codon sequence optimization for multiple species

[chaofan520]

Describe the solution you'd like
Can CodonTransformer perform codon sequence optimization for multiple species simultaneously? How should this be done?

Additional context

# Load model and tokenizer
tokenizer = AutoTokenizer.from_pretrained("adibvafa/CodonTransformer")
model = BigBirdForMaskedLM.from_pretrained("adibvafa/CodonTransformer").to(device)


# Set your input data
protein = "MALWMRLLPLLALLALWGPDPAAAFVNQHLCGSHLVEALYLVCGERGFFYTPKTRREAEDLQVGQVELGG"
organism = "Escherichia coli general" + "Saccharomyces cerevisiae"

[chaofan520]

@Adibvafa

[Adibvafa]

@chaofan520
Hello! Thank you for opening this PR.
It's technically possible but needs a bit of coding. What determins the organism is an embedding vector. You can technically get embedding vector of two different organisms, average them (or sum them, which ever works best) and use that as your organism embedding.