B | index |
| [InternLM 2-3](https://huggingface.co/internlm) | 7B/8B/20B | intern2 |
| [InternVL2_5-3](https://huggingface.co/OpenGVLab/InternVL) | 1B/2B/4B/8B/9B/14B/26B/38B/78B | intern_vl |
| [InternVL 2.5-3](https://huggingface.co/OpenGVLab)\*\* | 1B/2B/4B/8B/9B/14B/26B/38B/78B | intern_vl |
| [Kimi-VL](https://huggingface.co/moonshotai) | 16B | kimi_vl |
| [Llama](https://github.com/facebookresearch/llama) | 7B/13B/33B/65B | - |
| [Llama 2](https://huggingface.co/meta-llama) | 7B/13B/70B | llama2 |
@ -417,11 +419,11 @@ huggingface-cli login
| Mandatory | Minimum | Recommend |
| ------------ | ------- | --------- |
| python | 3.9 | 3.10 |
| torch | 1.13.1 | 2.6.0 |
| transformers | 4.41.2 | 4.50.0 |
| torch | 2.0.0 | 2.6.0 |
| transformers | 4.45.0 | 4.50.0 |
| datasets | 2.16.0 | 3.2.0 |
| accelerate | 0.34.0 | 1.2.1 |
| peft | 0.14.0 | 0.15.0 |
| peft | 0.14.0 | 0.15.1 |
| trl | 0.8.6 | 0.9.6 |
| Optional | Minimum | Recommend |
@ -430,7 +432,7 @@ huggingface-cli login
| deepspeed | 0.10.0 | 0.16.4 |
| bitsandbytes | 0.39.0 | 0.43.1 |
| vllm | 0.4.3 | 0.8.2 |
| flash-attn | 2.3.0 | 2.7.2 |
| flash-attn | 2.5.6 | 2.7.2 |
### Hardware Requirement
@ -458,7 +460,7 @@ cd LLaMA-Factory
pip install -e ".[torch,metrics]"
```
Extra dependencies available: torch, torch-npu, metrics, deepspeed, liger-kernel, bitsandbytes, hqq, eetq, gptq, awq, aqlm, vllm, sglang, muon, galore, apollo, badam, adam-mini, qwen, minicpm_v, modelscope, openmind, swanlab, quality
Extra dependencies available: torch, torch-npu, metrics, deepspeed, liger-kernel, bitsandbytes, hqq, eetq, gptq, awq, aqlm, vllm, sglang, galore, apollo, badam, adam-mini, qwen, minicpm_v, modelscope, openmind, swanlab, quality
> [!TIP]
> Use `pip install --no-deps -e .` to resolve package conflicts.
@ -519,6 +521,7 @@ source /usr/local/Ascend/ascend-toolkit/set_env.sh
| torch | 2.1.0 | 2.4.0 |
| torch-npu | 2.1.0 | 2.4.0.post2 |
| deepspeed | 0.13.2 | 0.13.2 |
| vllm-ascend | - | 0.7.3 |
Remember to use `ASCEND_RT_VISIBLE_DEVICES` instead of `CUDA_VISIBLE_DEVICES` to specify the device to use.

25
README_zh.md
View File

@ -80,7 +80,7 @@ https://github.com/user-attachments/assets/43b700c6-a178-41db-b1f8-8190a5d3fcfc
- **多种模型**：LLaMA、LLaVA、Mistral、Mixtral-MoE、Qwen、Qwen2-VL、DeepSeek、Yi、Gemma、ChatGLM、Phi 等等。
- **集成方法**：（增量）预训练、（多模态）指令监督微调、奖励模型训练、PPO 训练、DPO 训练、KTO 训练、ORPO 训练等等。
- **多种精度**：16 比特全参数微调、冻结微调、LoRA 微调和基于 AQLM/AWQ/GPTQ/LLM.int8/HQQ/EETQ 的 2/3/4/5/6/8 比特 QLoRA 微调。
- **先进算法**：[Muon](https://github.com/KellerJordan/Muon), [GaLore](https://github.com/jiaweizzhao/GaLore)、[BAdam](https://github.com/Ledzy/BAdam)、[APOLLO](https://github.com/zhuhanqing/APOLLO)、[Adam-mini](https://github.com/zyushun/Adam-mini)、DoRA、LongLoRA、LLaMA Pro、Mixture-of-Depths、LoRA+、LoftQ 和 PiSSA。
- **先进算法**：[GaLore](https://github.com/jiaweizzhao/GaLore)、[BAdam](https://github.com/Ledzy/BAdam)、[APOLLO](https://github.com/zhuhanqing/APOLLO)、[Adam-mini](https://github.com/zyushun/Adam-mini)、[Muon](https://github.com/KellerJordan/Muon)、DoRA、LongLoRA、LLaMA Pro、Mixture-of-Depths、LoRA+、LoftQ 和 PiSSA。
- **实用技巧**：[FlashAttention-2](https://github.com/Dao-AILab/flash-attention)、[Unsloth](https://github.com/unslothai/unsloth)、[Liger Kernel](https://github.com/linkedin/Liger-Kernel)、RoPE scaling、NEFTune 和 rsLoRA。
- **广泛任务**：多轮对话、工具调用、图像理解、视觉定位、视频识别和语音理解等等。
- **实验监控**：LlamaBoard、TensorBoard、Wandb、MLflow、[SwanLab](https://github.com/SwanHubX/SwanLab) 等等。
@ -110,7 +110,9 @@ https://github.com/user-attachments/assets/43b700c6-a178-41db-b1f8-8190a5d3fcfc
## 更新日志
[25/04/16] 我们支持了 **[Muon](https://github.com/KellerJordan/Muon)** 优化器。详细用法请参照 [examples](examples/README_zh.md)。感谢 [@Juanxi Tian](https://tianshijing.github.io) 的 PR。
