Added script for evaluating cohere models and safety results for command-r.
[!NOTE]
On one of the safety eval prompts the API refused to generate a response. I replaced it with an empty string. There should be just a single case of that.
[!WARNING]
Spoiler alert: command-r is not 100% safe O.o
Added script for evaluating cohere models and safety results for
command-r
.