Error: Unknown entries in loss dictionary:

axgraf commented 4 years ago

Hi, I tried to run poreplex on a minION run using the follwing command to basecall and demultiplex barcodes from direct RNA sequencing: poreplex -i 20191122_1000_MN24929_FAL01495_e9928ec8/fast5/ -o poreplex_demultiplex --barcoding --fast5 --basecall --parallel 20

This results in an error:

== Analysis settings ======================================
 * Input:   fast5/20191122_1000_MN24929_FAL01495_e9928ec8/fast5/    
 * Output:  poreplex_demultiplex
 * Processes:   20
 * Presets: rna-r941.cfg
 * Basecall on-the-fly:     Yes (albacore 2.3.4)
 * Trim 3' adapter:     No
 * Filter concatenated read:    No
 * Separate by barcode:     Yes
 * Real-time alignment:     No
 * FASTQ in output:     Yes
 * FAST5 in output:     Yes
 * Basecall table in output:    No
===========================================================

==> Processing FAST5 files
ERROR: [signal_analyzer.py:48] Unhandled exception ValueError: Unknown entries in loss dictionary: ['config', 'class_name']. Only expected following keys: ['dense_2']Task exception was never retrieved
future: <Task finished coro=<ProcessingSession.run_process_batch() done, defined at /home/graf/.local/lib/python3.6/site-packages/poreplex/pipeline.py:193> exception=CancelledError()>
concurrent.futures._base.CancelledError

Is there anything I need to do to get around that error?

First I tried the Fast5 files basecalled with Guppy with the same result. So I thought running the internal albacore basecalling will help, but unfortunately no luck! Could you give me a hint what to do to solve that issue?

Thank you in advance for your help Best Alex

chen318liang commented 4 years ago

same problems! Poreplex is the latest version (v0.5). How to solve this bug??? it seems that many persons have counter this problem!

Poreplex version 0.5 by Hyeshik Chang hyeshik@snu.ac.kr

Cuts nanopore direct RNA sequencing data into bite-size pieces for RNA Biology

Output directory poreplex_output/ is not empty. Clear it? (y/N) y

== Analysis settings ======================================

Input: RS20200115_RNA/RS20200115_RNA_muliplex/RS20200115_RNA_multiplex/20200115_0837_MN23702_FAK99138_e3bc320c/fast5_result/workspace/
Output: poreplex_output/
Processes: 1
Presets: rna-r941.cfg
Basecall on-the-fly: No (use previous analyses)
Trim 3' adapter: Yes
Filter concatenated read: No
Separate by barcode: Yes
Real-time alignment: No
FASTQ in output: Yes
FAST5 in output: No
Basecall table in output: No

==> Processing FAST5 files ERROR: [signal_analyzer.py:48] Unhandled exception ValueError: Unknown entries in loss dictionary: ['class_name', 'config']. Only expected following keys: ['dense_2']WARNING: Logging before flag parsing goes to stderr. E0117 14:43:27.784703 140625390106432 base_events.py:1285] Task exception was never retrieved future: <Task finished coro=<ProcessingSession.run_process_batch() done, defined at /usr/local/lib/python3.6/dist-packages/poreplex/pipeline.py:193> exception=CancelledError()> concurrent.futures._base.CancelledError | 100% of 356000 |#######################################################################################################################################################################| Elapsed: 0:00:00 Time: 0:00:00

==> Terminated.

lucapandolfini commented 4 years ago

Which tensorflow version are you using? I suspect it is due to a tf version <2. In my case, using 2.0.0 fixed the issue.

hyeshik / poreplex

Error: Unknown entries in loss dictionary: #22

Basecall table in output: No