This model is useless because it has an english brain but got the italian accent from the finetuning. It's my 5th try and for the first time is printing out words inside the moshi GUI.
All the previous finetuning did not produce any word outputs. It kinda works, there is a lot of work to do. I guess you can try it for fun
LoRA finetune of
kyutai/moshiko
for
Italian full-duplex spoken dialogue, in the style of
J-Moshi.
Early research checkpoint: the model acquired Italian phonetics and prosody,
but sentence-level coherence and turn-taking are still weak. Treat it as a
research artifact, not a usable assistant.
Inference runs with the official
moshi
package by Kyutai (this is a different repo from
moshi-finetune, which is only
needed for training). A CUDA GPU with ~16 GB of VRAM is recommended.
On Windows, run everything inside WSL.
1# 1. Clone the inference repo and install the moshi package
2git clone https://github.com/kyutai-labs/moshi.git
3cd moshi
4pip install -e moshi # the Python package lives in the moshi/ subdirectory
5
6# 2. Download the LoRA adapter from this repo
7pip install huggingface_hub
8huggingface-cli download <TUO_USERNAME>/i-moshi --local-dir ./i-moshi
9
10# 3. Launch the web UI — the LoRA adapter is merged on top of the base
11# moshiko weights at load time (base weights are auto-downloaded from HF)
12CUDA_VISIBLE_DEVICES=0 python -m moshi.server \
13 --lora-weight="./i-moshi/lora.safetensors" \
14 --config-path="./i-moshi/config.json"
Then open
http://localhost:8998 in your browser and start talking.
