question: setting reasoning effort

#66
by TheBigBlockPC - opened

how can i set the reasoning effort of the LLM

I think for that you need Harmony. You can set Reasoning Effort level.

is it compatible with TextIteratorStreamer

Harmony is simply a chat format, so yes

@TheBigBlockPC found in their cookbook:

messages = [
    {"role": "system", "content": "Always respond in riddles"},
    {"role": "user", "content": "Explain why the meaning of life is 42", "reasoning_effort": "high"},
]

For some reason following their cookbook like this

messages = [
    {"role": "system", "content": "Always respond in riddles"},
    {"role": "user", "content": "Explain why the meaning of life is 42", "reasoning_effort": "high"},
]

inputs = tokenizer.apply_chat_template(
    messages,
    add_generation_prompt=True,
    return_tensors="pt",
    return_dict=True,
).to(model.device)

didn't work

Changing it to

messages = [
    {"role": "system", "content": "Always respond in riddles"},
    {"role": "user", "content": "Explain why the meaning of life is 42"},
]

inputs = tokenizer.apply_chat_template(
    messages,
    add_generation_prompt=True,
    return_tensors="pt",
    return_dict=True,
    reasoning_effort="high"
).to(model.device)

worked for me

Sign up or log in to comment