TypeError: 'NoneType' object is not subscriptable

[annabechang]

I am having this error while trying to load the model.

from detoxify import Detoxify

model = Detoxify('original', device="cuda")

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In [15], line 3
      1 from detoxify import Detoxify
----> 3 results = Detoxify('original').predict('some text')

File ~/.conda/envs/py/lib/python3.9/site-packages/detoxify/detoxify.py:103, in Detoxify.__init__(self, model_type, checkpoint, device, huggingface_config_path)
    101 def __init__(self, model_type="original", checkpoint=PRETRAINED_MODEL, device="cpu", huggingface_config_path=None):
    102     super().__init__()
--> 103     self.model, self.tokenizer, self.class_names = load_checkpoint(
    104         model_type=model_type,
    105         checkpoint=checkpoint,
    106         device=device,
    107         huggingface_config_path=huggingface_config_path,
    108     )
    109     self.device = device
    110     self.model.to(self.device)

File ~/.conda/envs/py/lib/python3.9/site-packages/detoxify/detoxify.py:56, in load_checkpoint(model_type, checkpoint, device, huggingface_config_path)
     50 change_names = {
     51     "toxic": "toxicity",
     52     "identity_hate": "identity_attack",
     53     "severe_toxic": "severe_toxicity",
     54 }
     55 class_names = [change_names.get(cl, cl) for cl in class_names]
---> 56 model, tokenizer = get_model_and_tokenizer(
     57     **loaded["config"]["arch"]["args"],
     58     state_dict=loaded["state_dict"],
     59     huggingface_config_path=huggingface_config_path,
     60 )
     62 return model, tokenizer, class_names

File ~/.conda/envs/py/lib/python3.9/site-packages/detoxify/detoxify.py:20, in get_model_and_tokenizer(model_type, model_name, tokenizer_name, num_classes, state_dict, huggingface_config_path)
     16 def get_model_and_tokenizer(
     17     model_type, model_name, tokenizer_name, num_classes, state_dict, huggingface_config_path=None
     18 ):
     19     model_class = getattr(transformers, model_name)
---> 20     model = model_class.from_pretrained(
     21         pretrained_model_name_or_path=None,
     22         config=huggingface_config_path or model_type,
     23         num_labels=num_classes,
     24         state_dict=state_dict,
     25         local_files_only=huggingface_config_path is not None,
     26     )
     27     tokenizer = getattr(transformers, tokenizer_name).from_pretrained(
     28         huggingface_config_path or model_type,
     29         local_files_only=huggingface_config_path is not None,
     30         # TODO: may be needed to let it work with Kaggle competition
     31         # model_max_length=512,
     32     )
     34     return model, tokenizer

File ~/.conda/envs/py/lib/python3.9/site-packages/transformers/modeling_utils.py:2379, in PreTrainedModel.from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs)
   2369     if dtype_orig is not None:
   2370         torch.set_default_dtype(dtype_orig)
   2372     (
   2373         model,
   2374         missing_keys,
   2375         unexpected_keys,
   2376         mismatched_keys,
   2377         offload_index,
   2378         error_msgs,
-> 2379     ) = cls._load_pretrained_model(
   2380         model,
   2381         state_dict,
   2382         loaded_state_dict_keys,  # XXX: rename?
   2383         resolved_archive_file,
   2384         pretrained_model_name_or_path,
   2385         ignore_mismatched_sizes=ignore_mismatched_sizes,
   2386         sharded_metadata=sharded_metadata,
   2387         _fast_init=_fast_init,
   2388         low_cpu_mem_usage=low_cpu_mem_usage,
   2389         device_map=device_map,
   2390         offload_folder=offload_folder,
   2391         offload_state_dict=offload_state_dict,
   2392         dtype=torch_dtype,
   2393         load_in_8bit=load_in_8bit,
   2394     )
   2396 model.is_loaded_in_8bit = load_in_8bit
   2398 # make sure token embedding weights are still tied if needed

