# TF's alternative for HuggingFace's deprecated \`TF\`\ models? **URL:** \ **Category:** TensorFlow\ **Tags:** tf-model, models\ **Created:** [September 23, 2025, 2:50am UTC](https://discuss.ai.google.dev/t/tfs-alternative-for-huggingfaces-deprecated-tf-foo-models/106202 "2025-09-23T02:50:29Z")\ **Posts on this page:** 8\ **Page:** 1
**Author:** ![khteh](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/khteh/32/61930_2.png) [@khteh](https://discuss.ai.google.dev/u/khteh)\ **Post date:** [September 23, 2025, 2:50am UTC](https://discuss.ai.google.dev/t/tfs-alternative-for-huggingfaces-deprecated-tf-foo-models/106202/1 "2025-09-23T02:50:29Z")
HuggingFace has deprecated `TF` models. What is the TF alternative for me to use the transformers tokenizer and BERT models? Specifically, I am looking for the following: ```auto from transformers import AutoTokenizer, TFBertForMaskedLM MODEL = "bert-base-uncased" tokenizer = AutoTokenizer.from_pretrained(MODEL) model = TFBertForMaskedLM.from_pretrained(MODEL) ``` > > > \### System Info > > \`\`\` > \- \`transformers\` version: 4.56.1 > \- Platform: Linux-6.14.0-2…9-generic-x86\_64-with-glibc2.41 > \- Python version: 3.13.3 > \- Huggingface\_hub version: 0.34.4 > \- Safetensors version: 0.6.2 > \- Accelerate version: not installed > \- Accelerate config: not found > \- DeepSpeed version: not installed > \- PyTorch version (accelerator?): not installed (NA) > \- Tensorflow version (GPU?): 2.20.0 (True) > \- Flax version (CPU?/GPU?/TPU?): not installed (NA) > \- Jax version: not installed > \- JaxLib version: not installed > \- Using distributed or parallel set-up in script?: \ > \`\`\` > > \### Who can help? > > \`\`\` > model = TFBertForMaskedLM.from\_pretrained(MODEL) > File "/home/khteh/.local/share/virtualenvs/pAIthon-GaqEDHQT/lib/python3.13/site-packages/transformers/modeling\_tf\_utils.py", line 2964, in from\_pretrained > return load\_pytorch\_state\_dict\_in\_tf2\_model( > model, > ...\<6 lines\>... > tf\_to\_pt\_weight\_rename=tf\_to\_pt\_weight\_rename, > ) > File "/home/khteh/.local/share/virtualenvs/pAIthon-GaqEDHQT/lib/python3.13/site-packages/transformers/modeling\_tf\_pytorch\_utils.py", line 333, in load\_pytorch\_state\_dict\_in\_tf2\_model > for key in pt\_state\_dict: > ^^^^^^^^^^^^^ > TypeError: 'builtins.safe\_open' object is not iterable > \`\`\` > > \### Information > > \- \[\] The official example scripts > \- \[x\] My own modified scripts > > \### Tasks > > \- \[\] An officially supported task in the \`examples\` folder (such as GLUE/SQuAD, ...) > \- \[x\] My own task or dataset (give details below) > > \### Reproduction > > \`\`\` > from transformers import AutoTokenizer, TFBertForMaskedLM > model = TFBertForMaskedLM.from\_pretrained(MODEL) > \`\`\` > > \### Expected behavior > > NO error ---
**Author:** ![Divya\_Sree\_Kayyuri](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/divya_sree_kayyuri/32/29791_2.png) [@Divya\_Sree\_Kayyuri](https://discuss.ai.google.dev/u/Divya_Sree_Kayyuri)\ **Post date:** [September 24, 2025, 8:40am UTC](https://discuss.ai.google.dev/t/tfs-alternative-for-huggingfaces-deprecated-tf-foo-models/106202/3 "2025-09-24T08:40:52Z")
Hi @khteh, The recommended approach is to switching from TF to PyTorch. ```auto from transformers import AutoTokenizer, AutoModelForMaskedLM MODEL = “bert-base-uncased” tokenizer = AutoTokenizer.from_pretrained(MODEL) model = AutoModelForMaskedLM.from_pretrained(MODEL) ``` ---
**Author:** ![khteh](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/khteh/32/61930_2.png) [@khteh](https://discuss.ai.google.dev/u/khteh)\ **Post date:** [September 24, 2025, 8:45am UTC](https://discuss.ai.google.dev/t/tfs-alternative-for-huggingfaces-deprecated-tf-foo-models/106202/4 "2025-09-24T08:45:03Z")
Do I need to install `PyTorch` library for this to work? If “yes”, wouldn’t that introduce too many and sometimes conflicting dependencies in my project? Is there no solution at all from `TF`? ---
**Author:** ![Divya\_Sree\_Kayyuri](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/divya_sree_kayyuri/32/29791_2.png) [@Divya\_Sree\_Kayyuri](https://discuss.ai.google.dev/u/Divya_Sree_Kayyuri)\ **Post date:** [September 24, 2025, 8:49am UTC](https://discuss.ai.google.dev/t/tfs-alternative-for-huggingfaces-deprecated-tf-foo-models/106202/5 "2025-09-24T08:49:55Z")
Could you tried with older version of transformers?? ---
**Author:** ![khteh](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/khteh/32/61930_2.png) [@khteh](https://discuss.ai.google.dev/u/khteh)\ **Post date:** [September 24, 2025, 8:51am UTC](https://discuss.ai.google.dev/t/tfs-alternative-for-huggingfaces-deprecated-tf-foo-models/106202/6 "2025-09-24T08:51:04Z")
No I can’t and don’t do that! ---
**Author:** ![Divya\_Sree\_Kayyuri](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/divya_sree_kayyuri/32/29791_2.png) [@Divya\_Sree\_Kayyuri](https://discuss.ai.google.dev/u/Divya_Sree_Kayyuri)\ **Post date:** [November 7, 2025, 6:52am UTC](https://discuss.ai.google.dev/t/tfs-alternative-for-huggingfaces-deprecated-tf-foo-models/106202/8 "2025-11-07T06:52:39Z")
Hi @khteh, Could you please refer to this [document](https://huggingface.co/docs/transformers/v4.49.0/en/model_doc/bert#transformers.BertForMaskedLM) to switch `TFBertForMaskedLM` model to PyTorch. ---
**Author:** ![khteh](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/khteh/32/61930_2.png) [@khteh](https://discuss.ai.google.dev/u/khteh)\ **Post date:** [November 7, 2025, 8:41am UTC](https://discuss.ai.google.dev/t/tfs-alternative-for-huggingfaces-deprecated-tf-foo-models/106202/10 "2025-11-07T08:41:05Z")
What you mean is that using tensorflow/keras models is discouraged / a dead end with HuggingFace? ---
**Author:** ![Divya\_Sree\_Kayyuri](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/divya_sree_kayyuri/32/29791_2.png) [@Divya\_Sree\_Kayyuri](https://discuss.ai.google.dev/u/Divya_Sree_Kayyuri)\ **Post date:** [November 18, 2025, 4:26am UTC](https://discuss.ai.google.dev/t/tfs-alternative-for-huggingfaces-deprecated-tf-foo-models/106202/12 "2025-11-18T04:26:47Z")
Hi , TensorFlow support in Transformers is now deprecated. If possible, I would recommend switching to the PyTorch version. Thanks!