LLaMA-Factory

LLaMA-Factory copied to clipboard

Qwen72B-Chat，预训练后，合并导出模型，通过Vllm运行报错

1

预训练是通过配置zero3： ``` { "fp16": { "enabled": "auto", "loss_scale": 0, "loss_scale_window": 1000, "initial_scale_power": 16, "hysteresis": 2, "min_loss_scale": 1 }, "bf16": { "enabled": "auto" }, "optimizer": { "type": "AdamW", "params": { "lr":...

Fire-Star

[Feature]后续会支持llava等多模态模型的训练吗

1

### Reminder - [X] I have read the README and searched the existing issues. ### Reproduction 如题 ### Expected behavior _No response_ ### System Info _No response_ ### Others _No...

Vincent131499

DPO训练Lora后，模型的生成结果是乱码

4

### Reminder - [X] I have read the README and searched the existing issues. ### Reproduction 您好，我通过SFT训练了一个能够正常使用的lora模型。现在想进一步通过DPO阶段的训练来优化lora模型的效果。但是我通过以下脚本训练后，输出的结果是乱码（随机重复的数字或字符串）。数据集我反复检查了是没有问题的。请问我是哪里出错了呢？ 另外，我的目的是继续训练Lora，训练的输出希望是优化后的lora模型。这个参数adapter_name_or_path 我看介绍说的是path to sft checkpoint. 那这里我应该放的是lora模型，还是将lora和base合并后的模型呢？ 非常感谢！ > CUDA_VISIBLE_DEVICES=0 deepspeed --num_gpus=1 /root/LLaMA-Factory/src/train_bash.py \...

IvoryTower800

### Reminder - [X] I have read the README and searched the existing issues. ### Reproduction 可以直接从web上选择checkpoint么，这样就可以随时abort，adapter可以直接选择checkpoint ### Expected behavior _No response_ ### System Info _No response_ ### Others _No...

tiger55cn

Brikarl

Bloom跑pt的full时，跑到60steps后，loss趋近于0

1

### Reminder - [X] I have read the README and searched the existing issues. ### Reproduction nohup deepspeed --num_gpus 8 --master_port=5544 src/train_bash.py \ --deepspeed ds_config.json \ --stage pt \ --do_train...

wqc007

### Reminder - [X] I have read the README and searched the existing issues. ### Reproduction #1882 When using the qwen model, flash-attn is automatically enabled. How can I confirm...

KelleyYin

PPO使用zero3加载全参训练的奖励模型，奖励模型加载失败。

12

PPO阶段使用zero2能正常开启训练，但是使用zero3就会出现奖励模型加载出错。 训练参数：  报错如下：  ds_config_zero3.config如下： 

Luoxiaohei41

Help: Need guidance on correctly saving and loading optimizer, scheduler, and scaler during training checkpoints

2

### Reminder - [X] I have read the README and searched the existing issues. ### Reproduction In my training script, I'm attempting to resume training from a checkpoint. However, upon...

GavinZhao19

### Reminder - [X] I have read the README and searched the existing issues. ### Reproduction ***** train metrics ***** epoch = 3.0 train_loss = 1.6394 train_runtime = 3:33:32.94 train_samples_per_second...

1Jenifer