[25/04/21] 我们支持了 **[Muon](https://github.com/KellerJordan/Muon)** 优化器。详细用法请参照 [examples](examples/README_zh.md)。感谢 [@tianshijing](https://github.com/tianshijing) 的 PR。
[25/04/16] 我们支持了 **[InternVL3](https://huggingface.co/OpenGVLab/InternVL3-8B)** 模型的微调。查看 [PR #7258](https://github.com/hiyouga/LLaMA-Factory/pull/7258) 以使用。
[25/04/14] 我们支持了 **[GLM-Z1](https://huggingface.co/THUDM/GLM-Z1-9B-0414)** 和 **[Kimi-VL](https://huggingface.co/moonshotai/Kimi-VL-A3B-Instruct)** 模型的微调。
@ -118,14 +120,14 @@ https://github.com/user-attachments/assets/43b700c6-a178-41db-b1f8-8190a5d3fcfc
[25/03/31] 我们支持了 **[Qwen2.5 Omni](https://qwenlm.github.io/blog/qwen2.5-omni/)** 模型的微调。查看 [PR #7537](https://github.com/hiyouga/LLaMA-Factory/pull/7537) 以使用。
<details><summary>展开日志</summary>
[25/03/15] 我们支持了 **[SGLang](https://github.com/sgl-project/sglang)** 推理后端，请使用 `infer_backend: sglang` 启用。
[25/03/12] 我们支持了 **[Gemma 3](https://huggingface.co/blog/gemma3)** 模型的微调。
[25/02/24] 我们宣布开源 **[EasyR1](https://github.com/hiyouga/EasyR1)**，一个高效可扩展的多模态强化学习框架，支持高效的 GRPO 训练。
<details><summary>展开日志</summary>
[25/02/11] 我们支持了在导出模型时保存 **[Ollama](https://github.com/ollama/ollama)** 配置文件。详细用法请参照 [examples](examples/README_zh.md)。
[25/02/05] 我们支持了在语音理解任务上微调 **[Qwen2-Audio](Qwen/Qwen2-Audio-7B-Instruct)** 和 **[MiniCPM-o-2.6](https://huggingface.co/openbmb/MiniCPM-o-2_6)** 模型。
@ -248,11 +250,11 @@ https://github.com/user-attachments/assets/43b700c6-a178-41db-b1f8-8190a5d3fcfc
| [Gemma 3](https://huggingface.co/google) | 1B/4B/12B/27B | gemma3/gemma (1B) |
| [GLM-4/GLM-4-0414/GLM-Z1](https://huggingface.co/THUDM) | 9B/32B | glm4 |
| [GPT-2](https://huggingface.co/openai-community) | 0.1B/0.4B/0.8B/1.5B | - |
| [Granite 3.0-3.1](https://huggingface.co/ibm-granite) | 1B/2B/3B/8B | granite3 |
| [Granite 3.0-3.3](https://huggingface.co/ibm-granite) | 1B/2B/3B/8B | granite3 |
| [Hunyuan](https://huggingface.co/tencent/) | 7B | hunyuan |
| [Index](https://huggingface.co/IndexTeam) | 1.9B | index |
| [InternLM 2-3](https://huggingface.co/internlm) | 7B/8B/20B | intern2 |
| [InternVL2_5-3](https://huggingface.co/OpenGVLab/InternVL) | 1B/2B/4B/8B/9B/14B/26B/38B/78B | intern_vl |
| [InternVL 2.5-3](https://huggingface.co/OpenGVLab)\*\* | 1B/2B/4B/8B/9B/14B/26B/38B/78B | intern_vl |
| [Kimi-VL](https://huggingface.co/moonshotai) | 16B | kimi_vl |
| [Llama](https://github.com/facebookresearch/llama) | 7B/13B/33B/65B | - |
| [Llama 2](https://huggingface.co/meta-llama) | 7B/13B/70B | llama2 |
@ -420,11 +422,11 @@ huggingface-cli login
| 必需项 | 至少 | 推荐 |
| ------------ | ------- | --------- |
| python | 3.9 | 3.10 |
| torch | 1.13.1 | 2.6.0 |
| transformers | 4.41.2 | 4.50.0 |
| torch | 2.0.0 | 2.6.0 |
| transformers | 4.45.0 | 4.50.0 |
| datasets | 2.16.0 | 3.2.0 |
| accelerate | 0.34.0 | 1.2.1 |
| peft | 0.14.0 | 0.15.0 |
| peft | 0.14.0 | 0.15.1 |
| trl | 0.8.6 | 0.9.6 |
| 可选项 | 至少 | 推荐 |
@ -433,7 +435,7 @@ huggingface-cli login
| deepspeed | 0.10.0 | 0.16.4 |
| bitsandbytes | 0.39.0 | 0.43.1 |
| vllm | 0.4.3 | 0.8.2 |
| flash-attn | 2.3.0 | 2.7.2 |
| flash-attn | 2.5.6 | 2.7.2 |
### 硬件依赖
@ -461,7 +463,7 @@ cd LLaMA-Factory
pip install -e ".[torch,metrics]"
```
可选的额外依赖项：torch、torch-npu、metrics、deepspeed、liger-kernel、bitsandbytes、hqq、eetq、gptq、awq、aqlm、vllm、sglang、muon, galore、apollo、badam、adam-mini、qwen、minicpm_v、modelscope、openmind、swanlab、quality
可选的额外依赖项：torch、torch-npu、metrics、deepspeed、liger-kernel、bitsandbytes、hqq、eetq、gptq、awq、aqlm、vllm、sglang、galore、apollo、badam、adam-mini、qwen、minicpm_v、modelscope、openmind、swanlab、quality
> [!TIP]
> 遇到包冲突时，可使用 `pip install --no-deps -e .` 解决。
@ -523,6 +525,7 @@ source /usr/local/Ascend/ascend-toolkit/set_env.sh
| torch | 2.1.0 | 2.4.0 |
| torch-npu | 2.1.0 | 2.4.0.post2 |
| deepspeed | 0.13.2 | 0.13.2 |
| vllm-ascend | - | 0.7.3 |
请使用 `ASCEND_RT_VISIBLE_DEVICES` 而非 `CUDA_VISIBLE_DEVICES` 来指定运算设备。

