ONNX export and model card

README.md

@@ -0,0 +1,70 @@
+---
+license: cc-by-nc-4.0
+base_model: MahmoudAshraf/mms-300m-1130-forced-aligner
+library_name: onnx
+pipeline_tag: automatic-speech-recognition
+tags:
+  - forced-alignment
+  - singing
+  - lyrics-alignment
+  - ultrastar
+  - onnx
+---
+
+# Tune Studio aligner
+
+A CTC forced aligner for **sung** lyrics, plus a small head that finds where syllables start. Made for
+Tune Studio, which times lyrics to a vocal stem to make UltraStar karaoke charts.
+
+- `aligner.onnx`: [MahmoudAshraf/mms-300m-1130-forced-aligner](https://huggingface.co/MahmoudAshraf/mms-300m-1130-forced-aligner)
+  (Meta's MMS-300m wav2vec2, a speech aligner) fine-tuned on singing.
+- `boundary.onnx`: a syllable-start head that reads the aligner's last hidden layer.
+- `vocab.json`, `tokenizer_config.json`, `config.json`: the aligner's vocabulary and configuration, unchanged from the base model.
+
+## Use
+
+**aligner.onnx**: input `audio` `[batch, samples]` float32, 16 kHz mono, raw (not normalized). Outputs `logits`
+`[batch, frames, 32]` (CTC, before log-softmax) and `hidden` `[batch, frames, 1024]`, one frame per 20 ms (320
+samples). Run a song in 30 s windows with 2 s of context on each side, the way
+[ctc-forced-aligner](https://github.com/MahmoudAshraf97/ctc-forced-aligner) does (`generate_emissions`), and keep
+each window's middle 30 s. Text is romanized and tokenized as in ctc-forced-aligner.
+
+**boundary.onnx**: input `hidden` `[1, frames, 1024]`, the aligner's hidden states for the **whole song** (the
+windows joined). Output `probs` `[1, frames]`: the probability that a syllable starts in each 20 ms frame. Tune
+Studio moves each aligned syllable start to a peak (≥ 0.3) within ±0.3 s, assigning each peak to at most one
+syllable in order.
+
+## Training
+
+- Aligner: 4000 steps at learning rate 6e-5, feature encoder frozen, on about 39,000 sung lines from the
+  hand-made UltraStar charts of 669 songs (the train split), with CTC targets from the lyrics. Held-out CTC loss
+  2.81 → 0.77.
+- Boundary head: Linear → bidirectional GRU (192) → Linear, trained on the same songs to predict the charts'
+  syllable starts (Gaussian targets, σ = 20 ms). On held-out songs, 79.6% of chart syllable starts have a peak
+  within ±40 ms and 94.0% within ±80 ms.
+
+## Results
+
+Measured once on 74 held-out songs, English and German, against their hand-made charts: Tune Studio's chart
+pipeline today (this aligner, the boundary head, the note refiners and some rule changes) vs. its first version
+(the stock aligner, no learned steps):
+
+| | first version | today | songs better / worse |
+|---|---|---|---|
+| syllable F1 | 0.788 | 0.875 | 66 / 6 |
+| note F1 | 0.609 | 0.698 | 67 / 5 |
+| frame F1 | 0.526 | 0.589 | 60 / 10 |
+| exact pitch | 0.792 | 0.851 | 69 / 4 |
+
+On its own (same pipeline otherwise, 58 development songs), the fine-tuned aligner raised syllable F1 by 0.047 and
+note F1 by 0.029, and the boundary head on top of it syllable F1 by another 0.026.
+
+Both versions split words into syllables with tables built partly from the test charts, which flatters both,
+German most; the difference between them is not affected.
+
+The ONNX files were checked against the PyTorch models: identical CTC alignment paths and syllable-start peaks on
+full songs.
+
+## License
+
+CC BY-NC 4.0, inherited from the MMS base model: free for non-commercial use, with attribution.

