Duplicate from hi-paris/CosyVoice2-0.5B-EU

.gitattributes

@@ -0,0 +1,36 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+sample_audio_prompt.wav filter=lfs diff=lfs merge=lfs -text

CosyVoice-BlankEN/config.json

@@ -0,0 +1,27 @@
+{
+  "architectures": [
+    "Qwen2ForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "eos_token_id": 151645,
+  "hidden_act": "silu",
+  "hidden_size": 896,
+  "initializer_range": 0.02,
+  "intermediate_size": 4864,
+  "max_position_embeddings": 32768,
+  "max_window_layers": 24,
+  "model_type": "qwen2",
+  "num_attention_heads": 14,
+  "num_hidden_layers": 24,
+  "num_key_value_heads": 2,
+  "rms_norm_eps": 1e-06,
+  "rope_theta": 1000000.0,
+  "sliding_window": 32768,
+  "tie_word_embeddings": true,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.40.1",
+  "use_cache": true,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}

CosyVoice-BlankEN/generation_config.json

@@ -0,0 +1,14 @@
+{
+  "bos_token_id": 151643,
+  "pad_token_id": 151643,
+  "do_sample": true,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "repetition_penalty": 1.1,
+  "temperature": 0.7,
+  "top_p": 0.8,
+  "top_k": 20,
+  "transformers_version": "4.37.0"
+}

CosyVoice-BlankEN/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:130282af0dfa9fe5840737cc49a0d339d06075f83c5a315c3372c9a0740d0b96
+size 988097824

CosyVoice-BlankEN/tokenizer_config.json

@@ -0,0 +1,40 @@
+{
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": ["<|im_start|>", "<|im_end|>"],
+  "bos_token": null,
+  "chat_template": "{% for message in messages %}{% if loop.first and messages[0]['role'] != 'system' %}{{ '<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n' }}{% endif %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\n' }}{% endif %}",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "model_max_length": 32768,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

README.md

@@ -0,0 +1,100 @@
+---
+license: apache-2.0
+language:
+- en
+- de
+- fr
+- zh
+- ko
+- ja
+base_model:
+- FunAudioLLM/CosyVoice2-0.5B
+- Qwen/Qwen3-0.6B
+- utter-project/EuroLLM-1.7B-Instruct
+- mistralai/Mistral-7B-v0.3
+pipeline_tag: text-to-speech
+---
+
+<p align="center">
+  <img src="https://hi-paris.github.io/CosyVoice2-EU/cosyvoice2-logo-clear.png"
+       alt="CosyVoice2-EU logo" width="260">
+</p>
+
+
+# CosyVoice2-0.5B-EU — FR/DE Zero-Shot Voice Cloning (CosyVoice2)
+
+**Europeanized CosyVoice2 for French & German.**  
+Plug-and-play zero-shot voice cloning with streaming support, bilingual training (FR+DE), and a simple CLI via the companion PyPI package.
+
+**👉 PyPI:** `cosyvoice2-eu` (current: **0.2.7**) at https://pypi.org/project/cosyvoice2-eu/  
+**👉 Demo:** https://hi-paris.github.io/CosyVoice2-EU/ 
+**👉 Built on:** FunAudioLLM **CosyVoice2** (semantic LM + chunk-aware flow + HiFi-GAN)
+
+---
+
+## TL;DR
+High-quality **French/German** zero-shot TTS (text + short reference audio) built on **CosyVoice2**. Optimized for sentence-to-paragraph narration, bilingual FR+DE adaptation, and easy local inference.
+While this model is optimized for French and German, it remains fully compatible with the original CosyVoice2 languages — English, Chinese, Japanese, Korean, and their dialects.
+
+---
+
+## Quickstart (CLI)
+
+Install:
+```bash
+pip install cosyvoice2-eu
+```
+
+French example:
+```bash
+cosy2-eu   --text "Salut ! Je vous présente CosyVoice 2, un système de synthèse vocale très avancé."   --prompt path/to/french_ref.wav   --out out_fr.wav
+```
+
+German example:
+```bash
+cosy2-eu   --text "Hallo! Ich präsentiere CosyVoice 2 – ein fortschrittliches TTS-System."   --prompt path/to/german_ref.wav   --out out_de.wav
+```
+
+> First run downloads the model from this repo and caches it locally.  
+> Tip: You can experiment with prompts for style control using `"<style>. <|endofprompt|> <text>"`, e.g., "Speak cheerfully. <|endofprompt|>  Hallo! Wie geht es Ihnen heute?"
+
+---
+
+## What you get
+- **Zero-shot voice cloning** for **FR/DE** (reference audio → cloned timbre & style).
+- **Bilingual adaptation** (FR+DE) on top of CosyVoice2 for stronger data efficiency. While this model adds support for French and German, it remains fully compatible with the original CosyVoice2 languages — English, Chinese, Japanese, Korean, and their dialects.
+- **Streaming & non-streaming** synthesis supported by the underlying architecture.
+- **Simple local inference**: one pip install, one CLI (`cosy2-eu`).
+- **Interoperable components** (text→semantic LM, flow decoder, HiFi-GAN vocoder).
+
+Also compatible with original CosyVoice2 languages (EN/ZH/JA/KO & dialects).
+
+---
+
+## Inputs / Outputs
+- **Input:** text (FR/DE) + short **reference audio** (mono WAV recommended).  
+- **Output:** synthesized WAV cloning the reference speaker’s timbre, speaking the input text in FR/DE.
+
+---
+
+## Notes & limitations
+- FR/DE were adapted under constrained open-data budgets; extreme edge cases (very noisy prompts, long numerics, heavy code-switching) may require careful prompting or additional fine-tuning.
+- Voice cloning carries **misuse risks** (impersonation, fraud). Use only with consent and follow local laws/policies.
+
+---
+
+## License & attribution
+- **License:** Apache-2.0 (see card metadata / repo).  
+- Built on **CosyVoice2** by FunAudioLLM; please cite their work (see below).
+
+
+---
+
+**Links**
+- PyPI (inference CLI): https://pypi.org/project/cosyvoice2-eu/  
+- Upstream project: https://github.com/FunAudioLLM/CosyVoice  
+- CosyVoice2 paper & page: https://arxiv.org/abs/2412.10117 • https://funaudiollm.github.io/cosyvoice2/
+
+---
+
+*If you use CosyVoice2-0.5B-EU in research or products, please add a short acknowledgment and share feedback or samples—we’re continuously improving FR/DE expressiveness and robustness.*