File ~/.conda/envs/py/lib/python3.9/site-packages/transformers/modeling_utils.py:2572, in PreTrainedModel._load_pretrained_model(cls, model, state_dict, loaded_keys, resolved_archive_file, pretrained_model_name_or_path, ignore_mismatched_sizes, sharded_metadata, _fast_init, low_cpu_mem_usage, device_map, offload_folder, offload_state_dict, dtype, load_in_8bit)
   2569                 del state_dict[checkpoint_key]
   2570     return mismatched_keys
-> 2572 folder = os.path.sep.join(resolved_archive_file[0].split(os.path.sep)[:-1])
   2573 if device_map is not None and is_safetensors:
   2574     param_device_map = expand_device_map(device_map, original_loaded_keys)

TypeError: 'NoneType' object is not subscriptable

pip install information:

Collecting detoxify
  Downloading detoxify-0.5.0-py3-none-any.whl (12 kB)
Collecting transformers!=4.18.0
  Downloading transformers-4.25.1-py3-none-any.whl (5.8 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.8/5.8 MB 75.2 MB/s eta 0:00:0000:0100:01
Collecting torch>=1.7.0
  Downloading torch-1.13.0-cp39-cp39-manylinux1_x86_64.whl (890.2 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 890.2/890.2 MB 3.6 MB/s eta 0:00:0000:0100:01
Collecting sentencepiece>=0.1.94
  Downloading sentencepiece-0.1.97-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.3 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.3/1.3 MB 107.1 MB/s eta 0:00:00
Collecting typing-extensions
  Downloading typing_extensions-4.4.0-py3-none-any.whl (26 kB)
Collecting nvidia-cuda-nvrtc-cu11==11.7.99
  Downloading nvidia_cuda_nvrtc_cu11-11.7.99-2-py3-none-manylinux1_x86_64.whl (21.0 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 21.0/21.0 MB 77.9 MB/s eta 0:00:0000:0100:01
Collecting nvidia-cublas-cu11==11.10.3.66
  Downloading nvidia_cublas_cu11-11.10.3.66-py3-none-manylinux1_x86_64.whl (317.1 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 317.1/317.1 MB 8.9 MB/s eta 0:00:0000:0100:01
Collecting nvidia-cuda-runtime-cu11==11.7.99
  Downloading nvidia_cuda_runtime_cu11-11.7.99-py3-none-manylinux1_x86_64.whl (849 kB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 849.3/849.3 kB 112.2 MB/s eta 0:00:00
Collecting nvidia-cudnn-cu11==8.5.0.96
  Downloading nvidia_cudnn_cu11-8.5.0.96-2-py3-none-manylinux1_x86_64.whl (557.1 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 557.1/557.1 MB 6.0 MB/s eta 0:00:0000:0100:01
Requirement already satisfied: wheel in /home/annahaz/.conda/envs/py/lib/python3.9/site-packages (from nvidia-cublas-cu11==11.10.3.66->torch>=1.7.0->detoxify) (0.37.1)
Requirement already satisfied: setuptools in /home/annahaz/.conda/envs/py/lib/python3.9/site-packages (from nvidia-cublas-cu11==11.10.3.66->torch>=1.7.0->detoxify) (63.4.1)
Collecting regex!=2019.12.17
  Downloading regex-2022.10.31-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (769 kB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 770.0/770.0 kB 116.5 MB/s eta 0:00:00
Requirement already satisfied: numpy>=1.17 in /home/annahaz/.conda/envs/py/lib/python3.9/site-packages (from transformers!=4.18.0->detoxify) (1.23.4)
Requirement already satisfied: tqdm>=4.27 in /home/annahaz/.conda/envs/py/lib/python3.9/site-packages (from transformers!=4.18.0->detoxify) (4.64.1)
Requirement already satisfied: pyyaml>=5.1 in /home/annahaz/.conda/envs/py/lib/python3.9/site-packages (from transformers!=4.18.0->detoxify) (6.0)
Collecting filelock
  Downloading filelock-3.8.2-py3-none-any.whl (10 kB)
Requirement already satisfied: packaging>=20.0 in /home/annahaz/.conda/envs/py/lib/python3.9/site-packages (from transformers!=4.18.0->detoxify) (21.3)
Requirement already satisfied: requests in /home/annahaz/.conda/envs/py/lib/python3.9/site-packages (from transformers!=4.18.0->detoxify) (2.28.1)
Collecting huggingface-hub<1.0,>=0.10.0
  Downloading huggingface_hub-0.11.1-py3-none-any.whl (182 kB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 182.4/182.4 kB 103.0 MB/s eta 0:00:00
Collecting tokenizers!=0.11.3,<0.14,>=0.11.1
  Downloading tokenizers-0.13.2-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (7.6 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 7.6/7.6 MB 33.4 MB/s eta 0:00:0000:0100:01m
Requirement already satisfied: pyparsing!=3.0.5,>=2.0.2 in /home/annahaz/.conda/envs/py/lib/python3.9/site-packages (from packaging>=20.0->transformers!=4.18.0->detoxify) (3.0.9)
Requirement already satisfied: certifi>=2017.4.17 in /home/annahaz/.conda/envs/py/lib/python3.9/site-packages (from requests->transformers!=4.18.0->detoxify) (2022.9.24)
Requirement already satisfied: charset-normalizer<3,>=2 in /home/annahaz/.conda/envs/py/lib/python3.9/site-packages (from requests->transformers!=4.18.0->detoxify) (2.1.1)
Requirement already satisfied: urllib3<1.27,>=1.21.1 in /home/annahaz/.conda/envs/py/lib/python3.9/site-packages (from requests->transformers!=4.18.0->detoxify) (1.26.12)
Requirement already satisfied: idna<4,>=2.5 in /home/annahaz/.conda/envs/py/lib/python3.9/site-packages (from requests->transformers!=4.18.0->detoxify) (3.4)
Installing collected packages: tokenizers, sentencepiece, typing-extensions, regex, nvidia-cuda-runtime-cu11, nvidia-cuda-nvrtc-cu11, nvidia-cublas-cu11, filelock, nvidia-cudnn-cu11, huggingface-hub, transformers, torch, detoxify
Successfully installed detoxify-0.5.0 filelock-3.8.2 huggingface-hub-0.11.1 nvidia-cublas-cu11-11.10.3.66 nvidia-cuda-nvrtc-cu11-11.7.99 nvidia-cuda-runtime-cu11-11.7.99 nvidia-cudnn-cu11-8.5.0.96 regex-2022.10.31 sentencepiece-0.1.97 tokenizers-0.13.2 torch-1.13.0 transformers-4.25.1 typing-extensions-4.4.0

