My goal is to use GPT-2 for Midi Data - like museNet.
The data input are numbers like as follows:
23 24 177 12 23 12 15
i trained the model from scratch described in this post: https://discuss.huggingface.co/t/gpt2-training-from-scratch-in-german/1157
While using pipeline(“text-generation”), the output length is very short. Is this due to the data input? how can i force a min_length with pipeline?