aligner.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:597a1d775df8f7fe3c94310568751ef9253645b4d8cdf21bb7a3da1ab80225a1
+size 1262284621

boundary.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c408211885ef43a867cfb107e52eaa7399a3df84985764c578a72c8949252736
+size 2570198

config.json

@@ -0,0 +1,108 @@
+{
+  "activation_dropout": 0.0,
+  "adapter_attn_dim": null,
+  "adapter_kernel_size": 3,
+  "adapter_stride": 2,
+  "add_adapter": false,
+  "apply_spec_augment": true,
+  "architectures": [
+    "Wav2Vec2ForCTC"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 1,
+  "classifier_proj_size": 256,
+  "codevector_dim": 768,
+  "contrastive_logits_temperature": 0.1,
+  "conv_bias": true,
+  "conv_dim": [
+    512,
+    512,
+    512,
+    512,
+    512,
+    512,
+    512
+  ],
+  "conv_kernel": [
+    10,
+    3,
+    3,
+    3,
+    3,
+    2,
+    2
+  ],
+  "conv_stride": [
+    5,
+    2,
+    2,
+    2,
+    2,
+    2,
+    2
+  ],
+  "ctc_loss_reduction": "mean",
+  "ctc_zero_infinity": true,
+  "diversity_loss_weight": 0.1,
+  "do_stable_layer_norm": true,
+  "dtype": "float32",
+  "eos_token_id": 2,
+  "feat_extract_activation": "gelu",
+  "feat_extract_dropout": 0.0,
+  "feat_extract_norm": "layer",
+  "feat_proj_dropout": 0.0,
+  "feat_quantizer_dropout": 0.0,
+  "final_dropout": 0.0,
+  "gradient_checkpointing": false,
+  "hidden_act": "gelu",
+  "hidden_dropout": 0.0,
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "layer_norm_eps": 1e-05,
+  "layerdrop": 0.0,
+  "mask_feature_length": 10,
+  "mask_feature_min_masks": 0,
+  "mask_feature_prob": 0.0,
+  "mask_time_length": 10,
+  "mask_time_min_masks": 2,
+  "mask_time_prob": 0.0,
+  "model_type": "wav2vec2",
+  "num_adapter_layers": 3,
+  "num_attention_heads": 16,
+  "num_codevector_groups": 2,
+  "num_codevectors_per_group": 320,
+  "num_conv_pos_embedding_groups": 16,
+  "num_conv_pos_embeddings": 128,
+  "num_feat_extract_layers": 7,
+  "num_hidden_layers": 24,
+  "num_negatives": 100,
+  "output_hidden_size": 1024,
+  "pad_token_id": 0,
+  "proj_codevector_dim": 768,
+  "tdnn_dilation": [
+    1,
+    2,
+    3,
+    1,
+    1
+  ],
+  "tdnn_dim": [
+    512,
+    512,
+    512,
+    512,
+    1500
+  ],
+  "tdnn_kernel": [
+    5,
+    3,
+    3,
+    1,
+    1
+  ],
+  "transformers_version": "5.19.0",
+  "use_weighted_layer_sum": false,
+  "vocab_size": 31,
+  "xvector_output_dim": 512
+}

tokenizer_config.json

@@ -0,0 +1,50 @@
+{
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<blank>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": false
+    },
+    "1": {
+      "content": "<pad>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": false
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": false
+    },
+    "3": {
+      "content": "<unk>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "backend": "custom",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": true,
+  "do_lower_case": false,
+  "eos_token": "</s>",
+  "is_local": false,
+  "local_files_only": false,
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "<pad>",
+  "replace_word_delimiter_char": " ",
+  "target_lang": null,
+  "tokenizer_class": "Wav2Vec2CTCTokenizer",
+  "unk_token": "<unk>",
+  "word_delimiter_token": "None"
+}

vocab.json

@@ -0,0 +1,33 @@
+{
+  "'": 28,
+  "</s>": 2,
+  "<blank>": 0,
+  "<pad>": 1,
+  "<unk>": 3,
+  "a": 4,
+  "b": 20,
+  "c": 23,
+  "d": 16,
+  "e": 6,
+  "f": 27,
+  "g": 17,
+  "h": 18,
+  "i": 5,
+  "j": 25,
+  "k": 14,
+  "l": 15,
+  "m": 13,
+  "n": 7,
+  "o": 8,
+  "p": 21,
+  "q": 29,
+  "r": 12,
+  "s": 11,
+  "t": 10,
+  "u": 9,
+  "v": 24,
+  "w": 22,
+  "x": 30,
+  "y": 19,
+  "z": 26
+}