Can ct2 be used to convert a fine-tuned version of starcoder?
#4
by
Hodophile
- opened
Hello, I fine-tuned the original starcoder model with my own repository and want to adapt it to the quantized version after fine-tuning. Due to extra parameters adapted during fine-tuning, the original conversion code fails as expected. What could I do to fix this issue?
Thanks in advance.
Yes, you can start also with other models that are in GPTBigCode format.
Is this command failing?
ct2-transformers-converter --model bigcode/starcoder --output_dir ~/tmp-ct2fast-starcoder --force --copy_files merges.txt tokenizer.json README.md tokenizer_config.json vocab.json generation_config.json special_tokens_map.json .gitattributes --quantization int8_float16 --trust_remote_code