random_token_example error

[tanyaroosta]

Hi,
I have pulled the latest repo and compiled the package. When running the random_token_example/distill.py, I run into an issue as shown below. I am not clear how to fix it, as it doesn't seem distill.py takes any user input, and so it should be running without any problems.

---

AttributeError Traceback (most recent call last)
in
29 with distiller:
30 distiller.train(optimizer, dataloader, num_epochs=num_epochs,
---> 31 scheduler_class=scheduler_class, scheduler_args=scheduler_args, callback=callback_fun)

~/opt/anaconda3/lib/python3.7/site-packages/textbrewer/distiller_basic.py in train(self, optimizer, dataloader, num_epochs, scheduler_class, scheduler_args, scheduler, max_grad_norm, num_steps, callback, batch_postprocessor, **args)
281 self.train_with_num_steps(optimizer, scheduler, tqdm_disable, dataloader, max_grad_norm, num_steps, callback, batch_postprocessor, **args)
282 else:
--> 283 self.train_with_num_epochs(optimizer, scheduler, tqdm_disable, dataloader, max_grad_norm, num_epochs, callback, batch_postprocessor, **args)
284
285

~/opt/anaconda3/lib/python3.7/site-packages/textbrewer/distiller_basic.py in train_with_num_epochs(self, optimizer, scheduler, tqdm_disable, dataloader, max_grad_norm, num_epochs, callback, batch_postprocessor, **args)
244 if (global_step%train_steps_per_epoch in checkpoints)
245 and ((current_epoch+1)%self.t_config.ckpt_epoch_frequency==0 or current_epoch+1==num_epochs):
--> 246 self.save_and_callback(global_step, step, current_epoch, callback)
247
248 logger.info(f"Epoch {current_epoch+1} finished")

~/opt/anaconda3/lib/python3.7/site-packages/textbrewer/distiller_general.py in save_and_callback(self, global_step, step, epoch, callback)
63 self.model_T._forward_hooks = OrderedDict()
64
---> 65 super(GeneralDistiller, self).save_and_callback(global_step, step, epoch, callback)
66
67 if self.has_custom_matches:

~/opt/anaconda3/lib/python3.7/site-packages/textbrewer/distiller_basic.py in save_and_callback(self, global_step, step, epoch, callback)
36 if callback is not None:
37 logger.info("Running callback function...")
---> 38 callback(model=self.model_S, step=global_step)
39 self.model_S.train()
40

in predict(model, eval_dataset, step, device)
49 with torch.no_grad():
50 logits, _ = model(input_ids=input_ids, attention_mask=attention_mask)
---> 51 cpu_logits = logits.detach().cpu()
52 for i in range(len(cpu_logits)):
53 pred_logits.append(cpu_logits[i].numpy())

AttributeError: 'str' object has no attribute 'detach'

[airaria]

Hi, is your transformers version >= 4.0? I think this problem is caused by the breaking change between transformers 4 and transformers 3. In transformers 4, the output of the model is a dict-like object by default, rather than a tuple. The solution is 1. to call the model with the argument return_dict=False (line 50 in your program): ```python logits, _ = model(input_ids=input_ids, attention_mask=attention_mask, return_dict=False) ``` or 2. unpatch the output correctly: ```python output = model(input_ids=input_ids, attention_mask=attention_mask) logits = output[’logits’] ``` or 3. using transformers 3. Hope it helps. On 09/16/2021 ***@***.***> wrote： Hi, I have pulled the latest repo and compiled the package. When running the random_token_example/distill.py, I run into an issue as shown below. I am not clear how to fix it, as it doesn't seem distill.py takes any user input, and so it should be running without any problems. AttributeError Traceback (most recent call last) in 29 with distiller: 30 distiller.train(optimizer, dataloader, num_epochs=num_epochs, ---> 31 scheduler_class=scheduler_class, scheduler_args=scheduler_args, callback=callback_fun) ~/opt/anaconda3/lib/python3.7/site-packages/textbrewer/distiller_basic.py in train(self, optimizer, dataloader, num_epochs, scheduler_class, scheduler_args, scheduler, max_grad_norm, num_steps, callback, batch_postprocessor, **args) 281 self.train_with_num_steps(optimizer, scheduler, tqdm_disable, dataloader, max_grad_norm, num_steps, callback, batch_postprocessor, **args) 282 else: --> 283 self.train_with_num_epochs(optimizer, scheduler, tqdm_disable, dataloader, max_grad_norm, num_epochs, callback, batch_postprocessor, **args) 284 285 ~/opt/anaconda3/lib/python3.7/site-packages/textbrewer/distiller_basic.py in train_with_num_epochs(self, optimizer, scheduler, tqdm_disable, dataloader, max_grad_norm, num_epochs, callback, batch_postprocessor, **args) 244 if (global_step%train_steps_per_epoch in checkpoints) 245 and ((current_epoch+1)%self.t_config.ckpt_epoch_frequency==0 or current_epoch+1==num_epochs): --> 246 self.save_and_callback(global_step, step, current_epoch, callback) 247 248 logger.info(f"Epoch {current_epoch+1} finished") ~/opt/anaconda3/lib/python3.7/site-packages/textbrewer/distiller_general.py in save_and_callback(self, global_step, step, epoch, callback) 63 self.model_T._forward_hooks = OrderedDict() 64 ---> 65 super(GeneralDistiller, self).save_and_callback(global_step, step, epoch, callback) 66 67 if self.has_custom_matches: ~/opt/anaconda3/lib/python3.7/site-packages/textbrewer/distiller_basic.py in save_and_callback(self, global_step, step, epoch, callback) 36 if callback is not None: 37 logger.info("Running callback function...") ---> 38 callback(model=self.model_S, step=global_step) 39 self.model_S.train() 40 in predict(model, eval_dataset, step, device) 49 with torch.no_grad(): 50 logits, _ = model(input_ids=input_ids, attention_mask=attention_mask) ---> 51 cpu_logits = logits.detach().cpu() 52 for i in range(len(cpu_logits)): 53 pred_logits.append(cpu_logits[i].numpy()) AttributeError: 'str' object has no attribute 'detach' — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android.