additional information
python 3.9.13 haa1d7c7_2
on linux

[robvanvolt]

Me too... python 3.8 on Linux. Maybe this is related to the latest release? It worked fine before.

[user1342]

As a quick fix for this. On version 0.5.0, I was able to get around this error by commenting out line 2572 in modeling_utils.py (see this fork of transformers):

Replace:
folder = os.path.sep.join(resolved_archive_file[0].split(os.path.sep)[:-1])
with
#folder = os.path.sep.join(resolved_archive_file[0].split(os.path.sep)[:-1])

I'm unsure if this fix will have repercussions later down the line, however, for the time being it seems to be working.

[annabechang]

As a quick fix for this. On version 0.5.0, I was able to get around this error by commenting out line 2572 in modeling_utils.py (see this fork of transformers):

Replace: folder = os.path.sep.join(resolved_archive_file[0].split(os.path.sep)[:-1]) with #folder = os.path.sep.join(resolved_archive_file[0].split(os.path.sep)[:-1])

I'm unsure if this fix will have repercussions later down the line, however, for the time being it seems to be working.

thank you, this work for me as well!

[chrisji]

As a workaround, the v4.22.1 release of transformers appears to work as expected. So if installing via pip:

pip install detoxify
pip install transformers==4.22.1

(Other versions between the latest release and v4.22.1 may also work)

[laurahanu]

Hey all, thanks for raising this and good to see there's a workaround!

We will look into this asap!

[laurahanu]

Hey everyone, pinned the transformers version to 4.22.1 as a quick fix for now as we investigate this problem further, everything should hopefully work as expected if you reinstall the package/upgrade the transformers version!

[MartinPicc]

It is working for me with version 4.24.0, the next version however raises this error.
I believe it comes from this specific PR on transformers library merged in 4.25.1 huggingface/transformers#20321

However, it's not clear for me if this is an expected behavior on tranformers' side or if they introduced a bug

[MartinPicc]

It is apparently a bug on transformers side, and it should be fixed in next release with this PR: huggingface/transformers#21542

If you need a solution right now, you can also install transformers from their main branch directly, the PR is merged already