The recommended format is SavedModel. I am new to Pytorch and still wasnt able to figure one this out yet! Could it be possible that you had gradient_accumulation_steps>1? This function uses Python's pickle utility for serialization. Use this simple code snippet. If you use summary as a column name, you will see the error message. module . Staging Ground Beta 1 Recap, and Reviewers needed for Beta 2. I don't know how you defined the tokenizer and what you assigned the "tokenizer" variable to, but this can be a solution to your problem: This saves everything about the tokenizer and with the your_model.save_pretrained('results/tokenizer/') you get: If you are using from pytorch_pretrained_bert import BertForSequenceClassification then that attribute is not available (as you can see from the code). Im not sure which notebook you are referencing. Dataparallel. Stack Exchange Network Stack Exchange network consists of 180 Q&A communities including Stack Overflow , the largest, most trusted online community for developers to learn, share their knowledge, and build their careers. PYTORCHGPU. Hi, i meet the same problem, have you solved this problem? . They are generally the std values of the dataset on which the backbone has been trained on rpn_anchor_generator (AnchorGenerator): module that generates the anchors for a set of feature maps. type(self).name, name)) AttributeError: 'DataParallel' object has no attribute 'save_pretrained'. I have just followed this tutorial on how to train my own tokenizer. aaa = open(r'C:\Users\hahaha\.spyder-py3\py. Applying LIME interpretation on my fine-tuned BERT for sequence classification model? I get this error: AttributeError: 'list' object has no attribute 'split. I have just followed this tutorial on how to train my own tokenizer. model = BERT_CLASS. def save_checkpoint(state, is_best, filename = 'checkpoint.pth.tar'): . You can either add a nn.DataParallel temporarily in your network for loading purposes, or you can load the weights file, create a new ordered dict without the module prefix, and load it back. RuntimeError: module must have its parameters and buffers on device cuda:0 (device_ids[0]) but found. Modified 1 year, 11 months ago. dir, epoch, is_best=is . By clicking Sign up for GitHub, you agree to our terms of service and It means you need to change the model.function() to model.module.function() in the following codes. Orari Messe Chiese Barletta, This can be done by either setting CUDA_VISIBLE_DEVICES for every process or by calling: >>> torch.cuda.set_device(i) Copy to clipboard. AttributeError: 'BertModel' object has no attribute 'save_pretrained' The text was updated successfully, but these errors were encountered: Copy link Member LysandreJik commented Feb 18, 2020. recognizer. To learn more, see our tips on writing great answers. Already on GitHub? Generally, check the type of object you are using before you call the lower() method. trainer.save_pretrained (modeldir) AttributeError: 'Trainer' object has no attribute 'save_pretrained' Transformers version 4.8.0 sgugger December 20, 2021, 1:54pm 2 I don't knoe where you read that code, but Trainer does not have a save_pretrained method. If a column in your DataFrame uses a protected keyword as the column name, you will get an error message. I guess you could find some help from this Sign in I have the same issue when I use multi-host training (2 multigpu instances) and set up gradient_accumulation_steps to 10. Copy link SachinKalsi commented Jul 26, 2021. - the incident has nothing to do with me; can I use this this way? You signed in with another tab or window. ModuleAttributeError: 'DataParallel' object has no attribute 'log_weights'. What is the purpose of this D-shaped ring at the base of the tongue on my hiking boots? import os thank in advance. Connect and share knowledge within a single location that is structured and easy to search. You can either add a nn.DataParallel temporarily in your network for loading purposes, or you can load the weights file, create a new ordered dict without the module prefix, and load it back. model = nn.DataParallel (model,device_ids= [0,1]) AttributeError: 'DataParallel' object has no attribute '****'. I see - will take a look at that. Solution: Just remove show method from your expression, and if you need to show a data frame in the middle, call it on a standalone line without chaining with other expressions: To use . Thanks. . RuntimeError: module must have its parameters and buffers on device cuda:0 (device_ids[0]) but found. I use Anaconda, for res in results: Reply. Why is there a voltage on my HDMI and coaxial cables? Commento A Zacinto Riflessioni Personali, So just to recap (in case other people find it helpful), to train the RNNLearner.language_model with FastAI with multiple GPUs we do the following: Once we have our learn object, parallelize the model by executing learn.model = torch.nn.DataParallel (learn.model) Train as instructed in the docs. where i is from 0 to N-1. You can either add a nn.DataParallel temporarily in your network for loading purposes, or you can load the weights file, create a new ordered dict without the module prefix, and load it back. Any reason to save a pretrained BERT tokenizer? Hi, from_pretrained appeared in an older version of the library. please use read/write OR save/load consistantly (both write different files) berak AttributeError: module 'cv2' has no attribute 'face_LBPHFaceRecognizer' I am using python 3.6 and opencv_3.4.3. This only happens when MULTIPLE GPUs are used. model.save_weights TensorFlow Checkpoint 2 save_formatsave_format = "tf"save_format = "h5" path.h5.hdf5HDF5 loading pretrained model pytorch. Wrap the model with model = nn.DataParallel(model). Expected behavior. colombian street rappers Menu. torch.nn.modules.module.ModuleAttributeError: 'Model' object has no attribute '_non_persistent_buffers_set' python pytorch .. Dataparallel DataparallelDistributed DataparallelDP 1.1 Dartaparallel Dataparallel net = nn.Dataparallel(net . If you want to train a language model from scratch on masked language modeling, its in this notebook. DistributedDataParallel is proven to be significantly faster than torch.nn.DataParallel for single-node multi-GPU data parallel training. But I am not quite sure on how to pass the train dataset to the trainer API. Viewed 12k times 1 I am trying to use a conditional statement to generate a raster with binary values from a raster with probability values (floating point raster). Prezzo Mattoni Forati 8x25x50, How Intuit democratizes AI development across teams through reusability. For further reading on AttributeErrors involving the list object, go to the articles: How to Solve Python AttributeError: list object has no attribute split. torch GPUmodel.state_dict (), modelmodel. "After the incident", I started to be more careful not to trip over things. XXX Thank you for your contributions. Fine tuning resnet: 'DataParallel' object has no attribute 'fc' vision yang_yang1 (Yang Yang) March 13, 2018, 7:27am #1 When I tried to fine tuning my resnet module, and run the following code: ignored_params = list (map (id, model.fc.parameters ())) base_params = filter (lambda p: id not in ignored_params, model.parameters ()) Nenhum produto no carrinho. rpn_head (nn.Module): module that computes the objectness and regression deltas from the RPN rpn_pre_nms_top_n_train (int): number of proposals to keep import utils Roberta Roberta adsbygoogle window.adsbygoogle .push world clydesdale show 2022 tickets; kelowna airport covid testing. @sgugger Do I replace the following with where I saved my trained tokenizer? Sign up for a free GitHub account to open an issue and contact its maintainers and the community. privacy statement. Thanks in advance. CLASS torch.nn.DataParallel (module, device_ids=None, output_device=None, dim=0) moduledevice_idsoutput_device. Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX. import numpy as np Contributo Covelco 2020, I have all the features extracted and saved in the disk. . 1 Like Read documentation. I was using the default version published in AWS Sagemaker. For example, 'DistributedDataParallel' object has no attribute 'save_pretrained'. File /usr/local/lib/python2.7/dist-packages/torch/nn/modules/module.py, line 508, in load_state_dict pytorchnn.DataParrallel. How to Solve Python AttributeError: list object has no attribute shape. It means you need to change the model.function () to model.module.function () in the following codes. for name, param in state_dict.items(): import numpy as np I realize where I have gone wrong. Did any DOS compatibility layers exist for any UNIX-like systems before DOS started to become outmoded? I was wondering if you can share the train.py file. 1.. I wanted to train it on multi gpus using the huggingface trainer API. token = generate_token(ip,username) student.save() It is the default when you use model.save (). Accepted answer. non food items that contain algae dataparallel' object has no attribute save_pretrained. Transformers is our natural language processing library and our hub is now open to all ML models, with support from libraries like Flair , Asteroid , ESPnet , Pyannote, and more to come. Thanks for replying. from scipy impo, PUT 500 load model from pth file. Lex Fridman Political Views, I expect the attribute to be available, especially since the wrapper in Pytorch ensures that all attributes of the wrapped model are accessible. Thanks for your help! How to save my tokenizer using save_pretrained. If you are trying to access the fc layer in the resnet50 wrapped by the DataParallel model, you can use model.module.fc, as DataParallel stores the provided model as self.module: github.com pytorch/pytorch/blob/df8d6eeb19423848b20cd727bc4a728337b73829/torch/nn/parallel/data_parallel.py#L131 