[tanyaroosta]

Thanks for your reply and help. Tanya

…

On Wed, Sep 15, 2021 at 6:00 PM Ziqing Yang ***@***.***> wrote: Hi, is your transformers version >= 4.0? I think this problem is caused by the breaking change between transformers 4 and transformers 3. In transformers 4, the output of the model is a dict-like object by default, rather than a tuple. The solution is 1. to call the model with the argument return_dict=False (line 50 in your program): ```python logits, _ = model(input_ids=input_ids, attention_mask=attention_mask, return_dict=False) ``` or 2. unpatch the output correctly: ```python output = model(input_ids=input_ids, attention_mask=attention_mask) logits = output[’logits’] ``` or 3. using transformers 3. Hope it helps. On 09/16/2021 ***@***.***> wrote： Hi, I have pulled the latest repo and compiled the package. When running the random_token_example/distill.py, I run into an issue as shown below. I am not clear how to fix it, as it doesn't seem distill.py takes any user input, and so it should be running without any problems. AttributeError Traceback (most recent call last) in 29 with distiller: 30 distiller.train(optimizer, dataloader, num_epochs=num_epochs, ---> 31 scheduler_class=scheduler_class, scheduler_args=scheduler_args, callback=callback_fun) ~/opt/anaconda3/lib/python3.7/site-packages/textbrewer/distiller_basic.py in train(self, optimizer, dataloader, num_epochs, scheduler_class, scheduler_args, scheduler, max_grad_norm, num_steps, callback, batch_postprocessor, **args) 281 self.train_with_num_steps(optimizer, scheduler, tqdm_disable, dataloader, max_grad_norm, num_steps, callback, batch_postprocessor, **args) 282 else: --> 283 self.train_with_num_epochs(optimizer, scheduler, tqdm_disable, dataloader, max_grad_norm, num_epochs, callback, batch_postprocessor, **args) 284 285 ~/opt/anaconda3/lib/python3.7/site-packages/textbrewer/distiller_basic.py in train_with_num_epochs(self, optimizer, scheduler, tqdm_disable, dataloader, max_grad_norm, num_epochs, callback, batch_postprocessor, **args) 244 if (global_step%train_steps_per_epoch in checkpoints) 245 and ((current_epoch+1)%self.t_config.ckpt_epoch_frequency==0 or current_epoch+1==num_epochs): --> 246 self.save_and_callback(global_step, step, current_epoch, callback) 247 248 logger.info(f"Epoch {current_epoch+1} finished") ~/opt/anaconda3/lib/python3.7/site-packages/textbrewer/distiller_general.py in save_and_callback(self, global_step, step, epoch, callback) 63 self.model_T._forward_hooks = OrderedDict() 64 ---> 65 super(GeneralDistiller, self).save_and_callback(global_step, step, epoch, callback) 66 67 if self.has_custom_matches: ~/opt/anaconda3/lib/python3.7/site-packages/textbrewer/distiller_basic.py in save_and_callback(self, global_step, step, epoch, callback) 36 if callback is not None: 37 logger.info("Running callback function...") ---> 38 callback(model=self.model_S, step=global_step) 39 self.model_S.train() 40 in predict(model, eval_dataset, step, device) 49 with torch.no_grad(): 50 logits, _ = model(input_ids=input_ids, attention_mask=attention_mask) ---> 51 cpu_logits = logits.detach().cpu() 52 for i in range(len(cpu_logits)): 53 pred_logits.append(cpu_logits[i].numpy()) AttributeError: 'str' object has no attribute 'detach' — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#81 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AHMA5SI7KHBDBNS772PPEP3UCE6RDANCNFSM5EDMXCRA> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.