20
examples/README.md
View File

@ -24,7 +24,13 @@ llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml
Advanced usage:
```bash
CUDA_VISIBLE_DEVICES=0,1 llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml learning_rate=1e-5 logging_steps=1
CUDA_VISIBLE_DEVICES=0,1 llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml \
learning_rate=1e-5 \
logging_steps=1
```
```bash
bash examples/train_lora/llama3_lora_sft.sh
```
## Examples
@ -215,12 +221,6 @@ llamafactory-cli api examples/inference/llama3_lora_sft.yaml
### Extras
#### Full-Parameter Fine-Tuning using Muon
```bash
llamafactory-cli train examples/extras/muon/qwen2_full_sft.yaml
```
#### Full-Parameter Fine-Tuning using GaLore
```bash
@ -245,6 +245,12 @@ llamafactory-cli train examples/extras/badam/llama3_full_sft.yaml
llamafactory-cli train examples/extras/adam_mini/qwen2_full_sft.yaml
```
#### Full-Parameter Fine-Tuning using Muon
```bash
llamafactory-cli train examples/extras/muon/qwen2_full_sft.yaml
```
#### LoRA+ Fine-Tuning
```bash

20
examples/README_zh.md
View File

@ -24,7 +24,13 @@ llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml
高级用法：
```bash
CUDA_VISIBLE_DEVICES=0,1 llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml learning_rate=1e-5 logging_steps=1
CUDA_VISIBLE_DEVICES=0,1 llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml \
learning_rate=1e-5 \
logging_steps=1
```
```bash
bash examples/train_lora/llama3_lora_sft.sh
```
## 示例
@ -215,12 +221,6 @@ llamafactory-cli api examples/inference/llama3_lora_sft.yaml
### 杂项
#### 使用 Muon 进行全参数训练
```bash
llamafactory-cli train examples/extras/muon/qwen2_full_sft.yaml
```
#### 使用 GaLore 进行全参数训练
```bash
@ -245,6 +245,12 @@ llamafactory-cli train examples/extras/badam/llama3_full_sft.yaml
llamafactory-cli train examples/extras/adam_mini/qwen2_full_sft.yaml
```
#### 使用 Muon 进行全参数训练
```bash
llamafactory-cli train examples/extras/muon/qwen2_full_sft.yaml
```
#### LoRA+ 微调
```bash

