ctheodoris

/

Geneformer

like 197

Hi, I encountered an error when running the examples provided (in silico perturbation), during the "isp_perturb_all". The error message is pasted below. When I look into this error, it seems that the problem comes from the different tensor shapes at cos_sim_shift function (Line 581, 592).

I changed Line 584 from
state_embs_dict[state].to("cuda"), # required to move to cuda here
to
state_embs_dict[state].to("cuda").unsqueeze(0), # required to move to cuda here

And it seems to solve the problem (btw, I have to change cuda to cpu in source code as I tested in a CPU environment). However I don't know if the same error happens to others, so I didn't send a push request.

IndexError Traceback (most recent call last)
Cell In[20], line 1
----> 1 isp.isp_perturb_all(
2 wl_model, filtered_input_data, zlayer_to_quant, "insilico_out/out_perturbed_wl"
3 )

File ~/miniconda3/envs/pytorch/lib/python3.11/site-packages/geneformer/in_silico_perturber.py:785, in InSilicoPerturber.isp_perturb_all(self, model, filtered_input_data, layer_to_quant, output_path_prefix)
778 original_cell_emb = pu.compute_nonpadded_cell_embedding(
779 full_original_emb, "mean_pool"
780 )
781 perturbation_cell_emb = pu.compute_nonpadded_cell_embedding(
782 full_perturbation_emb, "mean_pool"
783 )
--> 785 cell_cos_sims = pu.quant_cos_sims(
786 perturbation_cell_emb,
787 original_cell_emb,
788 self.cell_states_to_model,
789 self.state_embs_dict,
790 emb_mode="cell",
791 )
793 if self.emb_mode == "cell_and_gene":
794 # remove perturbed index for gene list
795 perturbed_gene_dict = {
796 gene: gene_list[:i] + gene_list[i + 1 :]
797 for i, gene in enumerate(gene_list)
798 }

File ~/miniconda3/envs/pytorch/lib/python3.11/site-packages/geneformer/perturber_utils.py:584, in quant_cos_sims(perturbation_emb, original_emb, cell_states_to_model, state_embs_dict, emb_mode)
582 cos_sims = dict(zip(possible_states, [[] for _ in range(len(possible_states))]))
583 for state in possible_states:
--> 584 cos_sims[state] = cos_sim_shift(
585 original_emb,
586 perturbation_emb,
587 #state_embs_dict[state].to("cuda"), # required to move to cuda here
588 state_embs_dict[state].to("cpu"), # required to move to cuda here
589 cos,
590 )
592 return cos_sims

File ~/miniconda3/envs/pytorch/lib/python3.11/site-packages/geneformer/perturber_utils.py:597, in cos_sim_shift(original_emb, perturbed_emb, end_emb, cos)
596 def cos_sim_shift(original_emb, perturbed_emb, end_emb, cos):
--> 597 origin_v_end = cos(original_emb, end_emb)
598 perturb_v_end = cos(perturbed_emb, end_emb)
600 return perturb_v_end - origin_v_end

File ~/miniconda3/envs/pytorch/lib/python3.11/site-packages/torch/nn/modules/module.py:1518, in Module._wrapped_call_impl(self, *args, **kwargs)
1516 return self._compiled_call_impl(*args, **kwargs) # type: ignore[misc]
1517 else:
-> 1518 return self._call_impl(*args, **kwargs)

File ~/miniconda3/envs/pytorch/lib/python3.11/site-packages/torch/nn/modules/module.py:1527, in Module._call_impl(self, *args, **kwargs)
1522 # If we don't have any hooks, we want to skip the rest of the logic in
1523 # this function, and just call forward.
1524 if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks
1525 or _global_backward_pre_hooks or _global_backward_hooks
1526 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1527 return forward_call(*args, **kwargs)
1529 try:
1530 result = None

File ~/miniconda3/envs/pytorch/lib/python3.11/site-packages/torch/nn/modules/distance.py:87, in CosineSimilarity.forward(self, x1, x2)
86 def forward(self, x1: Tensor, x2: Tensor) -> Tensor:
---> 87 return F.cosine_similarity(x1, x2, self.dim, self.eps)

IndexError: Dimension out of range (expected to be in range of [-1, 0], but got 1)