lyhue1991 / torchkeras Goto Github PK

When I'm using a multi-input model (such as aTransformer model), I get errors like TypeError: forward() missing 2 required positional arguments: 'tgt' and 'tgt_mask'.
I looked at the source code of torchkeras and thought that there was a problem with the way the parameters of the torchkeras.torchkeras.Model.forward() function were passed.

So I suggest modifying the torchkeras.torchkeras.Model.forward() function like this:

Change:

To:

    def forward(self, *x):  #Attention this line 
        if self.net:
            return self.net.forward(*x)  #Attention this line 
        else:
            raise NotImplementedError

Also, the place where the torchkeras.torchkeras.Model.forward() function is used should also be modified, as follows:

Change:

To:

        predictions = self.forward(*features)

and many more.

If my suggestion is wrong, please ignore my suggestion.

代码：官方给的实例
https://www.kaggle.com/code/lyhue1991/torchkeras-ddp-tpu-examples/notebook

运行环境：
pycharm 上的 run 按钮

报错信息：

我在多gpu 运行您的examples/ChatGLM2——transformers.ipynb 的训练脚本，报 如下错误：
RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:3 and cuda:0!
(when checking argument for argument target in method wrapper_CUDA_nll_loss_forward)
in :1 │
│ │
│ ❱ 1 keras_model.fit(train_data = dl_train, │
│ 2 │ │ │ │ val_data = dl_val, │
│ 3 │ │ │ │ epochs=100,patience=5, │
│ 4 │ │ │ │ monitor='val_loss',mode='min', │
│ │
│ /u01/liuys/anaconda3/envs/chatglm/lib/python3.11/site-packages/torchkeras/kerasmodel.py:204 in │
│ fit │
│ │
│ 201 │ │ │ │
│ 202 │ │ │ train_epoch_runner = self.EpochRunner(train_step_runner,should_quiet) │
│ 203 │ │ │ train_metrics = {'epoch':epoch} │
│ ❱ 204 │ │ │ train_metrics.update(train_epoch_runner(train_dataloader)) │
│ 205 │ │ │ │
│ 206 │ │ │ for name, metric in train_metrics.items(): │
│ 207 │ │ │ │ self.history[name] = self.history.get(name, []) + [metric] │
│ │
│ /u01/liuys/anaconda3/envs/chatglm/lib/python3.11/site-packages/torchkeras/kerasmodel.py:77 in │
│ call │
│ │
│ 74 │ │ │
│ 75 │ │ for step, batch in loop: │
│ 76 │ │ │ with self.accelerator.accumulate(self.net): │
│ ❱ 77 │ │ │ │ step_losses,step_metrics = self.steprunner(batch) │
│ 78 │ │ │ │ step_log = dict(step_losses,**step_metrics) │
│ 79 │ │ │ │ for k,v in step_losses.items(): │
│ 80 │ │ │ │ │ epoch_losses[k] = epoch_losses.get(k,0.0)+v │
│ │
│ in call:20 │
│ │
│ 17 │ │ │
│ 18 │ │ #loss │
│ 19 │ │ with self.accelerator.autocast(): │
│ ❱ 20 │ │ │ loss = self.net(input_ids=batch["input_ids"],labels=batch["labels"]).loss │
│ 21 │ │ │
│ 22 │ │ #backward() │
│ 23 │ │ if self.optimizer is not None and self.stage=="train": │
│ │
│ /u01/liuys/anaconda3/envs/chatglm/lib/python3.11/site-packages/torch/nn/modules/module.py:1501 │
│ in _call_impl │
│ │
│ 1498 │ │ if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks │
│ 1499 │ │ │ │ or _global_backward_pre_hooks or _global_backward_hooks │
│ 1500 │ │ │ │ or _global_forward_hooks or _global_forward_pre_hooks): │
│ ❱ 1501 │ │ │ return forward_call(*args, **kwargs) │
│ 1502 │ │ # Do not call functions when jit is used │
│ 1503 │ │ full_backward_hooks, non_full_backward_hooks = [], [] │
│ 1504 │ │ backward_pre_hooks = [] │
│ │
│ /u01/liuys/anaconda3/envs/chatglm/lib/python3.11/site-packages/accelerate/utils/operations.py:52 │
│ 1 in forward │
│ │
│ 518 │ model_forward = ConvertOutputsToFp32(model_forward) │
│ 519 │ │
│ 520 │ def forward(*args, **kwargs): │
│ ❱ 521 │ │ return model_forward(*args, **kwargs) │
│ 522 │ │
│ 523 │ # To act like a decorator so that it can be popped when doing `extract_model_from_pa │
│ 524 │ forward.wrapped = model_forward │
│ │
│ /u01/liuys/anaconda3/envs/chatglm/lib/python3.11/site-packages/accelerate/utils/operations.py:50 │
│ 9 in call │
│ │
│ 506 │ │ update_wrapper(self, model_forward) │
│ 507 │ │
│ 508 │ def call(self, *args, **kwargs): │