self.module = module self.device_ids = [] return To access the underlying module, you can use the module attribute: You signed in with another tab or window. Whereas News: 27/10/2018: Fix compatibility issues, Add tests, Add travis. QuerySet, Sign up for a free GitHub account to open an issue and contact its maintainers and the community. to your account, However, I keep running into: Well occasionally send you account related emails. scipy.io.savemat(file_name, mdict, appendmat=True, format='5', long_field_names=False, do_compression=False, oned_as='row') . type(self).name, name)) The model works well when I train it on a single GPU. You seem to use the same path variable in different scenarios (load entire model and load weights). Marotta Occhio Storto; Eccomi Ges Accordi Chitarra; Reggisella Carbonio 27,2 Usato; Fino Immobiliare San Pietro Vernotico; Casa Pinaldo Ginosa Marina Telefono; Nson Save Editor; to your account. How to Solve Python AttributeError: list object has no attribute strip How to Solve Python AttributeError: _csv.reader object has no attribute next To learn more about Python for data science and machine learning, go to the online courses page on Python for the most comprehensive courses available. DataParallel class torch.nn. import shutil, from config import Config By clicking Sign up for GitHub, you agree to our terms of service and 'DataParallel' object has no attribute 'generate'. Is there any way in Pytorch I might be able to extract the parameters in the pytorch model and use them? Pandas 'DataFrame' object has no attribute 'write' when trying to save it locally in Parquet file. Trying to understand how to get this basic Fourier Series. You will need the torch, torchvision and torchvision.models modules.. You might be able to call the method on your model_dm.wv object instead, but I'm not sure. model.save_pretrained(path) save and load fine-tuned bert classification model using tensorflow 2.0. how to use BertTokenizer to load Tokenizer model? The text was updated successfully, but these errors were encountered: DataParallel wraps the model. from_pretrained pytorchnn.DataParrallel. of a man with trust issues. AttributeError: 'dict' object has no attribute 'encode'. dataparallel' object has no attribute save_pretrained. warnings.warn(msg, SourceChangeWarning) AttributeError: str object has no attribute sortstrsort 1 Need to load a pretrained model, such as VGG 16 in Pytorch. You signed in with another tab or window. Well occasionally send you account related emails. Another solution would be to use AutoClasses. huggingface - save fine tuned model locally - and tokenizer too? 2 comments bilalghanem commented on Apr 27, 2022 edited bilalghanem added the label on Apr 27, 2022 on May 5, 2022 Sign up for free to join this conversation on GitHub . @AaronLeong Notably, if you use 'DataParallel', the model will be wrapped in DataParallel (). cerca indirizzo da nome e cognome dataparallel' object has no attribute save_pretrained Hi, Site design / logo 2023 Stack Exchange Inc; user contributions licensed under CC BY-SA. A command-line interface is provided to convert TensorFlow checkpoints in PyTorch models. DataParallel class torch.nn. Is it suspicious or odd to stand by the gate of a GA airport watching the planes? AttributeError: 'model' object has no attribute 'copy' Or AttributeError: 'DataParallel' object has no attribute 'copy' Or RuntimeError: module must have its parameters and buffers on device cuda:0 (device_ids[0]) but found At this time, we can load the model in the following way, first build the model, and then load the parameters. @classmethod def evaluate_checkpoint (cls, experiment_name: str, ckpt_name: str = "ckpt_latest.pth", ckpt_root_dir: str = None)-> None: """ Evaluate a checkpoint . Oh and running the same code without the ddp and using a 1 GPU instance works just fine but obviously takes much longer to complete However, I expected this not to be required anymore due to: Apparently this was never merged, so yeah. Distributed DataParallel modelmodelmodel object has no attribute xxxx bug To concatenate a string with another string, you use the concatenation operator (+). AttributeError: 'list' object has no attribute 'strip' So if 'list' object has no attribute 'strip' or 'split', how can I split a list? I have switched to 4.6.1 version, and the problem is gone. . """ The Trainer class, to easily train a Transformers from scratch or finetune it on a new task. It will be closed if no further activity occurs. Or are you installing transformers from git master branch? class torch.nn.DataParallel(module, device_ids=None, output_device=None, dim=0) [source] Implements data parallelism at the module level. Now, from training my tokenizer, I have wrapped it inside a Transformers object, so that I can use it with the transformers library: from transformers import BertTokenizerFast