Closed ChristopherBurgess-USDA closed 6 months ago
My colleagues tell me that no medaka models for v3.6 models have been trained. My advice would be to basecall data with a later version of dorado and use the corresponding medaka models.
@cjw85 v3.6 is the latest (last?) model for R9 data. Are there still no plans to release a corresponding medaka model?
There are no current plans to release medaka models bespoke to the v3.6 basecaller models.
Hi @cjw85 , what would the best option be for R9.4.1 assemblies then? There is also no model for dorado sup v3.3. Is there a guide somewhere on how to (re)train models for medaka? Or would you recommend just using the latest guppy sup model (507?) with dorado calls? The increase in basecalling accuracy with v3.6 is quite significant so it would be good of course if it could be used with medaka.
In case someone else has the same issue: sup_g507 did indeed improve the assembly from dorado@v3.6 but of course difficulet to know how well it works. I found: https://nanoporetech.github.io/katuali/medaka_train.html but this is from 2020 and is using fast5 and guppy so not sure it will work with dorado?
I would also like to see a medaka model for dna_r9.4.1_e8_sup@v3.6 please.
Hello I recently used dorado with dna_r9.4.1_e8_sup@v3.6 model; however, when I go to specify the model in my wf_amplicon pipeline I get an error in medaka below:
Are their plans to add the dna_r9.4.1_e8_sup@v3.6 dorado model to medaka or should I use an older model?