codegen-16B-mono (Salesforce) fails to load tokenizer and model

[weidotwisc]

System Info

• transformers version: 4.20.1
• Platform: Linux-4.18.0-193.19.1.el8_2.x86_64-x86_64-with-glibc2.28
• Python version: 3.9.10
• Huggingface_hub version: 0.8.1
• PyTorch version (GPU?): 1.7.1+cu110 (True)
• Tensorflow version (GPU?): not installed (NA)
• Flax version (CPU?/GPU?/TPU?): not installed (NA)
• Jax version: not installed
• JaxLib version: not installed
• Using GPU in script?: <N/A>
• Using distributed or parallel set-up in script?: <N/A>

Who can help?

Per https://huggingface.co/Salesforce/codegen-16B-mono?text=What+is+projection+matrix, I should be able to load the codegen tokenizer and model by doing
tokenizer = AutoTokenizer.from_pretrained("Salesforce/codegen-16B-mono")
model = AutoModelForCausalLM.from_pretrained("Salesforce/codegen-16B-mono")
@SaulLu When I do tokenizer = AutoTokenizer.from_pretrained("Salesforce/codegen-16B-mono"), I got this error:
"...huggingface_py3.9/lib/python3.9/site-packages/transformers/models/auto/tokenization_auto.py", line 576, in from_pretrained
raise ValueError(
ValueError: Tokenizer class CodeGenTokenizer does not exist or is not currently imported.

@LysandreJik When I do model = AutoModelForCausalLM.from_pretrained("Salesforce/codegen-16B-mono"), I got this error:
"... huggingface_py3.9/lib/python3.9/site-packages/transformers/models/auto/configuration_auto.py", line 725, in from_pretrained
config_class = CONFIG_MAPPING[config_dict["model_type"]]
File "/Volume0/userhomes/weiz/venvs/huggingface_py3.9/lib/python3.9/site-packages/transformers/models/auto/configuration_auto.py", line 432, in getitem
raise KeyError(key)
KeyError: 'codegen'
"

Thanks!
Wei

Information

• The official example scripts
• My own modified scripts

Tasks

• An officially supported task in the examples folder (such as GLUE/SQuAD, ...)
• My own task or dataset (give details below)

Reproduction

Follow the model card at https://huggingface.co/Salesforce/codegen-16B-mono?text=What+is+projection+matrix
from transformers import AutoTokenizer, AutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("Salesforce/codegen-16B-mono")
model = AutoModelForCausalLM.from_pretrained("Salesforce/codegen-16B-mono")

I then got the Errors:

"...huggingface_py3.9/lib/python3.9/site-packages/transformers/models/auto/tokenization_auto.py", line 576, in from_pretrained
raise ValueError(
ValueError: Tokenizer class CodeGenTokenizer does not exist or is not currently imported.

"... huggingface_py3.9/lib/python3.9/site-packages/transformers/models/auto/configuration_auto.py", line 725, in from_pretrained
config_class = CONFIG_MAPPING[config_dict["model_type"]]
File "/Volume0/userhomes/weiz/venvs/huggingface_py3.9/lib/python3.9/site-packages/transformers/models/auto/configuration_auto.py", line 432, in getitem
raise KeyError(key)
KeyError: 'codegen'
"

Expected behavior

The tokenizer and Model should be loaded successfully.

[SaulLu]

Hi @weidotwisc ,

You get these errors because CodeGen was only merged onto the master branch of the repo 5 days ago (PR #17443) and therefore has not been released yet. 😄

If you want to use it now without waiting for a release, you can install a transformers version on master. For example with pip by running pip install git+https://github.com/huggingface/transformers.git

[weidotwisc]

@SaulLu Thanks for the help! I am now able to load its tokenizer and model and follow through the model card example.

Thanks,

Wei