│ ❱ 509 │ │ return convert_to_fp32(self.model_forward(*args, **kwargs)) │
│ 510 │ │
│ 511 │ def getstate(self): │
│ 512 │ │ raise pickle.PicklingError( │
│ │
│ /u01/liuys/anaconda3/envs/chatglm/lib/python3.11/site-packages/torch/amp/autocast_mode.py:14 in │
│ decorate_autocast │
│ │
│ 11 │ @functools.wraps(func) │
│ 12 │ def decorate_autocast(*args, **kwargs): │
│ 13 │ │ with autocast_instance: │
│ ❱ 14 │ │ │ return func(*args, **kwargs) │
│ 15 │ decorate_autocast.__script_unsupported = '@autocast() decorator is not supported in │
│ 16 │ return decorate_autocast │
│ 17 │
│ │
│ /u01/liuys/anaconda3/envs/chatglm/lib/python3.11/site-packages/peft/peft_model.py:686 in forward │
│ │
│ 683 │ ): │
│ 684 │ │ peft_config = self.active_peft_config │
│ 685 │ │ if not isinstance(peft_config, PromptLearningConfig): │
│ ❱ 686 │ │ │ return self.base_model( │
│ 687 │ │ │ │ input_ids=input_ids, │
│ 688 │ │ │ │ attention_mask=attention_mask, │
│ 689 │ │ │ │ inputs_embeds=inputs_embeds, │
│ │
│ /u01/liuys/anaconda3/envs/chatglm/lib/python3.11/site-packages/torch/nn/modules/module.py:1501 │
│ in _call_impl │
│ │
│ 1498 │ │ if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks │
│ 1499 │ │ │ │ or _global_backward_pre_hooks or _global_backward_hooks │
│ 1500 │ │ │ │ or _global_forward_hooks or _global_forward_pre_hooks): │
│ ❱ 1501 │ │ │ return forward_call(*args, **kwargs) │
│ 1502 │ │ # Do not call functions when jit is used │
│ 1503 │ │ full_backward_hooks, non_full_backward_hooks = [], [] │
│ 1504 │ │ backward_pre_hooks = [] │
│ │
│ /u01/liuys/anaconda3/envs/chatglm/lib/python3.11/site-packages/accelerate/hooks.py:165 in │
│ new_forward │
│ │
│ 162 │ │ │ with torch.no_grad(): │
│ 163 │ │ │ │ output = old_forward(*args, **kwargs) │
│ 164 │ │ else: │
│ ❱ 165 │ │ │ output = old_forward(*args, **kwargs) │
│ 166 │ │ return module._hf_hook.post_forward(module, output) │
│ 167 │ │
│ 168 │ module.forward = new_forward │
│ │
│ /u01/liuys/.cache/huggingface/modules/transformers_modules/THUDM/chatglm2-6b/c57e892806dfe383cd5 │
│ caf09719628788fe96379/modeling_chatglm.py:957 in forward │
│ │
│ 954 │ │ │ shift_labels = labels[..., 1:].contiguous() │
│ 955 │ │ │ # Flatten the tokens │
│ 956 │ │ │ loss_fct = CrossEntropyLoss(ignore_index=-100) │
│ ❱ 957 │ │ │ loss = loss_fct(shift_logits.view(-1, shift_logits.size(-1)), shift_labels.v │
│ 958 │ │ │ │
│ 959 │ │ │ lm_logits = lm_logits.to(hidden_states.dtype) │
│ 960 │ │ │ loss = loss.to(hidden_states.dtype) │
│ │
│ /u01/liuys/anaconda3/envs/chatglm/lib/python3.11/site-packages/torch/nn/modules/module.py:1501 │
│ in _call_impl │
│ │
│ 1498 │ │ if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks │
│ 1499 │ │ │ │ or _global_backward_pre_hooks or _global_backward_hooks │
│ 1500 │ │ │ │ or _global_forward_hooks or _global_forward_pre_hooks): │
│ ❱ 1501 │ │ │ return forward_call(*args, **kwargs) │
│ 1502 │ │ # Do not call functions when jit is used │
│ 1503 │ │ full_backward_hooks, non_full_backward_hooks = [], [] │
│ 1504 │ │ backward_pre_hooks = [] │
│ │
│ /u01/liuys/anaconda3/envs/chatglm/lib/python3.11/site-packages/torch/nn/modules/loss.py:1174 in │
│ forward │
│ │
│ 1171 │ │ self.label_smoothing = label_smoothing │
│ 1172 │ │
│ 1173 │ def forward(self, input: Tensor, target: Tensor) -> Tensor: │
│ ❱ 1174 │ │ return F.cross_entropy(input, target, weight=self.weight, │
│ 1175 │ │ │ │ │ │ │ ignore_index=self.ignore_index, reduction=self.reduction, │
│ 1176 │ │ │ │ │ │ │ label_smoothing=self.label_smoothing) │
│ 1177 │
│ │
│ /u01/liuys/anaconda3/envs/chatglm/lib/python3.11/site-packages/torch/nn/functional.py:3029 in │
│ cross_entropy │
│ │
│ 3026 │ │ ) │
│ 3027 │ if size_average is not None or reduce is not None: │
│ 3028 │ │ reduction = _Reduction.legacy_get_string(size_average, reduce) │
│ ❱ 3029 │ return torch._C._nn.cross_entropy_loss(input, target, weight, _Reduction.get_enum(re │
│ 3030 │
│ 3031 │
│ 3032 def binary_cross_entropy(

