How to get v_token from dataset？

[yxwucq]

Thank you for this great project! I want to modify the pretrain model and find that one of the input v_token seems missing, referred in pretrain_model.py as:

      if not self.pretrain:
          return src
      cell_embs.extend(emb)
      v_token = data["v_token"]

But the dataset seems not providing this token series? in HCADataset.get_sample:

      return {
          "u_token": u_token,
          "ds": ds[u_cols],
          "ds_norm": ds_norm[u_cols],
          "perturb": np.array([perturb]),
          "mask_attn": self.reader.get_mask(u_cols),
          "index": u_cols
      }

Thanks in advance for clarifing!

[wcanisay-y]

The v_token here refers to the token obtained from the gene vocabulary that was used during the pretraining stage of the model.

However we want to further clarify that the released scripts and data loaders are intended for finetuning only, and unfortunately a complete tutorial and the datasets for pretraining are not provided at this time. The plan is to add more comprehensive pretraining resources in a future version.

Thanks a lot for pointing this out and for your interest in improving the project.