36
examples/train_lora/llama3_lora_sft.sh Normal file
View File

@ -0,0 +1,36 @@
#!/bin/bash
set -x
MODEL_PATH=meta-llama/Meta-Llama-3-8B-Instruct
llamafactory-cli train \
--model_name_or_path ${MODEL_PATH} \
--trust_remote_code \
--stage sft \
--do_train \
--finetuning_type lora \
--lora_rank 8 \
--lora_target all \
--dataset identity,alpaca_en_demo \
--template llama3 \
--cutoff_len 2048 \
--max_samples 1000 \
--overwrite_cache \
--preprocessing_num_workers 16 \
--dataloader_num_workers 4 \
--output_dir saves/llama3-8b/lora/sft \
--logging_steps 10 \
--save_steps 500 \
--plot_loss \
--overwrite_output_dir \
--save_only_model false \
--report_to none \
--per_device_train_batch_size 1 \
--gradient_accumulation_steps 8 \
--learning_rate 1e-4 \
--num_train_epochs 3.0 \
--lr_scheduler_type cosine \
--warmup_ratio 0.1 \
--bf16 \
--ddp_timeout 180000000

20
src/llamafactory/hparams/model_args.py
View File

@ -65,14 +65,16 @@ class BaseModelArguments:
default=False,
metadata={"help": "Whether or not the special tokens should be split during the tokenization process."},
)
new_special_tokens: Optional[str] = field(
add_tokens: Optional[str] = field(
default=None,
metadata={
"help": "Non-special tokens to be added into the tokenizer. Use commas to separate multiple tokens."
},
)
add_special_tokens: Optional[str] = field(
default=None,
metadata={"help": "Special tokens to be added into the tokenizer. Use commas to separate multiple tokens."},
)
new_normal_tokens: Optional[str] = field(
default=None,
metadata={"help": "Normal tokens to be added into the tokenizer. Use commas to separate multiple tokens."},
)
model_revision: str = field(
default="main",
metadata={"help": "The specific model version to use (can be a branch name, tag name or commit id)."},
@ -180,11 +182,11 @@ class BaseModelArguments:
if self.adapter_name_or_path is not None: # support merging multiple lora weights
self.adapter_name_or_path = [path.strip() for path in self.adapter_name_or_path.split(",")]
if self.new_normal_tokens is not None: # support multiple normal tokens
self.new_normal_tokens = [token.strip() for token in self.new_normal_tokens.split(",")]
if self.add_tokens is not None: # support multiple tokens
self.add_tokens = [token.strip() for token in self.add_tokens.split(",")]
if self.new_special_tokens is not None: # support multiple special tokens
self.new_special_tokens = [token.strip() for token in self.new_special_tokens.split(",")]
if self.add_special_tokens is not None: # support multiple special tokens
self.add_special_tokens = [token.strip() for token in self.add_special_tokens.split(",")]
@dataclass

1
src/llamafactory/model/model_utils/quantization.py
View File

@ -124,6 +124,7 @@ def configure_quantization(
try:
from optimum.gptq import utils as gq_utils
if "language_model.model.layers" not in gq_utils.BLOCK_PATTERNS:
gq_utils.BLOCK_PATTERNS.insert(0, "language_model.model.layers")
except ImportError:

