Update README.md

README.md

@@ -1,199 +1,291 @@
 ---
+base_model:
+- Qwen/Qwen3-4B-Instruct-2507
 library_name: transformers
-tags: []
+pipeline_tag: text-generation
+tags:
+- audio
+- speech
+- audio-codec
+- neural-audio-codec
+- spoken-language-modeling
+- codec-superb
+- qwen3
+datasets:
+- librispeech_asr
+metrics:
+- perplexity
+- pesq
+- stoi
 ---
 
-# Model Card for Model ID
+# LLM-Codec
 
-<!-- Provide a quick summary of what the model is/does. -->
+LLM-Codec is a neural audio codec checkpoint trained to produce discrete audio
+tokens that are both reconstructable and easier for autoregressive language
+models to predict.
 
+Model: https://huggingface.co/voidful/llm-codec
 
+Code: https://github.com/voidful/llm-codec
 
-## Model Details
+Usage reference: https://github.com/voidful/Codec-SUPERB
 
-### Model Description
+## Model Description
 
-<!-- Provide a longer summary of what this model is. -->
+Most neural audio codecs are trained for waveform reconstruction. Spoken
+language models, however, consume codec tokens with a next-token prediction
+objective. This mismatch can make acoustically valid variation appear as token
+uncertainty to the language model.
 
-This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
+LLM-Codec adapts a codec with language-model-facing objectives while keeping the
+deployed codec interface unchanged. The model is trained with:
 
-- **Developed by:** [More Information Needed]
-- **Funded by [optional]:** [More Information Needed]
-- **Shared by [optional]:** [More Information Needed]
-- **Model type:** [More Information Needed]
-- **Language(s) (NLP):** [More Information Needed]
-- **License:** [More Information Needed]
-- **Finetuned from model [optional]:** [More Information Needed]
+- Future Token Prediction (FTP): Medusa-style heads predict future audio tokens
+  from frozen-LLM hidden states.
+- Semantic Alignment (SA): audio-induced hidden states are aligned with paired
+  text hidden states inside a frozen LLM.
+- Differentiable Gumbel bridge: hard Gumbel-Softmax keeps discrete forward
+  tokens while enabling gradients to flow to the codec encoder.
+- Reconstruction losses: mel, multi-scale mel, multi-resolution STFT, complex
+  STFT, VQ, GAN, and feature matching losses.
 
-### Model Sources [optional]
+The deployed codec does not require the auxiliary FTP heads.
 
-<!-- Provide the basic links for the model. -->
+## Intended Use
 
-- **Repository:** [More Information Needed]
-- **Paper [optional]:** [More Information Needed]
-- **Demo [optional]:** [More Information Needed]
+This model is intended for research and development in:
 
-## Uses
+- audio tokenization for spoken language modeling
+- codec reconstruction experiments
+- token-level speech LM training
+- Codec-SUPERB style codec evaluation
+- speech token analysis and ablation studies
 
-<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+It is not a full text-to-speech system by itself. For speech generation, use the
+codec as the tokenizer/decoder inside a separate speech language modeling
+pipeline.
 
-### Direct Use
+## Out-of-Scope Use
 
-<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+Do not use this model for:
 
-[More Information Needed]
+- impersonation or unauthorized voice cloning
+- surveillance or speaker tracking without consent
+- high-stakes speaker, language, or identity decisions
+- generating deceptive audio content
 
-### Downstream Use [optional]
+## Installation
 
-<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+The easiest inference path is through the Codec-SUPERB `SoundCodec` interface.
 
-[More Information Needed]
+```bash
+git clone https://github.com/voidful/Codec-SUPERB.git
+cd Codec-SUPERB
+pip install -r requirements.txt
+export PYTHONPATH=$PWD:$PYTHONPATH
+```
 
-### Out-of-Scope Use
+If your environment supports editable installs, this is also convenient:
 
-<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+```bash
+pip install -e .
+```
 
-[More Information Needed]
+## Quick Start
 
-## Bias, Risks, and Limitations
+Load LLM-Codec through the Codec-SUPERB codec registry:
 
-<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+```python
+from SoundCodec import codec
 