new_tokenizer = BertTokenizerFast(tokenizer_object=tokenizer) Then, I try to save my tokenizer using this code: tokenizer.save_pretrained('/content . AttributeError: 'AddAskForm' object has no attribute 'save' 287 1 1. tf.keras.models.load_model () There are two formats you can use to save an entire model to disk: the TensorFlow SavedModel format, and the older Keras H5 format . the_model.load_state_dict(torch.load(path)) Have a question about this project? DataParallelinit_hidden(DataParallel object has no attribute init_hidden) 2018-10-30 16:56:48 RNN DataParallel Also don't try to save torch.save(model.parameters(), filepath). ModuleAttributeError: 'DataParallel' object has no attribute 'log_weights' NOTE. . Already have an account? AttributeError: 'NoneType' object has no attribute 'save' Simply finding pytorch loading model. Voli Neos In Tempo Reale, R.305-306, 3th floor, 48B Keangnam Tower, Pham Hung Street, Nam Tu Liem District, Ha Noi, Viet Nam, Tel:rotte nautiche in tempo reale Email: arbitro massa precedenti inter, , agenda 2030 attivit didattiche scuola secondaria, mirko e silvia primo appuntamento cognomi, rinuncia all'azione nei confronti di un solo convenuto fac simile. Note*: If you want to access the stdout (or) AttributeError: 'DataParallel' object has no attribute 'copy' RuntimeError: module must have its parameters and buffers on device cuda:0 (device_ids[0]) but found PSexcelself.workbook. Aruba Associare Metodo Di Pagamento, GPU0GPUGPUGPUbatch sizeGPU0 DataParallel[5]) . DataParallel (module, device_ids = None, output_device = None, dim = 0) [source] . AttributeError: 'DataParallel' object has no attribute 'save_pretrained'. You probably saved the model using nn.DataParallel, which stores the model in module, and now you are trying to load it without DataParallel. san jose police bike auction / agno3 + hcl precipitate / dataparallel' object has no attribute save_pretrained Publicerad 3 juli, 2022 av hsbc: a payment was attempted from a new device text dataparallel' object has no attribute save_pretrained Now, from training my tokenizer, I have wrapped it inside a Transformers object, so that I can use it with the transformers library: Then, I try to save my tokenizer using this code: However, from executing the code above, I get this error: If so, what is the correct approach to save it to my local files, so I can use it later? dataparallel' object has no attribute save_pretrained. Already on GitHub? Yes, try model.state_dict(), see the doc for more info. You are continuing to use pytorch_pretrained_bert instead transformers. Instead of inheriting from nn.Module you could inherit from PreTrainedModel, which is the abstract class we use for all models, that contains save_pretrained. . Whereas OK, here is the answer. Asking for help, clarification, or responding to other answers. and I am not able to load state dict also, I am looking for way to save my finetuned model with "save_pretrained". Browse other questions tagged, Where developers & technologists share private knowledge with coworkers, Reach developers & technologists worldwide, I tried your code your_model.save_pretrained('results/tokenizer/') but this error appears torch.nn.modules.module.ModuleAttributeError: 'BertForSequenceClassification' object has no attribute 'save_pretrained', Yes of course, now I try to update my answer making it more complete to explain better, I tried your updated solution but error appears torch.nn.modules.module.ModuleAttributeError: 'BertForSequenceClassification' object has no attribute 'save_pretrained', You are not using the code from my updated answer. Saving and doing Inference with Tensorflow BERT model. Please be sure to answer the question.Provide details and share your research! Since your file saves the entire model, torch.load(path) will return a DataParallel object. With the embedding size of 768, the total size of the word embedding table is ~ 4 (Bytes/FP32) * 30522 * 768 = 90 MB. . In order to get actual values you have to read the data and target content itself.. torch GPUmodel.state_dict (), modelmodel.module. "sklearn.datasets" is a scikit package, where it contains a method load_iris(). You are continuing to use, given that I fine-tuned the model and I want to save the finetuned version not the imported version and I could save the .bin file of my model using this code model_to_save = model.module if hasattr(model, 'module') else model # Only save the model it-self output_model_file = os.path.join(args.output_dir, "pytorch_model_task.bin") but i could not save other config files. A complete end-to-end MLOps pipeline used to build, deploy, monitor, improve, and scale a YOLOv7-based aerial object detection model - schwenkd/aerial-detection-mlops Since your file saves the entire model, torch.load (path) will return a DataParallel object. AttributeError: 'model' object has no attribute 'copy' . The recommended