Output adata files missing vars

[GriffinBINF]

I have previously had great success with cellbender but recently and inexplicably all of my cellbender outputs have been missing the genetic features entirely. Furthermore there appears to be a discrepancy between the metrics file and the output files.

Here's an example of the metrics.csv file for one of my runs:

total_raw_counts	76506168.000

1 total_output_counts 65411847.000
2 total_counts_removed 11094321.000
3 fraction_counts_removed 0.145
4 total_raw_counts_in_cells 76506152.000
5 total_counts_removed_from_cells 11094305.000
6 fraction_counts_removed_from_cells 0.145
7 average_counts_removed_per_cell 590.311
8 target_fpr 0.010
9 expected_cells 6050.000
10 found_cells 18794.000
11 output_average_counts_per_cell 3480.464
12 ratio_of_found_cells_to_expected_cells 3.106
13 found_empties 31206.000
14 fraction_of_analyzed_droplets_that_are_nonempty 0.376
15 convergence_indicator 0.473
16 overall_change_in_train_elbo 1302.979

The input adata for this file is

AnnData object with n_obs × n_vars = 737280 × 99522
var: 'gene_id', 'gene_name'
layers: 'ambiguous', 'spliced', 'unspliced'

However the output h5 file and the filtered h5 for this dataset is respectively

AnnData object with n_obs × n_vars = 737280 × 0
var: 'gene_ids', 'feature_types', 'genome'

AnnData object with n_obs × n_vars = 18794 × 0
var: 'gene_ids', 'feature_types', 'genome'

There are no apparent issues in the log file and the metrics file indicates that the filtering is not being overly aggressive, at least not in a way that explains the lack of vars columns in the output.

Please let me know if there is anything I can do to troubleshoot this issue.

Cheers,
Griffin