-[More Information Needed]
+print(codec.list_codec())
+model = codec.load_codec("llmcodec")
+```
 
-### Recommendations
+Encode and reconstruct one audio file:
 
-<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+```python
+from SoundCodec import codec
+import torchaudio
+import soundfile as sf
 
-Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+model = codec.load_codec("llmcodec")
 
-## How to Get Started with the Model
+waveform, sample_rate = torchaudio.load("sample_audio.wav")
+data_item = {
+    "audio": {
+        "array": waveform.numpy()[0],
+        "sampling_rate": sample_rate,
+    }
+}
 
-Use the code below to get started with the model.
+units = model.extract_unit(data_item).unit
+print("Unit shape:", units.shape)
 
-[More Information Needed]
+result = model.synth(data_item, local_save=False)
+reconstructed = result["audio"]["array"]
+reconstructed_sr = result["audio"].get("sampling_rate", sample_rate)
 
-## Training Details
+sf.write("reconstructed.wav", reconstructed, reconstructed_sr)
+```
 
-### Training Data
+## Batch Usage
 
-<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+Codec-SUPERB also provides batch APIs:
 
-[More Information Needed]
+```python
+from SoundCodec import codec
+import torchaudio
 
-### Training Procedure
+model = codec.load_codec("llmcodec")
 
-<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+audio_files = ["audio1.wav", "audio2.wav", "audio3.wav"]
+data_list = []
 
-#### Preprocessing [optional]
+for path in audio_files:
+    waveform, sample_rate = torchaudio.load(path)
+    data_list.append({
+        "id": path,
+        "audio": {
+            "array": waveform.numpy()[0],
+            "sampling_rate": sample_rate,
+        },
+    })
 
-[More Information Needed]
+batch_units = model.batch_extract_unit(data_list)
+batch_audio = model.batch_decode_unit(batch_units)
 
+results = model.batch_synth(data_list, local_save=False)
+for item in results:
+    print(item["unit"].shape, item["audio"]["array"].shape)
+```
 
-#### Training Hyperparameters
+For better throughput, group audio samples with similar lengths before batching.
 
-- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+## Codec-SUPERB Evaluation
 
-#### Speeds, Sizes, Times [optional]
+To evaluate LLM-Codec with Codec-SUPERB-tiny:
 
-<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+```bash
+PYTHONPATH=. python3 scripts/dataset_creator.py \
+  --dataset voidful/codec-superb-tiny
 
-[More Information Needed]
+PYTHONPATH=. python3 scripts/benchmarking.py \
+  --dataset datasets/voidful/codec-superb-tiny_synth \
+  --models llmcodec
+```
 
-## Evaluation
+## Model Files
 
-<!-- This section describes the evaluation protocols and provides the results. -->
+The model repository provides:
 
-### Testing Data, Factors & Metrics
+- codec weights as `llm-codec.pt`
+- a tokenizer extended with `<CODEC_*>` audio tokens
+- Qwen-compatible model artifacts containing trained audio-token embeddings
 
-#### Testing Data
+The codec uses 20,480 audio tokens with the canonical token format:
 
-<!-- This should link to a Dataset Card if possible. -->
+```text
+<CODEC_0>, <CODEC_1>, ..., <CODEC_20479>
+```
 
-[More Information Needed]
+## Training Data
 
-#### Factors
+The codec was trained on LibriSpeech `train-clean-100` with paired transcripts.
+The validation split used during training is LibriSpeech `validation`.
 
-<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+Because training is speech-centric and transcript-supervised, performance may be
+weaker on non-English speech, conversational speech, music, environmental audio,
+or audio with strong noise and overlap.
 