[tanyaroosta]

Hi Ziqing, I have another question on the msra notebook. I was able to run it to the point of training, however, when I started the training, I repeatedly saw this error and the notebook finally froze: [I 07:37:39.289 NotebookApp] Saving file at /examples/notebook_examples/msra_ner.ipynb [E 07:37:39.302 NotebookApp] Notebook JSON is invalid: {'version_major': 2, 'version_minor': 0, 'model_id': '2a114915871b4c7988a94fa7a633b646'} is not valid under any of the given schemas Failed validating 'oneOf' in display_data['properties']['data']['patternProperties']['^(?!application/json$)[a-zA-Z0-9]+/[a-zA-Z0-9\\-\\+\\.]+$']: On instance['cells'][7]['outputs'][0]['data']['application/vnd.jupyter.widget-view+json']: {'model_id': '2a114915871b4c7988a94fa7a633b646', 'version_major': 2, 'version_minor': 0} Is there an issue on my end? Thanks, Tanya

…

On Thu, Sep 16, 2021 at 7:50 AM Tanya Roosta ***@***.***> wrote: Thanks for your reply and help. Tanya On Wed, Sep 15, 2021 at 6:00 PM Ziqing Yang ***@***.***> wrote: > Hi, > is your transformers version >= 4.0? > I think this problem is caused by the breaking change between > transformers 4 and transformers 3. > In transformers 4, the output of the model is a dict-like object by > default, rather than a tuple. > > > The solution is > 1. to call the model with the argument return_dict=False (line 50 in your > program): > ```python > logits, _ = model(input_ids=input_ids, attention_mask=attention_mask, > return_dict=False) > ``` > or > 2. unpatch the output correctly: > ```python > output = model(input_ids=input_ids, attention_mask=attention_mask) > logits = output[’logits’] > ``` > or > 3. using transformers 3. > > > > > Hope it helps. > On 09/16/2021 ***@***.***> wrote： > > Hi, > I have pulled the latest repo and compiled the package. When running the > random_token_example/distill.py, I run into an issue as shown below. I am > not clear how to fix it, as it doesn't seem distill.py takes any user > input, and so it should be running without any problems. > > AttributeError Traceback (most recent call last) > in > 29 with distiller: > 30 distiller.train(optimizer, dataloader, num_epochs=num_epochs, > ---> 31 scheduler_class=scheduler_class, scheduler_args=scheduler_args, > callback=callback_fun) > > ~/opt/anaconda3/lib/python3.7/site-packages/textbrewer/distiller_basic.py > in train(self, optimizer, dataloader, num_epochs, scheduler_class, > scheduler_args, scheduler, max_grad_norm, num_steps, callback, > batch_postprocessor, **args) > 281 self.train_with_num_steps(optimizer, scheduler, tqdm_disable, > dataloader, max_grad_norm, num_steps, callback, batch_postprocessor, **args) > 282 else: > --> 283 self.train_with_num_epochs(optimizer, scheduler, tqdm_disable, > dataloader, max_grad_norm, num_epochs, callback, batch_postprocessor, > **args) > 284 > 285 > > ~/opt/anaconda3/lib/python3.7/site-packages/textbrewer/distiller_basic.py > in train_with_num_epochs(self, optimizer, scheduler, tqdm_disable, > dataloader, max_grad_norm, num_epochs, callback, batch_postprocessor, > **args) > 244 if (global_step%train_steps_per_epoch in checkpoints) > 245 and ((current_epoch+1)%self.t_config.ckpt_epoch_frequency==0 or > current_epoch+1==num_epochs): > --> 246 self.save_and_callback(global_step, step, current_epoch, callback) > 247 > 248 logger.info(f"Epoch {current_epoch+1} finished") > > ~/opt/anaconda3/lib/python3.7/site-packages/textbrewer/distiller_general.py > in save_and_callback(self, global_step, step, epoch, callback) > 63 self.model_T._forward_hooks = OrderedDict() > 64 > ---> 65 super(GeneralDistiller, self).save_and_callback(global_step, > step, epoch, callback) > 66 > 67 if self.has_custom_matches: > > ~/opt/anaconda3/lib/python3.7/site-packages/textbrewer/distiller_basic.py > in save_and_callback(self, global_step, step, epoch, callback) > 36 if callback is not None: > 37 logger.info("Running callback function...") > ---> 38 callback(model=self.model_S, step=global_step) > 39 self.model_S.train() > 40 > > in predict(model, eval_dataset, step, device) > 49 with torch.no_grad(): > 50 logits, _ = model(input_ids=input_ids, attention_mask=attention_mask) > ---> 51 cpu_logits = logits.detach().cpu() > 52 for i in range(len(cpu_logits)): > 53 pred_logits.append(cpu_logits[i].numpy()) > > AttributeError: 'str' object has no attribute 'detach' > > — > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub, or unsubscribe. > Triage notifications on the go with GitHub Mobile for iOS or Android. > > — > You are receiving this because you authored the thread. > Reply to this email directly, view it on GitHub > <#81 (comment)>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AHMA5SI7KHBDBNS772PPEP3UCE6RDANCNFSM5EDMXCRA> > . > Triage notifications on the go with GitHub Mobile for iOS > <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> > or Android > <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>. > >

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[stale]

Closing the issue, since no updates observed. Feel free to re-open if you need any further assistance.