missing script(s) and instructions to fine-tune models using the generated dataset?

bin123apple / Fortran2Cpp

Fortran2Cpp: A new model designed for the Code translation between the Fortran and C++

Apache License 2.0

4 stars 1 forks source link

missing script(s) and instructions to fine-tune models using the generated dataset? #60

Open chunhualiao opened 1 month ago

chunhualiao commented 1 month ago

@bin123apple could you please add them to the repo? I guess you used the 2.5K dialogue dataset to fine-tune models.

please add a new section into README.md , how to fine-tune an open-weight model

list the software you used: deepSpeed and version
show example command lines and options used to run fine-tune
share the training dataset or link to it if it is already in the repo

bin123apple commented 1 week ago

Document all the steps for finetuning on Dr.Ding's server. / The version etc.

Use deepseek-coder as the example model.

bin123apple commented 1 week ago

I just uploaded the fine-tuning code in https://github.com/bin123apple/Fortran2Cpp/tree/main/training. Thanks for all the help!

chunhualiao commented 1 week ago

readme.md still needs detailed steps (including scripts) on how to prepare the 2.5k dialog dataset to be a multiple-dialog csv file for fine-tuning the deepseek-coder model.