Result

Generate unconditional sample

Unconditional sample refers to randomly generate sample without taking into account any user input. Think of it as random sample. Make sure you are in the src directory and type the following in the command prompt:

$ python generate_unconditional_samples --model_name lyric

Wait for it to run and you should be able to see samples being generated one by one. It will keep on running until you stop it with Ctrl+C.

The most important parameters are top_k and temperate.

• top_k: Integer value controlling diversity. 1 means only 1 word is considered for each step (token), resulting in deterministic completions, while 40 means 40 words are considered at each step. 0 (default) is a special setting meaning no restrictions. 40 generally is a good value.

• temperature: Float value controlling randomness in boltzmann distribution. Lower temperature results in less random completions. As the temperature approaches zero, the model will become deterministic and repetitive. Higher temperature results in more random completions. Default value is 1.

Generate interactive conditional sample

Interactive conditional sample refers to generate samples based on user input. In other words, you input some text and GPT-2 will do its best to fill in the rest. The command and parameters available is the same as that of unconditional sample. Do not enter your input together in the command as you will be prompt for the input later on in the command prompt. Type the following command:

$ python interactive_conditional_samples --temperature 0.8 --top_k 40 --model_name lyric