campplus.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a6ac6a63997761ae2997373e2ee1c47040854b4b759ea41ec48e4e42df0f4d73
+size 28303423

cosyvoice2.yaml

@@ -0,0 +1,237 @@
+# set random seed, so that you may reproduce your result.
+__set_seed1: !apply:random.seed [1986]
+__set_seed2: !apply:numpy.random.seed [1986]
+__set_seed3: !apply:torch.manual_seed [1986]
+__set_seed4: !apply:torch.cuda.manual_seed_all [1986]
+
+# fixed params
+sample_rate: 24000
+llm_input_size: 896
+llm_output_size: 896
+spk_embed_dim: 192
+qwen_pretrain_path: ''
+token_frame_rate: 25
+token_mel_ratio: 2
+
+# stream related params
+chunk_size: 25 # streaming inference chunk size, in token
+num_decoding_left_chunks: -1 # streaming inference flow decoder left chunk size, <0 means use all left chunks
+
+# model params
+# for all class/function included in this repo, we use !<name> or !<new> for intialization, so that user may find all corresponding class/function according to one single yaml.
+# for system/third_party class/function, we do not require this.
+llm: !new:cosyvoice.llm.llm.Qwen2LM
+    # llm_input_size/llm_output_size will be auto-set from backbone.hidden_size, so 896 here is ignored
+    llm_input_size: 0
+    llm_output_size: 0
+    speech_token_size: 6561
+    length_normalized_loss: True
+    lsm_weight: 0
+    mix_ratio: [5, 15]
+    llm: !new:cosyvoice.llm.llm.HFBackbone     # backbone-agnostic
+        pretrain_path: !ref <qwen_pretrain_path> # e.g., "Qwen/Qwen3-0.6B" or "mistralai/Mistral-7B-Instruct-v0.3"
+    sampling: !name:cosyvoice.utils.common.ras_sampling
+        top_p: 0.8
+        top_k: 25
+        win_size: 10
+        tau_r: 0.1
+
+flow: !new:cosyvoice.flow.flow.CausalMaskedDiffWithXvec
+    input_size: 512
+    output_size: 80
+    spk_embed_dim: !ref <spk_embed_dim>
+    output_type: 'mel'
+    vocab_size: 6561
+    input_frame_rate: !ref <token_frame_rate>
+    only_mask_loss: True
+    token_mel_ratio: !ref <token_mel_ratio>
+    pre_lookahead_len: 3
+    encoder: !new:cosyvoice.transformer.upsample_encoder.UpsampleConformerEncoder
+        output_size: 512
+        attention_heads: 8
+        linear_units: 2048
+        num_blocks: 6
+        dropout_rate: 0.1
+        positional_dropout_rate: 0.1
+        attention_dropout_rate: 0.1
+        normalize_before: True
+        input_layer: 'linear'
+        pos_enc_layer_type: 'rel_pos_espnet'
+        selfattention_layer_type: 'rel_selfattn'
+        input_size: 512
+        use_cnn_module: False
+        macaron_style: False
+        static_chunk_size: !ref <chunk_size>
+    decoder: !new:cosyvoice.flow.flow_matching.CausalConditionalCFM
+        in_channels: 240
+        n_spks: 1
+        spk_emb_dim: 80
+        cfm_params: !new:omegaconf.DictConfig
+            content:
+                sigma_min: 1e-06
+                solver: 'euler'
+                t_scheduler: 'cosine'
+                training_cfg_rate: 0.2
+                inference_cfg_rate: 0.7
+                reg_loss_type: 'l1'
+        estimator: !new:cosyvoice.flow.decoder.CausalConditionalDecoder
+            in_channels: 320
+            out_channels: 80
+            channels: [256]
+            dropout: 0.0
+            attention_head_dim: 64
+            n_blocks: 4
+            num_mid_blocks: 12
+            num_heads: 8
+            act_fn: 'gelu'
+            static_chunk_size: !ref <chunk_size> * <token_mel_ratio>
+            num_decoding_left_chunks: !ref <num_decoding_left_chunks>
+
+hift: !new:cosyvoice.hifigan.generator.HiFTGenerator
+    in_channels: 80
+    base_channels: 512
+    nb_harmonics: 8
+    sampling_rate: !ref <sample_rate>
+    nsf_alpha: 0.1
+    nsf_sigma: 0.003
+    nsf_voiced_threshold: 10
+    upsample_rates: [8, 5, 3]
+    upsample_kernel_sizes: [16, 11, 7]
+    istft_params:
+        n_fft: 16
+        hop_len: 4
+    resblock_kernel_sizes: [3, 7, 11]
+    resblock_dilation_sizes: [[1, 3, 5], [1, 3, 5], [1, 3, 5]]
+    source_resblock_kernel_sizes: [7, 7, 11]
+    source_resblock_dilation_sizes: [[1, 3, 5], [1, 3, 5], [1, 3, 5]]
+    lrelu_slope: 0.1
+    audio_limit: 0.99