Description:
链接失效:《Kaggle免费GPU使用攻略 https://www.bilibili.com/video/BV1oa411u7uR/%E3%80%8B

大家有遇到同样情况的嘛，感觉是loss fn出了问题

ValueError: Can't find 'adapter_config.json' at '/data/jupyter/root/torchkeras/examples/single_chatglm2'

代码：官方给的实例
https://www.kaggle.com/code/lyhue1991/torchkeras-ddp-tpu-examples/notebook

运行环境：
pycharm 上的 run 按钮

报错信息：

怎样使用gpu

where is this dataset "dfner_13k.pkl"

self.accelerator.backward(loss)
RuntimeError: Expected is_sm80 to be true, but got false.
accelerate=0.21 accelerate=0.20 都不行

hi, i get a problem when i use torch.summary function to print the multi-input model structure, i don't know how to pass parameters to "input_shape", i tried tuple like ((10, ), (10, 1)) and list like [(10, ), (10, )], can you help me?thanks in advance.

need the latest version

TypeError: ChatGLMForConditionalGeneration.stream_chat() takes from 3 to 9 positional arguments but 11 were given

多训练几次就没啥问题了！

TypeError: transformers.tokenization_utils.PreTrainedTokenizer.init() got multiple values for keyword argument 'clean_up_tokenization_spaces'

TypeError Traceback (most recent call last)
Cell In[13], line 3
1 from transformers import AutoModel,AutoTokenizer
2 model_name = "/mnt/workspace/chatglm2-6b-AdaLoRA/chatglm2-6b-梦中情炉"
----> 3 tokenizer = AutoTokenizer.from_pretrained(
4 model_name, trust_remote_code=True)
5 model = AutoModel.from_pretrained(model_name,
6 trust_remote_code=True).half().cuda()

File /home/pai/lib/python3.9/site-packages/transformers/models/auto/tokenization_auto.py:679, in AutoTokenizer.from_pretrained(cls, pretrained_model_name_or_path, *inputs, **kwargs)
675 if tokenizer_class is None:
676 raise ValueError(
677 f"Tokenizer class {tokenizer_class_candidate} does not exist or is not currently imported."
678 )
--> 679 return tokenizer_class.from_pretrained(pretrained_model_name_or_path, *inputs, **kwargs)
681 # Otherwise we have to be creative.
682 # if model is an encoder decoder, the encoder tokenizer class is used by default
683 if isinstance(config, EncoderDecoderConfig):

File /home/pai/lib/python3.9/site-packages/transformers/tokenization_utils_base.py:1804, in PreTrainedTokenizerBase.from_pretrained(cls, pretrained_model_name_or_path, *init_inputs, **kwargs)
1801 else:
1802 logger.info(f"loading file {file_path} from cache at {resolved_vocab_files[file_id]}")
-> 1804 return cls._from_pretrained(
1805 resolved_vocab_files,
1806 pretrained_model_name_or_path,
1807 init_configuration,
1808 *init_inputs,
1809 use_auth_token=use_auth_token,
1810 cache_dir=cache_dir,
1811 local_files_only=local_files_only,
1812 _commit_hash=commit_hash,
1813 **kwargs,
1814 )

File /home/pai/lib/python3.9/site-packages/transformers/tokenization_utils_base.py:1958, in PreTrainedTokenizerBase._from_pretrained(cls, resolved_vocab_files, pretrained_model_name_or_path, init_configuration, use_auth_token, cache_dir, local_files_only, _commit_hash, *init_inputs, **kwargs)
1956 # Instantiate tokenizer.
1957 try:
-> 1958 tokenizer = cls(*init_inputs, **init_kwargs)
1959 except OSError:
1960 raise OSError(
1961 "Unable to load vocabulary from file. "
1962 "Please check that the provided vocabulary is accessible and not corrupted."
1963 )

File ~/.cache/huggingface/modules/transformers_modules/chatglm2-6b-梦中情炉/tokenization_chatglm.py:69, in ChatGLMTokenizer.init(self, vocab_file, padding_side, **kwargs)
68 def init(self, vocab_file, padding_side="left", **kwargs):
---> 69 super().init(padding_side=padding_side, clean_up_tokenization_spaces=False, **kwargs)
70 self.name = "GLMTokenizer"
72 self.vocab_file = vocab_file

TypeError: transformers.tokenization_utils.PreTrainedTokenizer.init() got multiple values for keyword argument 'clean_up_tokenization_spaces'

================================================================================2023-07-06 14:03:03
Epoch 1 / 100

