chatglm-maths

chatglm-6b微调/LORA/PPO/推理, 样本为自动生成的整数/小数加减乘除运算, 可gpu/cpu

Github

https://github.com/yongzhuo/chatglm-maths

踩坑

1. eps=1e-5(不要改小), 半精度float16, 以及LN采用的是Post-LN(泛化性更好) + DeepNorm, 【害, Attention前也有LN】目的是大模型为了防止梯度溢出等;
2. 模型输入输出, 默认的tokenization_chatglm.py/modeling_chatglm.py不能用, 因为那是完全为生成generate设置的, 需要自己写好所有缩入参数, 或者机子改成适配的;
   2.1 ChatGLMModel中, get_masks()正常, get_position_ids()函数中‘context_length = seq.index(150004) + 1’ 改为 ‘context_length = len(seq)’;
   2.2 训练输入input_ids格式暂定为(训练后post-padding, 推理前pre-padding[tokenization_chatglm.py默认pre-padding])
       x: prompt_1 + "_" + text_1 + "\n" + prompt_2 + [gMASK] + [BOS] + "_" + text_2 + [PAD]*N
   2.3 训练输入label_ids格式暂定为(CrossEntropyLoss默认忽略-100不参与计算loss)  
       y = [-100]*len(text_1) + [BOS] + text_2 + [EOS] + [-100]*N
   2.4 注意position/mask(自带的只是推理用的batch_size=1, 所以训练输入还得自己写), 可参考GLM-130的README.md, huozhe 查看GLM-1源码https://github.com/THUDM/GLM/blob/main/tasks/seq2seq/dataset.py
3. 注意chatglm-6b权重是float16的, 不过计算loss时候会转成float32计算, 最后loss再转回float16更新梯度;
4. ChatGLMTokenizer有时候会报奇奇怪怪的错误, 建议生成时候设置max_new_tokens, 最大{"max_new_tokens": 2048}; decode有时候会出现不存在id;
5. 低秩自适应LORA, RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!
   尝试 transformers升级到最新, get_peft_model后再.cuda(), device_map={'':torch.cuda.current_device()},

微调数据

原始数据来自https://github.com/LYH-YF/MWPToolkit

处理后的微调数据(算式/解方程)-MWP: https://huggingface.co/datasets/Macropodus/MWP-Instruct
大数加减乘除来自: https://github.com/liutiedong/goat.git

LoRA权重

Baichuan-7B-GPT4ForALL: https://huggingface.co/Macropodus/MWP-Instruct
Bloomz-7B-GPT4ForALL: https://huggingface.co/Macropodus/MWP-Instruct
ChatGLM-6B-GPT4ForALL: https://huggingface.co/Macropodus/MWP-Instruct
LlaMA-7B-GPT4ForALL: https://huggingface.co/Macropodus/MWP-Instruct
ChatGLM-6B-MWP: https://huggingface.co/Macropodus/MWP-Instruct

数据集-中文

环境配置

transformers>=4.26.1
cpm_kernels==1.0.11
icetk==0.0.4
torch>=1.10.1
rouge==1.0.1
nltk==3.6.6
peft>=0.2.0
numpy
tqdm

lion_pytorch
macropodus
trl>=0.4.1

微调-计算题

lora
微调: python c00_toy_lora_train_6b.py
推理: python p00_toy_lora_predict_6b.py

ppo
训练: python t10_toy_trl_train_ppo.py
测试: python t10_toy_trl_predict_ppo.py

6b
微调: python c00_toy_cpu_train_6b.py
推理: python p00_toy_cpu_predit_6b.py

small-layer
微调: python c01_toy_cpu_train_small.py
推理: python p01_toy_cpu_predict_small.py

参考/感谢

推理日志toy

generator_calculate_line: ('13+75=', '13+75=88')
tokenizer.vocab_size: 150344
eval:   0%|                                                                                                                                                                      | 0/1 [00:00<?, ?it/s]batch_query: ['简便运算: 98+83= 剖析: 98+83=181']
batch_qtext_0: 简便运算: 98+83= 剖析:
batch_qans_0: 98+83=181
response_0: 98+83=171
{'rouge-1': 0.0, 'rouge-2': 0.0, 'rouge-l': 0.0, 'bleu': 0.0}
请输入:
25.31+86.35=
请稍等...
25.31+86.35=101.66

微调日志toy

generator_calculate_line: ('13+75=', '13+75=88')
tokenizer.vocab_size: 150344
Loading checkpoint shards: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 8/8 [00:10<00:00,  1.31s/it]
transformer.word_embeddings.weight False
......
transformer.layers.26.mlp.dense_4h_to_h.bias False
transformer.layers.27.input_layernorm.weight True
transformer.layers.27.input_layernorm.bias True
transformer.layers.27.attention.query_key_value.weight True
transformer.layers.27.attention.query_key_value.bias True
transformer.layers.27.attention.dense.weight True
transformer.layers.27.attention.dense.bias True
transformer.layers.27.post_attention_layernorm.weight True
transformer.layers.27.post_attention_layernorm.bias True
transformer.layers.27.mlp.dense_h_to_4h.weight True
transformer.layers.27.mlp.dense_h_to_4h.bias True
transformer.layers.27.mlp.dense_4h_to_h.weight True
transformer.layers.27.mlp.dense_4h_to_h.bias True
transformer.final_layernorm.weight True
transformer.final_layernorm.bias True
model.chat start
13+75=88, but that's not the correct answer. The correct answer is 13+75=88, which is 90.
/anaconda3/envs/py371/lib/python3.7/site-packages/transformers/optimization.py:395: FutureWarning: This implementation of AdamW is deprecated and will be removed in a future version. Use the PyTorch implementation torch.optim.AdamW instead, or set `no_deprecation_warning=True` to disable this warning
  FutureWarning,   
epoch:   0%|


---
license: cc-by-nc-4.0
---