+    f0_predictor: !new:cosyvoice.hifigan.f0_predictor.ConvRNNF0Predictor
+        num_class: 1
+        in_channels: 80
+        cond_channels: 512
+
+# gan related module
+mel_spec_transform1: !name:matcha.utils.audio.mel_spectrogram
+    n_fft: 1920
+    num_mels: 80
+    sampling_rate: !ref <sample_rate>
+    hop_size: 480
+    win_size: 1920
+    fmin: 0
+    fmax: null
+    center: False
+hifigan: !new:cosyvoice.hifigan.hifigan.HiFiGan
+    generator: !ref <hift>
+    discriminator: !new:cosyvoice.hifigan.discriminator.MultipleDiscriminator
+        mpd: !new:matcha.hifigan.models.MultiPeriodDiscriminator
+        mrd: !new:cosyvoice.hifigan.discriminator.MultiResSpecDiscriminator
+    mel_spec_transform: [
+        !ref <mel_spec_transform1>
+    ]
+
+# processor functions
+parquet_opener: !name:cosyvoice.dataset.processor.parquet_opener
+get_tokenizer: !name:cosyvoice.tokenizer.tokenizer.get_qwen_tokenizer
+    token_path: !ref <qwen_pretrain_path>
+    skip_special_tokens: True
+    # add_additional_specials: auto-detected based on token_path (True for blanken/CosyVoice models, False for custom HF backbones)
+allowed_special: 'all'
+tokenize: !name:cosyvoice.dataset.processor.tokenize
+    get_tokenizer: !ref <get_tokenizer>
+    allowed_special: !ref <allowed_special>
+filter: !name:cosyvoice.dataset.processor.filter
+    max_length: 40960
+    min_length: 100
+    token_max_length: 512 # not sure if this can just be changed?
+    token_min_length: 1
+resample: !name:cosyvoice.dataset.processor.resample
+    resample_rate: !ref <sample_rate>
+truncate: !name:cosyvoice.dataset.processor.truncate
+    truncate_length: 24480 # must be a multiplier of hop_size
+feat_extractor: !name:matcha.utils.audio.mel_spectrogram
+    n_fft: 1920
+    num_mels: 80
+    sampling_rate: !ref <sample_rate>
+    hop_size: 480
+    win_size: 1920
+    fmin: 0
+    fmax: 8000
+    center: False
+compute_fbank: !name:cosyvoice.dataset.processor.compute_fbank
+    feat_extractor: !ref <feat_extractor>
+    token_mel_ratio: 2
+compute_f0: !name:cosyvoice.dataset.processor.compute_f0
+    sample_rate: !ref <sample_rate>
+    hop_size: 480
+parse_embedding: !name:cosyvoice.dataset.processor.parse_embedding
+    normalize: True
+shuffle: !name:cosyvoice.dataset.processor.shuffle
+    shuffle_size: 1000
+sort: !name:cosyvoice.dataset.processor.sort
+    sort_size: 500  # sort_size should be less than shuffle_size
+batch: !name:cosyvoice.dataset.processor.batch
+    batch_type: 'dynamic'
+    max_frames_in_batch: 3000
+padding: !name:cosyvoice.dataset.processor.padding
+    use_spk_embedding: True # change to True during sft
+
+
+# dataset processor pipeline
+data_pipeline: [
+    !ref <parquet_opener>,
+    !ref <tokenize>,
+    !ref <filter>,
+    !ref <resample>,
+    !ref <compute_fbank>,
+    !ref <parse_embedding>,
+    !ref <shuffle>,
+    !ref <sort>,
+    !ref <batch>,
+    !ref <padding>,
+]
+data_pipeline_gan: [
+    !ref <parquet_opener>,
+    !ref <tokenize>,
+    !ref <filter>,
+    !ref <resample>,
+    !ref <truncate>,
+    !ref <compute_fbank>,
+    !ref <compute_f0>,
+    !ref <parse_embedding>,
+    !ref <shuffle>,
+    !ref <sort>,
+    !ref <batch>,
+    !ref <padding>,
+]
+
+# llm flow train conf
+train_conf:
+    optim: adamw
+    optim_conf:
+        lr: 1e-5 # change to 1e-5 during sft
+        # weight_decay: 0.01
+    scheduler: constantlr # change to constantlr during sft
+    scheduler_conf:
+        warmup_steps: 2500
+    max_epoch: 30 # 200
+    grad_clip: 5
+    accum_grad: 4
+    log_interval: 100
+    save_per_step: 3000 # -1 this is where you can set the step-wise validation checkpoint, -1 means no step-wise validation checkpoint
+
+# gan train conf
+train_conf_gan:
+    optim: adam
+    optim_conf:
+        lr: 0.0002 # use small lr for gan training
+    scheduler: constantlr
+    optim_d: adam
+    optim_conf_d:
+        lr: 0.0002 # use small lr for gan training
+    scheduler_d: constantlr
+    max_epoch: 20
+    grad_clip: 5
+    accum_grad: 1 # in gan training, accum_grad must be 1
+    log_interval: 100
+    save_per_step: -1