╭─────────────────────────────── Traceback (most recent call last) ────────────────────────────────╮
│ /torchkeras/train.py:230 in │
│ │
│ 227 │ │ optimizer=torch.optim.AdamW(model.parameters(),lr=2e-6)) │
│ 228 ckpt_path = './waimai_chatglm4' │
│ 229 │
│ ❱ 230 keras_model.fit(train_data = dl_train, │
│ 231 │ │ │ │ val_data = dl_val, │
│ 232 │ │ │ │ epochs=100,patience=5, │
│ 233 │ │ │ │ monitor='val_loss',mode='min', │
│ │
│ /torchkeras/torchkeras/kerasmodel.py:204 in fit │
│ │
│ 201 │ │ │ │ │ loss_fn = self.loss_fn, │
│ 202 │ │ │ │ │ accelerator = self.accelerator, │
│ 203 │ │ │ │ │ stage="train", │
│ ❱ 204 │ │ │ │ │ metrics_dict=deepcopy(self.metrics_dict), │
│ 205 │ │ │ │ │ optimizer = self.optimizer if epoch>0 else None, │
│ 206 │ │ │ │ │ lr_scheduler = self.lr_scheduler if epoch>0 else None │
│ 207 │ │ │ ) │
│ │
│ /usr/lib/python3.8/copy.py:172 in deepcopy │
│ │
│ 169 │ │ │ │ if isinstance(rv, str): │
│ 170 │ │ │ │ │ y = x │
│ 171 │ │ │ │ else: │
│ ❱ 172 │ │ │ │ │ y = _reconstruct(x, memo, *rv) │
│ 173 │ │
│ 174 │ # If is its own copy, don't memoize. │
│ 175 │ if y is not x: │
│ │
│ /usr/lib/python3.8/copy.py:270 in _reconstruct │
│ │
│ 267 │ │
│ 268 │ if state is not None: │
│ 269 │ │ if deep: │
│ ❱ 270 │ │ │ state = deepcopy(state, memo) │
│ 271 │ │ if hasattr(y, 'setstate'): │
│ 272 │ │ │ y.setstate(state) │
│ 273 │ │ else: │
│ │
│ /usr/lib/python3.8/copy.py:146 in deepcopy │
│ │
│ 143 │ │
│ 144 │ copier = _deepcopy_dispatch.get(cls) │
│ 145 │ if copier is not None: │
│ ❱ 146 │ │ y = copier(x, memo) │
│ 147 │ else: │
│ 148 │ │ if issubclass(cls, type): │
│ 149 │ │ │ y = _deepcopy_atomic(x, memo) │
│ │
│ /usr/lib/python3.8/copy.py:230 in _deepcopy_dict │
│ │
│ 227 │ y = {} │
│ 228 │ memo[id(x)] = y │
│ 229 │ for key, value in x.items(): │
│ ❱ 230 │ │ y[deepcopy(key, memo)] = deepcopy(value, memo) │
│ 231 │ return y │
│ 232 d[dict] = _deepcopy_dict │
│ 233 if PyStringMap is not None: │
│ │
│ /usr/lib/python3.8/copy.py:161 in deepcopy │
│ │
│ 158 │ │ │ │ else: │
│ 159 │ │ │ │ │ reductor = getattr(x, "reduce_ex", None) │
│ 160 │ │ │ │ │ if reductor is not None: │
│ ❱ 161 │ │ │ │ │ │ rv = reductor(4) │
│ 162 │ │ │ │ │ else: │
│ 163 │ │ │ │ │ │ reductor = getattr(x, "reduce", None) │
│ 164 │ │ │ │ │ │ if reductor: │
│ │
│ /usr/local/lib/python3.8/dist-packages/accelerate/utils/operations.py:498 in getstate │
│ │
│ 495 │ │ return convert_to_fp32(self.model_forward(*args, **kwargs)) │
│ 496 │ │
│ 497 │ def getstate(self): │
│ ❱ 498 │ │ raise pickle.PicklingError( │
│ 499 │ │ │ "Cannot pickle a prepared model with automatic mixed precision, please unwra │
│ 500 │ │ ) │
│ 501 │
╰──────────────────────────────────────────────────────────────────────────────────────────────────╯
PicklingError: Cannot pickle a prepared model with automatic mixed precision, please unwrap the model with Accelerator.unwrap_model(model) before pickling it.

当我开始训练的时候报这个错误，能帮忙解决吗？

Very impressive job! After the Google Colab test, need to install one more package to run perfectly.

when I use model = nn.DataParallel(model, device_ids=model_device_ids), I got this Error:torch.nn.modules.module.ModuleAttributeError: 'DataParallel' object has no attribute 'compile'
can you fix this problem, Thank you very much!!

AttributeError: 'ChatGLMTokenizer' object has no attribute 'build_prompt'

是否能支持多卡训练

i set device='cuda:0' in "model.compile()", but the "model.fit()" will get a error: "RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:1 and cpu!" . I don't know why. when i don't set device , it will run on cpu correctly, but i want to run the code on gpu so i set the param "device", why it gets wrong?...

报错信息如下：

版本3.8.0，import torchkeras时总是出现这个，如果直接import torchkeras.LightModel也会报错，看了一下LightModel定义在lightmodel.py中，但为什么import不成功呢？

代码和数据都没改过，看起来没有要收敛的意思，求大神指点

我魔改了一下，不支持呢

您好，大佬。
BaiChuan13B_NER——transformers模型微调按照您的文件运行了一遍最后跑出来的结果变小，没有更改过文件，这是否是需要增加轮次或者怎么处理呢？和您的结果对不上

跑出来的结果：

这个它没报错误原因，看不出来咋回事，大佬可以协助解决嘛？

3.9.2版本的torchkeras中运行from torchkeras import LightModel 会出现cannot import name 'LightModel' from 'torchkeras'，之前版本是可以的，最新版LightModel被去除了吗？

提示torchkeras No module named 'accelerate'
pip安装accelerate后
提示AttributeError: module 'signal' has no attribute 'SIGKILL'

我的环境是window，py3.9.13，torchkeras3.2.5

不知道是peft的哪个版本才有 AdaLoraConfig，我用的是0.3.0
ImportError: cannot import name 'AdaLoraConfig' from 'peft'