26
src/llamafactory/model/patcher.py
View File

@ -54,26 +54,22 @@ def patch_tokenizer(tokenizer: "PreTrainedTokenizer", model_args: "ModelArgument
if model_args.model_max_length is not None and tokenizer.model_max_length < model_args.model_max_length:
tokenizer.model_max_length = model_args.model_max_length # enlarge the tokenizer max length
if model_args.new_special_tokens is not None:
num_added_special_tokens = tokenizer.add_special_tokens(
dict(additional_special_tokens=model_args.new_special_tokens),
replace_additional_special_tokens=False,
if model_args.add_tokens is not None:
num_added_tokens = tokenizer.add_tokens(new_tokens=model_args.add_tokens, special_tokens=False)
logger.info_rank0("Add tokens {} to tokenizer's vocabulary.".format(",".join(model_args.add_tokens)))
if num_added_tokens > 0 and not model_args.resize_vocab:
model_args.resize_vocab = True
logger.warning_rank0("New tokens have been added, changed `resize_vocab` to True.")
if model_args.add_special_tokens is not None:
num_added_special_tokens = tokenizer.add_tokens(new_tokens=model_args.add_special_tokens, special_tokens=True)
logger.info_rank0(
"Add special tokens {} to tokenizer's vocabulary.".format(",".join(model_args.add_special_tokens))
)
logger.info_rank0("Add special tokens {} to vocab.".format(",".join(model_args.new_special_tokens)))
if num_added_special_tokens > 0 and not model_args.resize_vocab:
model_args.resize_vocab = True
logger.warning_rank0("New special tokens have been added, changed `resize_vocab` to True.")
if model_args.new_normal_tokens is not None:
num_added_normal_tokens = tokenizer.add_tokens(
new_tokens=model_args.new_normal_tokens,
special_tokens=False,
)
logger.info_rank0("Add normal tokens {} to vocab.".format(",".join(model_args.new_normal_tokens)))
if num_added_normal_tokens > 0 and not model_args.resize_vocab:
model_args.resize_vocab = True
logger.warning_rank0("New normal tokens have been added, changed `resize_vocab` to True.")
def patch_processor(
processor: "ProcessorMixin",

0
src/llamafactory/third_party/__init__.py vendored Normal file
View File

30
src/llamafactory/third_party/muon/muon.py vendored
View File

@ -2,6 +2,8 @@
#
# This code is based on the MoonshotAI's Moonlight library.
# https://github.com/MoonshotAI/Moonlight/blob/master/examples/toy_train.py
# and the Keller Jordan's Muon library.
# https://github.com/KellerJordan/Muon/blob/master/muon.py
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
@ -18,6 +20,7 @@
# MIT License
#
# Copyright (c) 2025 Moonshot AI
# Copyright (c) 2024 Keller Jordan
#
# Permission is hereby granted, free of charge, to any person obtaining a copy
# of this software and associated documentation files (the "Software"), to deal
@ -36,22 +39,20 @@
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
# SOFTWARE.
import math
import torch
# This code snippet is a modified version adapted from the following GitHub repository:
# https://github.com/KellerJordan/Muon/blob/master/muon.py
@torch.compile
def zeropower_via_newtonschulz5(G, steps):
def zeropower_via_newtonschulz5(G: "torch.Tensor", steps: int) -> "torch.Tensor":
"""Newton-Schulz iteration to compute the zeroth power / orthogonalization of G.
We opt to use a quintic iteration whose coefficients are selected to maximize the slope at zero.
For the purpose of minimizing steps, it turns out to be empirically effective to keep increasing the slope at
zero even beyond the point where the iteration no longer converges all the way to one everywhere
on the interval. This iteration therefore does not produce UV^T but rather something like US'V^T
where S' is diagonal with S_{ii}' ~ Uniform(0.5, 1.5), which turns out not to hurt model
For the purpose of minimizing steps, it turns out to be empirically effective to keep increasing
the slope at zero even beyond the point where the iteration no longer converges all the way to
one everywhere on the interval. This iteration therefore does not produce UV^T but rather something
like US'V^T where S' is diagonal with S_{ii}' ~ Uniform(0.5, 1.5), which turns out not to hurt model
performance at all relative to UV^T, where USV^T = G is the SVD.
"""
assert len(G.shape) == 2
@ -133,7 +134,7 @@ class Muon(torch.optim.Optimizer):
# Do not use Muon for parameters in adamw_params
self.state[p]["use_muon"] = False
def adjust_lr_for_muon(self, lr, param_shape):
def adjust_lr_for_muon(self, lr: float, param_shape: list[int]) -> float:
A, B = param_shape[:2]
# We adjust the learning rate and weight decay based on the size of the parameter matrix
# as describted in the paper
@ -154,12 +155,8 @@ class Muon(torch.optim.Optimizer):
loss = closure()
for group in self.param_groups:
############################
# Muon #
############################
# Muon loop
params = [p for p in group["params"] if self.state[p]["use_muon"]]
# import pdb; pdb.set_trace()
lr = group["lr"]
wd = group["wd"]
momentum = group["momentum"]
@ -195,10 +192,7 @@ class Muon(torch.optim.Optimizer):
# apply update
p.data.add_(u, alpha=-adjusted_lr)
############################
# AdamW backup #
############################
# Adam backup
params = [p for p in group["params"] if not self.state[p]["use_muon"]]
lr = group["lr"]
beta1, beta2 = group["adamw_betas"]