flow.decoder.estimator.fp32.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:628863056c6462e7563c3ce0c04f3434b8d0eabd56cc1ccc58edf58ceb0d758f
+size 286312346

flow.encoder.fp16.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:308e0517a8c35e36d66f25e7e467dd0f57fadb7c5f6c1a5f743e9e82db53b559
+size 116706755

flow.encoder.fp32.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2819dc8568189a4dd3792669b2e57ac1be183be653b2f5cf97ad11031c17b7ef
+size 192369091

flow.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c51909e408ab4b9ab57179f99ce1c97a7f8f4618fccd68f067a1d9de85783062
+size 450569991

hifigan.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:04936834f1bdcfeb0973203bb53b9f8ea652b779a06f3ff3958ee7693d7bef96
+size 248993278

hift.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3386cc880324d4e98e05987b99107f49e40ed925b8ecc87c1f4939432d429879
+size 83390254

llm.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eaa230b528e30f7684dfa2bbc984ccbc336299c3d475f9722224a60d476f2c3a
+size 2567851323

sample_audio_prompt.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b4e3ab32edbbc2561141ccd12636be2c1bc304a6cbd8696d69213246dd4cf1e
+size 5690588

speech_tokenizer_v2.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d43342aa12163a80bf07bffb94c9de2e120a8df2f9917cd2f642e7f4219c6f71
+size 496082973