您好，自定义loss需要传入其他参数，但是StepRunner中self.loss_fn(preds,labels)只接受预测和标签，除了修改源码外，还有其他更优雅的方法吗

python 3.7.6
torchkeras 3.8.8
是版本不对吗？

执行
ds_train = ds_train_raw.map(
preprocess,
batched=True,
num_proc=4,
remove_columns=ds_train_raw.column_names
)

ds_val = ds_val_raw.map(
preprocess,
batched=True,
num_proc=4,
remove_columns=ds_val_raw.column_names
)
报错
AttributeError: 'ChatGLMTokenizer' object has no attribute 'build_prompt'

`from model import LeNet
import torchkeras
import torchmetrics
from torchvision import datasets
import torch.nn as nn
import torch
import torchvision.transforms as transforms
import torchvision

transform = transforms.Compose(
[transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])

batch_size = 4
net = LeNet()
model = torchkeras.KerasModel(net,
loss_fn = nn.BCEWithLogitsLoss(),
optimizer= torch.optim.Adam(net.parameters(),lr = 1e-4),
metrics_dict = {"acc":torchmetrics.Accuracy(task='binary')}
)

trainset = torchvision.datasets.CIFAR10(root='../../../../Dataset', train=True,
download=False, transform=transform)
trainloader = torch.utils.data.DataLoader(trainset, batch_size=batch_size,
shuffle=True, num_workers=2)

testset = torchvision.datasets.CIFAR10(root='../../../../Dataset', train=False,
download=False, transform=transform)
testloader = torch.utils.data.DataLoader(testset, batch_size=batch_size,
shuffle=False, num_workers=2)

dfhistory=model.fit(train_data=trainloader,
val_data=testloader,
epochs=20,
patience=3,
ckpt_path='checkpoint.pt',
monitor="val_acc",
mode="max",
plot=True,
)`

# 通过注册jupyter魔法命令可以很方便地在jupyter中测试ChatGLM 
from torchkeras.chat import ChatGLM 
chatglm = ChatGLM(model, tokenizer)

register magic %%chatglm sucessed ...
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[5], line 3
      1 # 通过注册jupyter魔法命令可以很方便地在jupyter中测试ChatGLM 
      2 from torchkeras.chat import ChatGLM 
----> 3 chatglm = ChatGLM(model, tokenizer)

File ~/anaconda3/envs/zdw/lib/python3.10/site-packages/torchkeras/chat/chatglm.py:27, in ChatGLM.__init__(self, model, tokenizer, stream, max_chat_rounds, history, max_length, num_beams, do_sample, top_p, temperature, logits_processor)
     24     print('register magic %%chatglm failed ...')
     25     print(err)
---> 27 response = self('你好')
     28 if not self.stream:
     29     print(response)

File ~/anaconda3/envs/zdw/lib/python3.10/site-packages/torchkeras/chat/chatglm.py:50, in ChatGLM.__call__(self, query)
     43     return response 
     45 result = self.model.stream_chat(self.tokenizer,
     46     query,self.history,None,self.max_length,
     47     self.do_sample,self.top_p,self.temperature,
     48     self.logits_processor,None)
---> 50 for response,history in result:
     51     print(response)
     52     clear_output(wait=True)

File ~/anaconda3/envs/zdw/lib/python3.10/site-packages/torch/utils/_contextlib.py:26, in _wrap_generator.<locals>.generator_context(*args, **kwargs)
     24 @functools.wraps(func)
     25 def generator_context(*args, **kwargs):
---> 26     gen = func(*args, **kwargs)
     28     # Generators are suspended and unsuspended at `yield`, hence we
     29     # make sure the grad mode is properly set every time the execution
     30     # flow returns into the wrapped generator and restored when it
     31     # returns through our `yield` to our caller (see PR #49017).
     32     try:
     33         # Issuing `None` to a generator fires it up

TypeError: ChatGLMForConditionalGeneration.stream_chat() takes from 3 to 9 positional arguments but 11 were given

请问是哪里出问题了啊

首先感谢开源这么有意义的项目！

发现torchkeras对对学习率衰减支持得不好，首先是lr_scheduler.step()没传入参数metrics，其次是使用lr_scheduler后，学习率明明已经下降，但processing bar显示的lr仍然不变…… 改起来应该不难，希望作者更新fix

我看教程好像是自动选择的，但是我实际使用中，还是显示的cpu，torchkeras版本已经更新到最新