format is SavedModel. How to Solve Python AttributeError: list object has no attribute strip How to Solve Python AttributeError: _csv.reader object has no attribute next To learn more about Python for data science and machine learning, go to the online courses page on Python for the most comprehensive courses available. lake mead launch ramps 0. No products in the cart. Copy link Owner. Hi, Did you find any workaround for this? What is wrong here? Publicado el . To use DistributedDataParallel on a host with N GPUs, you should spawn up N processes, ensuring that each process exclusively works on a single GPU from 0 to N-1. AttributeError: 'str' object has no attribute 'save' 778 0 2. self.model = model # Since if the model is wrapped by the `DataParallel` class, you won't be able to access its attributes # unless you write `model.module` which breaks the code compatibility. Tried tracking down the problem but cant seem to figure it out. The nature of simulating nature: A Q&A with IBM Quantum researcher Dr. Jamie We've added a "Necessary cookies only" option to the cookie consent popup. .load_state_dict (. Have a question about this project? Sign in . However, it is a mlflow project and you need docker with the nvidia-container thingy to run it. Django problem : "'tuple' object has no attribute 'save'" Home. So I'm trying to create a database and store data, that I get from django forms. The example below will show how to check the type It might be unintentional, but you called show on a data frame, which returns a None object, and then you try to use df2 as data frame, but its actually None. Have a question about this project? And, one more thing When I want to use my tokenizer for masked language modelling, do I use the pretrained model notebook? Solution: Just remove show method from your expression, and if you need to show a data frame in the middle, call it on a standalone line without chaining with other expressions: Please be sure to answer the question.Provide details and share your research! Showing session object has no attribute 'modified' Related Posts. How to save / serialize a trained model in theano? You will need the torch, torchvision and torchvision.models modules.. DataParallelinit_hidden(DataParallel object has no attribute init_hidden) 2018-10-30 16:56:48 RNN DataParallel thanks. Parameters In other words, we will see the stderr of both java commands executed on both machines. import scipy.misc 9. Is there any way to save all the details of my model? privacy statement. Well occasionally send you account related emails. AttributeError: DataParallel object has no Implements data parallelism at the module level. . model = BERT_CLASS. For further reading on AttributeErrors involving the list object, go to the articles: How to Solve Python AttributeError: list object has no attribute split. shean1488-3 Light Poster . new_tokenizer.save_pretrained(xxx) should work. Well occasionally send you account related emails. . venetian pool tickets; . Sign in # resre import rere, the entire model or just the weights? I wonder, if gradient_accumulation_steps is not compatible with multi-host training at all, or there are other parameters I need to tweak? If you are a member, please kindly clap. I am pretty sure the file saved the entire model. model = BERT_CLASS. import skimage.color Loading Google AI or OpenAI pre-trained weights or PyTorch dump. yhenon/pytorch-retinanet PytorchRetinanet visualize.pyAttributeError: 'collections.OrderedDict' object has no attribute 'cuda' . DataParallel. Powered by Discourse, best viewed with JavaScript enabled, Data parallelism error for pretrained model, pytorch/pytorch/blob/df8d6eeb19423848b20cd727bc4a728337b73829/torch/nn/parallel/data_parallel.py#L131, device_ids = list(range(torch.cuda.device_count())), self.device_ids = list(map(lambda x: _get_device_index(x, True), device_ids)), self.output_device = _get_device_index(output_device, True), self.src_device_obj = torch.device("cuda:{}".format(self.device_ids[0])). autocertificazione certificato contestuale di residenza e stato di famiglia; costo manodopera regione lazio 2020; taxi roma fiumicino telefono; carta d'identit del pinguino Can you try that? Have a question about this project? The DataFrame API contains a small number of protected keywords. dataparallel' object has no attribute save_pretrained. When it comes to saving and loading models, there are three core functions to be familiar with: torch.save : Saves a serialized object to disk. tf.keras.models.load_model () There are two formats you can use to save an entire model to disk: the TensorFlow SavedModel format, and the older Keras H5 format . I dont install transformers separately, just use the one that goes with Sagemaker. What you should do is use transformers which also integrate this functionality. How to Solve Python AttributeError: list object has no attribute strip How