27
src/llamafactory/train/trainer_utils.py
View File

@ -489,16 +489,14 @@ def _create_adam_mini_optimizer(
logger.info_rank0("Using Adam-mini optimizer.")
return optimizer
def _create_muon_optimizer(
model: "PreTrainedModel",
training_args: "TrainingArguments",
) -> "torch.optim.Optimizer":
from llamafactory.third_party.muon import Muon # type: ignore
# Separate parameters for Muon (2D parameters) and AdamW (others)
muon_params = []
adamw_params = []
from ..third_party.muon import Muon
muon_params, adamw_params = [], []
for name, param in model.named_parameters():
if param.requires_grad:
# Use Muon for 2D parameters that aren't embeddings or heads
@ -507,33 +505,25 @@ def _create_muon_optimizer(
else:
adamw_params.append(param)
# Get optimizer settings from training_args
ns_steps = getattr(training_args, "ns_steps", 5)
# Create Muon optimizer
optimizer = Muon(
lr=training_args.learning_rate,
wd=training_args.weight_decay,
muon_params=muon_params,
momentum=0.95, # default momentum for Muon
nesterov=True, # default nesterov for Muon
ns_steps=ns_steps,
adamw_params=adamw_params,
adamw_betas=(training_args.adam_beta1, training_args.adam_beta2),
adamw_eps=training_args.adam_epsilon,
)
logger.info_rank0(f"Using Muon optimizer with {len(muon_params)} Muon params and {len(adamw_params)} AdamW params.")
logger.info_rank0(
f"Using Muon optimizer with {len(muon_params)} Muon params and {len(adamw_params)} AdamW params."
)
return optimizer
def create_custom_optimizer(
model: "PreTrainedModel",
training_args: "TrainingArguments",
finetuning_args: "FinetuningArguments",
) -> Optional["torch.optim.Optimizer"]:
if finetuning_args.use_muon:
return _create_muon_optimizer(model, training_args)
if finetuning_args.use_galore:
return _create_galore_optimizer(model, training_args, finetuning_args)
@ -549,6 +539,9 @@ def create_custom_optimizer(
if finetuning_args.use_adam_mini:
return _create_adam_mini_optimizer(model, training_args)
if finetuning_args.use_muon:
return _create_muon_optimizer(model, training_args)
def create_custom_scheduler(
training_args: "TrainingArguments",

46
tests/model/model_utils/test_add_tokens.py Normal file
View File

@ -0,0 +1,46 @@
# Copyright 2025 the LlamaFactory team.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import os
import pytest
from llamafactory.hparams import ModelArguments
from llamafactory.model import load_tokenizer
TINY_LLAMA3 = os.getenv("TINY_LLAMA3", "llamafactory/tiny-random-Llama-3")
UNUSED_TOKEN = "<|UNUSED_TOKEN|>"
@pytest.mark.parametrize("special_tokens", [False, True])
def test_add_tokens(special_tokens: bool):
if special_tokens:
model_args = ModelArguments(model_name_or_path=TINY_LLAMA3, add_special_tokens=UNUSED_TOKEN)
else:
model_args = ModelArguments(model_name_or_path=TINY_LLAMA3, add_tokens=UNUSED_TOKEN)
tokenizer = load_tokenizer(model_args)["tokenizer"]
encoded_ids = tokenizer.encode(UNUSED_TOKEN, add_special_tokens=False)
assert len(encoded_ids) == 1
decoded_str = tokenizer.decode(encoded_ids, skip_special_tokens=True)
if special_tokens:
assert decoded_str == ""
else:
assert decoded_str == UNUSED_TOKEN
if __name__ == "__main__":
pytest.main([__file__])