ctheodoris

/

Geneformer

like 200

Hi, I found a bug when I tried to load the current gene embeddings function:

TypeError                                 Traceback (most recent call last)
Cell In[1], line 1
----> 1 from geneformer_new import EmbExtractor

File /gpfs/gibbs/pi/zhao/tl688/Geneformer/geneformer_new/__init__.py:4
      2 from . import pretrainer
      3 from . import collator_for_classification
----> 4 from . import in_silico_perturber
      5 from . import in_silico_perturber_stats
      6 from .tokenizer import TranscriptomeTokenizer

File /gpfs/gibbs/pi/zhao/tl688/Geneformer/geneformer_new/in_silico_perturber.py:40
     37 from datasets import Dataset
     38 from tqdm.auto import trange
---> 40 from . import perturber_utils as pu
     41 from .emb_extractor import get_embs
     42 from .tokenizer import TOKEN_DICTIONARY_FILE

File /gpfs/gibbs/pi/zhao/tl688/Geneformer/geneformer_new/perturber_utils.py:278
    270     return torch.stack(output_batch_list_padded)
    273 # removes perturbed indices
    274 # need to handle the various cases where a set of genes is overexpressed
    275 def remove_perturbed_indices_set(
    276     emb,
    277     perturb_type: str,
--> 278     indices_to_perturb: list[list],
    279     tokens_to_perturb: list[list],
    280     original_lengths: list[int],
    281     input_ids=None,
    282 ):
    283     if perturb_type == "overexpress":
    284         num_perturbed = len(tokens_to_perturb)

TypeError: 'type' object is not subscriptable

Could you please help me address it? Thanks.