to Solve Python AttributeError: _csv.reader object has no attribute next To learn more about Python for data science and machine learning, go to the online courses page on Python for the most comprehensive courses available. Powered by Discourse, best viewed with JavaScript enabled, AttributeError: 'DataParallel' object has no attribute 'items'. Keras API . answered Jul 17, 2018 at 9:10. djstrong. Difficulties with estimation of epsilon-delta limit proof, Relation between transaction data and transaction id. Immagini Sulla Violenza In Generale, I saw in your initial(first thread) code: Can you(or someone) please explain to me why a module cannot be instance of nn.ModuleList, nn.Sequential or self.pModel in order to obtain the weights of each layer? But when I want to parallel the data across several GPUs by doing model = nn.DataParallel(model), I can't save the model. Checkout the documentaiton for a list of its methods! workbook1.save (workbook1)workbook1.save (excel). Oh and running the same code without the ddp and using a 1 GPU instance works just fine but obviously takes much longer to complete. L:\spn\Anaconda3\lib\site-packages\torch\serialization.py:786: SourceChangeWarning: source code of class 'torch.nn.parallel.data_parallel.DataParallel' has changed. For further reading on AttributeErrors, go to the article: How to Solve Python AttributeError: numpy.ndarray object has no attribute append. In the forward pass, the module . Pretrained models for Pytorch (Work in progress) The goal of this repo is: to help to reproduce research papers results (transfer learning setups for instance), to access pretrained ConvNets with a unique interface/API inspired by torchvision. @zhangliyun9120 Hi, did you solve the problem? Solution 3. how to solve cv2.face_LBPHFaceRecognizer object has no attribute 'load' ? uhvardhan (Harshvardhan Uppaluru) October 4, 2018, 6:04am #5 you can retrieve the original source code by accessing the object's source attribute or set torch.nn.Module.dump_patches = True and use the patch tool to revert the changes. if the variable is of type list, then call the append method. SentimentClassifier object has no attribute 'save_pretrained' which is correct but I also want to know how can I save that model with my trained weights just like the base model so that I can Import it in few lines and use it. model.train_model --> model.module.train_model, @jytime I have tried this setting, but only one GPU can work well, user@ubuntu:~/rcnn$ nvidia-smi Sat Sep 22 15:31:48 2018 +-----------------------------------------------------------------------------+ | NVIDIA-SMI 396.45 Driver Version: 396.45 | |-------------------------------+----------------------+----------------------+ | GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. !:AttributeError:listsplit This is my code: : myList = ['hello'] myList.split() 2 To use DistributedDataParallel on a host with N GPUs, you should spawn up N processes, ensuring that each process exclusively works on a single GPU from 0 to N-1. The main part is run_nnet.py. 91 3. () torch.nn.DataParallel GPUBUG. Contribute to bkbillybk/YoloV5 by creating an account on DAGsHub. June 3, 2022 . How can I fix this ? When using DataParallel your original module will be in attribute module of the parallel module: Show activity on this post. Sirs: Modified 7 years, 10 months ago. 1.. openpyxl. You are saving the wrong tokenizer ;-). dataparallel' object has no attribute save_pretrained. News: 27/10/2018: Fix compatibility issues, Add tests, Add travis. to your account, Hey, I want to use EncoderDecoderModel for parallel trainging. You seem to use the same path variable in different scenarios (load entire model and load weights). I saved the binary model file by the following code, but when I used it to save tokenizer or config file I could not do it because I dnot know what file extension should I save tokenizer and I could not reach cofig file, Posted on . Thank you very much for that! Pretrained models for Pytorch (Work in progress) The goal of this repo is: to help to reproduce research papers results (transfer learning setups for instance), to access pretrained ConvNets with a unique interface/API inspired by torchvision. Otherwise, take the alternative path and ignore the append () attribute. March 17, 2020, 5:23pm #1 While trying to load a checkpoint into a resnet model I get this error ! The text was updated successfully, but these errors were encountered: @AaronLeong Notably, if you use 'DataParallel', the model will be wrapped in DataParallel(). So that I can transfer the parameters in Pytorch model to Keras. News: 27/10/2018: Fix compatibility issues, Add tests, Add travis. only thing I Need to load a pretrained model, such as VGG 16 in Pytorch. model nn.DataParallel module . import skimage.io, from pycocotools.coco import COCO AttributeError: 'DataParallel' object has no attribute 'save'.