Commit ca2cd16d authored by sanchit-gandhi's avatar sanchit-gandhi
Browse files

fix prompt for subset

parent d175b832
...@@ -330,13 +330,12 @@ def main(): ...@@ -330,13 +330,12 @@ def main():
PROMPT = """ We have seven keywords that describe different attributes of an audio sample spoken by a given speaker: the speaker's gender, the speaker's accent, the amount of reverberation in the sample (high or low reverberation), the amount of noise in the sample (how clear or noisy), how monotone or animated the sample is, the speaker's pitch (high or low voice), the speaker's speed (how fast or slow the speaker is speaking). PROMPT = """ We have seven keywords that describe different attributes of an audio sample spoken by a given speaker: the speaker's gender, the speaker's accent, the amount of reverberation in the sample (high or low reverberation), the amount of noise in the sample (how clear or noisy), how monotone or animated the sample is, the speaker's pitch (high or low voice), the speaker's speed (how fast or slow the speaker is speaking).
Given these keywords, form a coherent sentence that summarises the seven attributes in a meaningful way. You can change the order of the keywords in the sentence and use common synonyms for these words, provided that the sentence summarises the attributes clearly. Keep the sentence simple - don't introduce additional information other than the keywords provided. Only return the generated sentence, not any other assistant remarks. Given these keywords, form a coherent sentence that summarises the seven attributes in a meaningful way. You can change the order of the keywords in the sentence and use common synonyms for these words, provided that the sentence summarises the attributes clearly. Keep the sentence simple - don't introduce additional information other than the keywords provided. Only return the generated sentence, not any other assistant remarks.
For example, given the following descriptors: 'female', 'Hungarian', 'slightly roomy sounding', 'fairly noisy', 'quite monotone', 'fairly low pitch', 'very slowly', a valid sentence would be: 'a woman with a deep voice speaking slowly and somewhat monotonously with a Hungarian accent in an echoey room with background noise'. Note how the seven attributes have been combined together in a simple sentence, with the ordering changed but no additional information added. For example, given the following descriptors: 'female', 'Hungarian', 'slightly roomy sounding', 'fairly noisy', 'quite monotone', 'fairly low pitch', 'very slowly', a valid sentence would be: 'a woman with a deep voice speaking slowly and somewhat monotonously with a Hungarian accent in an echoey room with background noise'. Note how the seven attributes have been combined together in a simple sentence, with the ordering changed but no additional information added.
For the descriptors: {gender}, {accent}, {reverberation}, {noise}, {monotony}, {pitch}, {speaking_rate}, the corresponding sentence is:""" For the descriptors: {gender}, {accent}, {reverberation}, {noise}, {speech_monotony}, {pitch}, {speaking_rate}, the corresponding sentence is:"""
SUBSET_PROMPT = """ We have six keywords that describe different attributes of an audio sample spoken by a given speaker: the speaker's gender, the amount of reverberation in the sample (high or low reverberation), the amount of noise in the sample (how clear or noisy), how monotone or animated the sample is, the speaker's pitch (high or low voice), the speaker's speed (how fast or slow the speaker is speaking). SUBSET_PROMPT = """ We have six keywords that describe different attributes of an audio sample spoken by a given speaker: the speaker's gender, the amount of reverberation in the sample (high or low reverberation), the amount of noise in the sample (how clear or noisy), how monotone or animated the sample is, the speaker's pitch (high or low voice), the speaker's speed (how fast or slow the speaker is speaking).
Given these keywords, form a coherent sentence that summarises the six attributes in a meaningful way. You can change the order of the keywords in the sentence and use common synonyms for these words, provided that the sentence summarises the attributes clearly. Keep the sentence simple - don't introduce additional information other than the keywords provided. Only return the generated sentence, not any other assistant remarks. Given these keywords, form a coherent sentence that summarises the six attributes in a meaningful way. You can change the order of the keywords in the sentence and use common synonyms for these words, provided that the sentence summarises the attributes clearly. Keep the sentence simple - don't introduce additional information other than the keywords provided. Only return the generated sentence, not any other assistant remarks.
For example, given the following descriptors: 'female', 'slightly roomy sounding', 'fairly noisy', 'quite monotone', 'fairly low pitch', 'very slowly', a valid sentence would be: 'a woman with a deep voice speaking slowly and somewhat monotonously in an echoey room with background noise'. Note how the six attributes have been combined together in a simple sentence, with the ordering changed but no additional information added. For example, given the following descriptors: 'female', 'slightly roomy sounding', 'fairly noisy', 'quite monotone', 'fairly low pitch', 'very slowly', a valid sentence would be: 'a woman with a deep voice speaking slowly and somewhat monotonously in an echoey room with background noise'. Note how the six attributes have been combined together in a simple sentence, with the ordering changed but no additional information added.
For the descriptors: {gender}, {accent}, {reverberation}, {noise}, {monotony}, {pitch}, {speaking_rate}, the corresponding sentence is:""" For the descriptors: {gender}, {reverberation}, {noise}, {speech_monotony}, {pitch}, {speaking_rate}, the corresponding sentence is:"""
def prepare_dataset(sample): def prepare_dataset(sample):
sample_prompt = SUBSET_PROMPT sample_prompt = SUBSET_PROMPT
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment