vocab not being written correctly #27
Assignees
Comments
ajtritt
added a commit
that referenced
this issue
Mar 2, 2021
ajtritt
added a commit
that referenced
this issue
Mar 5, 2021
* add function to extract profile data * add assert to catch bad vocab. See #27 * add command to give dataset info * Clean up some things - clean up argparse arguments - use CSV logger - always use DDP. Remove SLURM/LSF support (until PL stabilizes) * clean up job sumission * update base model to work with PL * improve chunking efficiency * add command to exec * remove name from environment * remove read of downsample
ajtritt
added a commit
that referenced
this issue
Mar 5, 2021
* add function to extract profile data * add assert to catch bad vocab. See #27 * add command to give dataset info * Clean up some things - clean up argparse arguments - use CSV logger - always use DDP. Remove SLURM/LSF support (until PL stabilizes) * clean up job sumission * update base model to work with PL * improve chunking efficiency * add command to exec * remove name from environment * remove read of downsample * add classifier for ResNet feature models * add options for using classifier * add sensible statement for recurring error * work out kinks in ResNet classifier * add options for job submission script * set model subdirectory if starting classification from features * Resume from checkpoint of ResNetClassifier - Use LightningDataModule so we can determine number of model outputs before initializing model * clean up to make checkpointing and adding classifer work * Clean up for Cori (#28) * add function to extract profile data * add assert to catch bad vocab. See #27 * add command to give dataset info * Clean up some things - clean up argparse arguments - use CSV logger - always use DDP. Remove SLURM/LSF support (until PL stabilizes) * clean up job sumission * update base model to work with PL * improve chunking efficiency * add command to exec * remove name from environment * remove read of downsample * add arguments to job runner * add classifier for ResNet feature models * add options for using classifier * add sensible statement for recurring error * work out kinks in ResNet classifier * add options for job submission script * set model subdirectory if starting classification from features * Resume from checkpoint of ResNetClassifier - Use LightningDataModule so we can determine number of model outputs before initializing model * clean up to make checkpointing and adding classifer work * add arguments to job runner * add field that was missing after restarting from checkpoint
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
In a DeepIndexFile that is currently sitting on Cori at
$CSCRATCH/exabiome/deep-index/input/gtdb/r95/ar122_r95.input.h5, the vocabulary attribute on seq_table/sequence was [A T C G N], when it should be [A C Y W S K D V N T G R W S M H B N].This attribute should get set during conversion here:
deep-taxon/src/exabiome/gtdb/prepare_data.py
Line 306 in d4ddbf3
According to this line, the correct vocabulary should be returned:
deep-taxon/src/exabiome/sequence/convert.py
Lines 354 to 355 in d7f54dd
The text was updated successfully, but these errors were encountered: