[Question]How to use local TITAN in TRIDENT？ #118
Description
What's your question?
Thank you very much for developing Trident—it’s fantastic. I carefully read through issue #96 and your Q&A there. I’m experiencing the exact same problem, but no matter what I try I keep getting errors. I’d like to know what I should do. Below I’ll provide my error message and directory structure.
I downloaded all of the Titan model files from Hugging Face into ./GC_PDL1/Model/titan. Then, in
trident/patch_encoder_models/local_ckpts.json (line 21) I changed the path to "./GC_PDL1/Model/titan/conch_v15"
Additionally, in trident/slide_encoder_models/load.py (line 428) I modified the code to: "model = AutoModel.from_pretrained('./GC_PDL1/Model/titan', trust_remote_code=True)``
python run_batch_of_slides.py --task feat --wsi_dir ./wsis --job_dir ./trident_processed --slide_encoder titan --mag 20 --patch_size 512
However, I still keep encountering errors.
[PROCESSOR] Found 72 valid slides in /baksv/DPH/DPYY_zhangwj/GC_PDL1/Data/IHC_2.
Traceback (most recent call last):
File "/baksv/DPH/DPYY_zhangwj/GC_PDL1/Code/run_batch_of_slides.py", line 278, in
main()
File "/baksv/DPH/DPYY_zhangwj/GC_PDL1/Code/run_batch_of_slides.py", line 274, in main
run_task(processor, args)
File "/baksv/DPH/DPYY_zhangwj/GC_PDL1/Code/run_batch_of_slides.py", line 201, in run_task
encoder = encoder_factory(args.slide_encoder)
File "/baksv/DPH/DPYY_zhangwj/trident/trident/slide_encoder_models/load.py", line 49, in encoder_factory
return enc(pretrained=pretrained, freeze=freeze, **kwargs)
File "/baksv/DPH/DPYY_zhangwj/trident/trident/slide_encoder_models/load.py", line 422, in init
super().init(**build_kwargs)
File "/baksv/DPH/DPYY_zhangwj/trident/trident/slide_encoder_models/load.py", line 74, in init
self.model, self.precision, self.embedding_dim = self._build(**build_kwargs)
File "/baksv/DPH/DPYY_zhangwj/trident/trident/slide_encoder_models/load.py", line 428, in _build
model = AutoModel.from_pretrained('/baksv/DPH/DPYY_zhangwj/GC_PDL1/Model/titan', trust_remote_code=True)
File "/baksv/DPH/DPYY_zhangwj/anaconda3/envs/trident/lib/python3.10/site-packages/transformers/models/auto/auto_factory.py", line 444, in from_pretrained
return model_class.from_pretrained(pretrained_model_name_or_path, *model_args, config=config, **kwargs)
File "/baksv/DPH/DPYY_zhangwj/anaconda3/envs/trident/lib/python3.10/site-packages/transformers/modeling_utils.py", line 1493, in from_pretrained
model = cls(config, *model_args, **model_kwargs)
File "/public/home/DPYY_zhangwj/.cache/huggingface/modules/transformers_modules/local/modeling_titan.py", line 17, in init
self.text_encoder = build_text_tower(config.text_config)
File "/public/home/DPYY_zhangwj/.cache/huggingface/modules/transformers_modules/local/text_transformer.py", line 500, in build_text_tower
text = TextTransformer(
File "/public/home/DPYY_zhangwj/.cache/huggingface/modules/transformers_modules/local/text_transformer.py", line 410, in init
self.tokenizer = Conch_Tokenizer(context_length=context_length)
File "/public/home/DPYY_zhangwj/.cache/huggingface/modules/transformers_modules/local/conch_tokenizer.py", line 17, in init
self.tokenizer = PreTrainedTokenizerFast.from_pretrained("MahmoodLab/TITAN")
File "/baksv/DPH/DPYY_zhangwj/anaconda3/envs/trident/lib/python3.10/site-packages/transformers/tokenization_utils_base.py", line 1763, in from_pretrained
raise err
File "/baksv/DPH/DPYY_zhangwj/anaconda3/envs/trident/lib/python3.10/site-packages/transformers/tokenization_utils_base.py", line 1724, in from_pretrained
resolved_vocab_files[file_id] = cached_path(
File "/baksv/DPH/DPYY_zhangwj/anaconda3/envs/trident/lib/python3.10/site-packages/transformers/file_utils.py", line 1921, in cached_path
output_path = get_from_cache(
File "/baksv/DPH/DPYY_zhangwj/anaconda3/envs/trident/lib/python3.10/site-packages/transformers/file_utils.py", line 2125, in get_from_cache
_raise_for_status(r)
File "/baksv/DPH/DPYY_zhangwj/anaconda3/envs/trident/lib/python3.10/site-packages/transformers/file_utils.py", line 2052, in _raise_for_status
request.raise_for_status()
File "/baksv/DPH/DPYY_zhangwj/anaconda3/envs/trident/lib/python3.10/site-packages/requests/models.py", line 1026, in raise_for_status
raise HTTPError(http_error_msg, response=self)
requests.exceptions.HTTPError: 401 Client Error: Unauthorized for url: https://hf-mirror.com/MahmoodLab/TITAN/resolve/main/tokenizer.json