From 70bf283dbc90ce12031db78a25634196412ede72 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Alexandre=20D=C3=A9fossez?= Date: Wed, 18 Sep 2024 15:08:43 +0200 Subject: [PATCH] new repo --- README.md | 18 +++++++++--------- moshi/moshi/models/loaders.py | 2 +- moshi/moshi/server.py | 2 +- moshi_mlx/README.md | 4 ++-- moshi_mlx/moshi_mlx/local.py | 6 +++--- moshi_mlx/moshi_mlx/local_web.py | 8 ++++---- rust/moshi-backend/config-q8.json | 2 +- rust/moshi-backend/config.json | 2 +- rust/moshi-backend/src/main.rs | 2 +- 9 files changed, 23 insertions(+), 23 deletions(-) diff --git a/README.md b/README.md index 69a672b..3e413c0 100644 --- a/README.md +++ b/README.md @@ -64,12 +64,12 @@ We release three models: Depending on the backend, the file format and quantization available will vary. Here is the list of the HuggingFace repo with each model. Mimi is bundled in each of those, and always use the same checkpoint format. -- Moshika for PyTorch (bf16): [kmhf/moshika-pytorch-bf16](https://huggingface.co/kmhf/moshika-pytorch-bf16). -- Moshiko for PyTorch (bf16): [kmhf/moshiko-pytorch-bf16](https://huggingface.co/kmhf/moshiko-pytorch-bf16). -- Moshika for MLX (int4, int8, bf16): [kmhf/moshiko-mlx-q4](https://huggingface.co/kmhf/moshika-mlx-q4), [kmhf/moshiko-mlx-q8](https://huggingface.co/kmhf/moshika-mlx-q8), [kmhf/moshiko-mlx-bf16](https://huggingface.co/kmhf/moshika-mlx-bf16). -- Moshiko for MLX (int4, int8, bf16): [kmhf/moshiko-mlx-q4](https://huggingface.co/kmhf/moshiko-mlx-q4), [kmhf/moshiko-mlx-q8](https://huggingface.co/kmhf/moshiko-mlx-q8), [kmhf/moshiko-mlx-bf16](https://huggingface.co/kmhf/moshiko-mlx-bf16). -- Moshika for Rust/Candle (int8, bf16): [kmhf/moshika-candle-q8](https://huggingface.co/kmhf/moshika-candle-q8), [kmhf/moshiko-mlx-bf16](https://huggingface.co/kmhf/moshika-candle-bf16). -- Moshiko for Rust/Candle (int8, bf16): [kmhf/moshiko-candle-q8](https://huggingface.co/kmhf/moshiko-candle-q8), [kmhf/moshiko-mlx-bf16](https://huggingface.co/kmhf/moshiko-candle-bf16). +- Moshika for PyTorch (bf16): [kyutai/moshika-pytorch-bf16](https://huggingface.co/kyutai/moshika-pytorch-bf16). +- Moshiko for PyTorch (bf16): [kyutai/moshiko-pytorch-bf16](https://huggingface.co/kyutai/moshiko-pytorch-bf16). +- Moshika for MLX (int4, int8, bf16): [kyutai/moshiko-mlx-q4](https://huggingface.co/kyutai/moshika-mlx-q4), [kyutai/moshiko-mlx-q8](https://huggingface.co/kyutai/moshika-mlx-q8), [kyutai/moshiko-mlx-bf16](https://huggingface.co/kyutai/moshika-mlx-bf16). +- Moshiko for MLX (int4, int8, bf16): [kyutai/moshiko-mlx-q4](https://huggingface.co/kyutai/moshiko-mlx-q4), [kyutai/moshiko-mlx-q8](https://huggingface.co/kyutai/moshiko-mlx-q8), [kyutai/moshiko-mlx-bf16](https://huggingface.co/kyutai/moshiko-mlx-bf16). +- Moshika for Rust/Candle (int8, bf16): [kyutai/moshika-candle-q8](https://huggingface.co/kyutai/moshika-candle-q8), [kyutai/moshiko-mlx-bf16](https://huggingface.co/kyutai/moshika-candle-bf16). +- Moshiko for Rust/Candle (int8, bf16): [kyutai/moshiko-candle-q8](https://huggingface.co/kyutai/moshiko-candle-q8), [kyutai/moshiko-mlx-bf16](https://huggingface.co/kyutai/moshiko-candle-bf16). All models are released under the CC-BY 4.0 license. @@ -124,7 +124,7 @@ run the model, you can then use either the web UI or a command line client. Start the server with: ```bash -python -m moshi.server [--gradio-tunnel] [--hf-repo kmhf/moshika-pytorch-bf16] +python -m moshi.server [--gradio-tunnel] [--hf-repo kyutai/moshika-pytorch-bf16] ``` And then access the web UI on [localhost:8998](http://localhost:8998). If your GPU is on a distant machine @@ -156,8 +156,8 @@ Once you have installed `moshi_mlx`, you can run python -m moshi_mlx.local -q 4 # weights quantized to 4 bits python -m moshi_mlx.local -q 8 # weights quantized to 8 bits # And using a different pretrained model: -python -m moshi_mlx.local -q 4 --hf-repo kmhf/moshika-mlx-q4 -python -m moshi_mlx.local -q 8 --hf-repo kmhf/moshika-mlx-q8 +python -m moshi_mlx.local -q 4 --hf-repo kyutai/moshika-mlx-q4 +python -m moshi_mlx.local -q 8 --hf-repo kyutai/moshika-mlx-q8 # be careful to always match the `-q` and `--hf-repo` flag. ``` diff --git a/moshi/moshi/models/loaders.py b/moshi/moshi/models/loaders.py index cb7e2f5..ffb3512 100644 --- a/moshi/moshi/models/loaders.py +++ b/moshi/moshi/models/loaders.py @@ -18,7 +18,7 @@ TEXT_TOKENIZER_NAME = 'tokenizer_spm_32k_3.model' MOSHI_NAME = 'model.safetensors' MIMI_NAME = 'tokenizer-e351c8d8-checkpoint125.safetensors' -DEFAULT_REPO = 'kmhf/moshiko-pytorch-bf16' +DEFAULT_REPO = 'kyutai/moshiko-pytorch-bf16' _seanet_kwargs = { diff --git a/moshi/moshi/server.py b/moshi/moshi/server.py index 1ede14a..212ba9c 100644 --- a/moshi/moshi/server.py +++ b/moshi/moshi/server.py @@ -225,7 +225,7 @@ def main(): static_path: None | str = None if args.static is None: log("info", "retrieving the static content") - dist_tgz = hf_hub_download("kmhf/moshi-artifacts", "dist.tgz") + dist_tgz = hf_hub_download("kyutai/moshi-artifacts", "dist.tgz") dist_tgz = Path(dist_tgz) dist = dist_tgz.parent / "dist" if not dist.exists(): diff --git a/moshi_mlx/README.md b/moshi_mlx/README.md index 7269371..832ac0d 100644 --- a/moshi_mlx/README.md +++ b/moshi_mlx/README.md @@ -31,8 +31,8 @@ Once you have installed `moshi_mlx`, you can run python -m moshi_mlx.local -q 4 # weights quantized to 4 bits python -m moshi_mlx.local -q 8 # weights quantized to 8 bits # And using a different pretrained model: -python -m moshi_mlx.local -q 4 --hf-repo kmhf/moshika-mlx-q4 -python -m moshi_mlx.local -q 8 --hf-repo kmhf/moshika-mlx-q8 +python -m moshi_mlx.local -q 4 --hf-repo kyutai/moshika-mlx-q4 +python -m moshi_mlx.local -q 8 --hf-repo kyutai/moshika-mlx-q8 # be careful to always match the `-q` and `--hf-repo` flag. ``` diff --git a/moshi_mlx/moshi_mlx/local.py b/moshi_mlx/moshi_mlx/local.py index 883d4e6..812442f 100644 --- a/moshi_mlx/moshi_mlx/local.py +++ b/moshi_mlx/moshi_mlx/local.py @@ -263,11 +263,11 @@ def main(printer: AnyPrinter): if args.hf_repo is None: if args.quantized == 8: - args.hf_repo = 'kmhf/moshiko-mlx-q8' + args.hf_repo = 'kyutai/moshiko-mlx-q8' elif args.quantized == 4: - args.hf_repo = 'kmhf/moshiko-mlx-q4' + args.hf_repo = 'kyutai/moshiko-mlx-q4' elif args.quantized is None: - args.hf_repo = 'kmhf/moshiko-mlx-bf16' + args.hf_repo = 'kyutai/moshiko-mlx-bf16' else: print(f"Invalid value for quantized {args.quantized}") sys.exit(1) diff --git a/moshi_mlx/moshi_mlx/local_web.py b/moshi_mlx/moshi_mlx/local_web.py index 578a24a..8eb0070 100644 --- a/moshi_mlx/moshi_mlx/local_web.py +++ b/moshi_mlx/moshi_mlx/local_web.py @@ -314,7 +314,7 @@ async def go(): static_path: None | str = None if args.static is None: log("info", "retrieving the static content") - dist_tgz = hf_hub_download('kmhf/moshi-artifacts', "dist.tgz") + dist_tgz = hf_hub_download('kyutai/moshi-artifacts', "dist.tgz") dist_tgz = Path(dist_tgz) dist = dist_tgz.parent / "dist" if not dist.exists(): @@ -368,11 +368,11 @@ def main(): args = parser.parse_args() if args.hf_repo is None: if args.quantized == 8: - args.hf_repo = 'kmhf/moshiko-mlx-q8' + args.hf_repo = 'kyutai/moshiko-mlx-q8' elif args.quantized == 4: - args.hf_repo = 'kmhf/moshiko-mlx-q4' + args.hf_repo = 'kyutai/moshiko-mlx-q4' elif args.quantized is None: - args.hf_repo = 'kmhf/moshiko-mlx-bf16' + args.hf_repo = 'kyutai/moshiko-mlx-bf16' else: print(f"Invalid value for quantized {args.quantized}") sys.exit(1) diff --git a/rust/moshi-backend/config-q8.json b/rust/moshi-backend/config-q8.json index 5442c65..b5e74af 100644 --- a/rust/moshi-backend/config-q8.json +++ b/rust/moshi-backend/config-q8.json @@ -1,6 +1,6 @@ { "instance_name": "foo", - "hf_repo": "kmhf/moshiko-candle-q8", + "hf_repo": "kyutai/moshiko-candle-q8", "lm_model_file": "$HOME/tmp/model.q8.gguf", "text_tokenizer_file": "$HOME/tmp/tokenizer_spm_32k_3.model", "log_dir": "$HOME/tmp/moshi-logs", diff --git a/rust/moshi-backend/config.json b/rust/moshi-backend/config.json index cc03a45..eb7f82c 100644 --- a/rust/moshi-backend/config.json +++ b/rust/moshi-backend/config.json @@ -1,6 +1,6 @@ { "instance_name": "foo", - "hf_repo": "kmhf/moshiko-candle-bf16", + "hf_repo": "kyutai/moshiko-candle-bf16", "lm_model_file": "$HOME/tmp/model.safetensors", "text_tokenizer_file": "$HOME/tmp/tokenizer_spm_32k_3.model", "log_dir": "$HOME/tmp/moshi-logs", diff --git a/rust/moshi-backend/src/main.rs b/rust/moshi-backend/src/main.rs index 23bcc97..6f88890 100644 --- a/rust/moshi-backend/src/main.rs +++ b/rust/moshi-backend/src/main.rs @@ -126,7 +126,7 @@ async fn main() -> Result<()> { if !std::path::PathBuf::from(&config.static_dir).exists() { use hf_hub::api::tokio::Api; let api = Api::new()?; - let repo = api.model("kmhf/moshi-artifacts".to_string()); + let repo = api.model("kyutai/moshi-artifacts".to_string()); let dist_tgz = repo.get("dist.tgz").await?; if let Some(parent) = dist_tgz.parent() { let dist = parent.join("dist");