-[More Information Needed]
+## Training Procedure
 
-#### Metrics
+Base components:
 
-<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+- Base codec: AUV
+- Frozen LLM backbone: Qwen3-4B-Instruct
+- Token rate: 50 Hz
+- Audio vocabulary size: 20,480
+- Segment length: 4 seconds
 
-[More Information Needed]
+Losses:
 
-### Results
+- reconstruction mel loss
+- multi-scale mel loss
+- multi-resolution STFT loss
+- complex STFT loss with phase term
+- VQ commitment loss
+- Gumbel bridge cross entropy
+- Future Token Prediction loss
+- Semantic Alignment cosine loss
+- Semantic Alignment contrastive loss with memory bank
+- MPD/MSD GAN and feature matching losses
 
-[More Information Needed]
+## Evaluation Results
 
-#### Summary
+### Token Learnability
 
+SALMon speech coherence accuracy after token-level LM training:
 
+| Tokenizer | Overall accuracy |
+| --- | ---: |
+| WavTok-L | 48.3 |
+| BigCodec | 49.4 |
+| UniCodec | 50.1 |
+| AUV | 49.4 |
+| LLM-Codec | 61.6 |
 
-## Model Examination [optional]
+Token-level perplexity on LibriSpeech after 3 epochs of LM training:
 
-<!-- Relevant interpretability work for the model goes here -->
+| Tokenizer | Eval loss | Perplexity |
+| --- | ---: | ---: |
+| WavTok-L | 11.91 | 148,122 |
+| UniCodec | 11.92 | 150,197 |
+| BigCodec | 11.96 | 156,448 |
+| AUV | 11.98 | 159,768 |
+| LLM-Codec | 8.44 | 4,617 |
+
+### Reconstruction Quality
 
-[More Information Needed]
+Codec-SUPERB-tiny speech reconstruction:
 
-## Environmental Impact
+| Model | Mel lower is better | STFT lower is better | PESQ higher is better | STOI higher is better |
+| --- | ---: | ---: | ---: | ---: |
+| AUV base | 0.762 | 1.648 | 2.094 | 0.850 |
+| LLM-Codec | 0.724 | 1.599 | 2.102 | 0.859 |
 
-<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+## Limitations
 
-Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+- The semantic alignment objective depends on paired speech and text.
+- The model is primarily validated on read speech.
+- Downstream generation quality depends on the separate speech language model.
+- The model may preserve speaker identity information present in the input.
+- The Hugging Face `transformers` artifacts are not a standalone text chatbot;
+  they accompany the codec/tokenizer workflow.
 
-- **Hardware Type:** [More Information Needed]
-- **Hours used:** [More Information Needed]
-- **Cloud Provider:** [More Information Needed]
-- **Compute Region:** [More Information Needed]
-- **Carbon Emitted:** [More Information Needed]
+## Citation
+
+```bibtex
+@misc{chung2026llmcodec,
+  title = {LLM-Codec: Neural Audio Codec Meets Language Model Objectives},
+  author = {Chung, Ho-Lam and Chen, Yiming and Lee, Hung-yi},
+  year = {2026},
+  note = {Model and code available at https://github.com/voidful/llm-codec}
+}
+```
 
-## Technical Specifications [optional]
-
-### Model Architecture and Objective
-
-[More Information Needed]
-
-### Compute Infrastructure
-
-[More Information Needed]
-
-#### Hardware
-
-[More Information Needed]
-
-#### Software
-
-[More Information Needed]
-
-## Citation [optional]
-
-<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
-
-**BibTeX:**
-
-[More Information Needed]
-
-**APA:**
-
-[More Information Needed]
-
-## Glossary [optional]
-
-<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
-
-[More Information Needed]
-
-## More Information [optional]
-
-[More Information Needed]
-
-## Model Card Authors [optional]
-
-[More Information Needed]
-
-## Model Card Contact
-
-[More Information Needed]
+If you use the Codec-SUPERB interface or benchmark, please also cite
+Codec-SUPERB:
+
+```bibtex
+@inproceedings{wu-etal-2024-codec,
+  title = {Codec-SUPERB: An In-Depth Analysis of Sound Codec Models},
+  author = {Wu, Haibin and Chung, Ho-Lam and Lin, Yi-Cheng and Wu, Yuan-Kuei and Chen, Xuanjun and Pai, Yu-Chi and Wang, Hsiu-Hsuan and Chang, Kai-Wei and Liu, Alexander and Lee, Hung-yi},
+  booktitle = {Findings of the Association for Computational Linguistics: ACL 2024},
+  year = {2024},
+  url = {https://aclanthology.org/2024.findings-acl.616},
+  doi = {10.18653/v1/2024.findings-acl.616},
+  pages = {10330--10348}
+}
+```