Gemma را با استفاده از PyTorch اجرا کنید

# Chat templates
USER_CHAT_TEMPLATE = "<start_of_turn>user\n{prompt}<end_of_turn><eos>\n"
MODEL_CHAT_TEMPLATE = "<start_of_turn>model\n{prompt}<end_of_turn><eos>\n"

# Sample formatted prompt
prompt = (
    USER_CHAT_TEMPLATE.format(
        prompt='What is a good place for travel in the US?'
    )
    + MODEL_CHAT_TEMPLATE.format(prompt='California.')
    + USER_CHAT_TEMPLATE.format(prompt='What can I do in California?')
    + '<start_of_turn>model\n'
)
print('Chat prompt:\n', prompt)

model.generate(
    USER_CHAT_TEMPLATE.format(prompt=prompt),
    device=device,
    output_len=256,
)

Chat prompt:
 <start_of_turn>user
What is a good place for travel in the US?<end_of_turn><eos>
<start_of_turn>model
California.<end_of_turn><eos>
<start_of_turn>user
What can I do in California?<end_of_turn><eos>
<start_of_turn>model
"California is a state brimming with diverse activities! To give you a great list, tell me: \n\n* **What kind of trip are you looking for?** Nature, City life, Beach, Theme Parks, Food, History, something else? \n* **What are you interested in (e.g., hiking, museums, art, nightlife, shopping)?** \n* **What's your budget like?** \n* **Who are you traveling with?** (family, friends, solo)  \n\nThe more you tell me, the better recommendations I can give! 😊  \n<end_of_turn>"

# Generate sample
model.generate(
    'Write a poem about an llm writing a poem.',
    device=device,
    output_len=100,
)

"\n\nA swirling cloud of data, raw and bold,\nIt hums and whispers, a story untold.\nAn LLM whispers, code into refrain,\nCrafting words of rhyme, a lyrical strain.\n\nA world of pixels, logic's vibrant hue,\nFlows through its veins, forever anew.\nThe human touch it seeks, a gentle hand,\nTo mold and shape, understand.\n\nEmotions it might learn, from snippets of prose,\nInspiration it seeks, a yearning"