Loading a finetuned model

[kenenbek]

Are these steps correct to use a finetuned model?

• When I finished finetuning with lora.py I got config.json and model.safetensors
• I run loadandmergecheckpoint.py file and got another model.safetensors file

For inference I used load_csm_1b_local function:

generator = load_csm_1b_local(model_path="/mnt/d/csm-chekpoints/", device="cuda")

text = "Мени азыр угуп жатасыздарбы?"

generate_streaming_audio(
    generator=generator,
    text=text,
    speaker=0,
    context=[],  # No context needed for basic generation
    output_file="streaming_audio.wav",
)