Remove torchcodec dependency

README.md

@@ -5,6 +5,7 @@ We present LFM2-Audio-1.5B, [Liquid AI](https://www.liquid.ai/)'s first end-to-e
 LFM2-Audio supports two generation modes, interleaved and sequential, to maximize performance and quality across different tasks. Interleaved generation outputs text and audio tokens in a fixed interleaved pattern. This approach minimizes time to first audio output and number of tokens generated, making it ideal for naturally flowing real-time speech-to-speech interactions on resource constrained devices. Sequential generation mode, where the model decides when to switch modalities via special tokens, is suitable for non-conversational tasks, such as speech-to-text (ASR) or text-to-speech (TTS).
 
 ### Updates
+- [Finetuning](#finetuning) is now supported in both interleaved and sequential generation modes. Version 1.2.0 introduces data preparation tools and a lightweight trainer, enabling users to fine-tune models for a broad range of tasks, from ASR and TTS to function calling and end-to-end speech-to-speech chat.
 - [LFM2.5-Audio-1.5B](https://huggingface.co/LiquidAI/LFM2.5-Audio-1.5B) is released! This model is based on the stronger LFM2.5-1.2B base, and comes with a lightning fast LFM2 based audio detokenizer, stronger ASR, and better TTS voices. To use the new detokenizer, simply use `processor.decode`, see the examples below for more details. For the improved TTS voices, see the [TTS](#tts) section.
 
 ## Installation
@@ -15,6 +16,11 @@ pip install "liquid-audio [demo]" # optional, to install demo dependencies
 pip install flash-attn --no-build-isolation  # optional, to use flash attention 2. Will fallback to torch SDPA if not installed
 ```
 
+For installation on AMD ROCm, don't forget to specify the correct `pytorch` version and index, e.g.
+```bash
+pip install liquid-audio torch==2.12.0+rocm7.2 --extra-index-url https://download.pytorch.org/whl/rocm7.2
+```
+
 ## Usage
 Generation is handled by two generation modes, interleaved and sequential, accessible from the methods `LFM2AudioModel.generate_interleaved` and `LFM2AudioModel.generate_sequential` respectively. Both are generators that yield `torch.Tensor`s. Text tokens are represented by tensors with 1 entry, and audio tokens are tensors with 8 entries, corresponding to 8 [Mimi](https://huggingface.co/docs/transformers/en/model_doc/mimi) codebooks.
 
@@ -60,7 +66,7 @@ https://github.com/user-attachments/assets/d0d054b2-6d1d-49fb-94df-4aa0b6641990
 
 ```python
 import torch
-import torchaudio
+import soundfile as sf
 from liquid_audio import LFM2AudioModel, LFM2AudioProcessor, ChatState, LFMModality
 
 # Load models
@@ -77,7 +83,8 @@ chat.add_text("Respond with interleaved text and audio.")
 chat.end_turn()
 
 chat.new_turn("user")
-wav, sampling_rate = torchaudio.load("assets/question.wav")
+wav, sampling_rate = sf.read("assets/question.wav", dtype="float32")
+wav = torch.from_numpy(wav).unsqueeze(0)
 chat.add_audio(wav, sampling_rate)
 chat.end_turn()
 
@@ -102,7 +109,7 @@ for t in model.generate_interleaved(**chat, max_new_tokens=512, audio_temperatur
 # Mimi returns audio at 24kHz
 audio_codes = torch.stack(audio_out[:-1], 1).unsqueeze(0)
 waveform = processor.decode(audio_codes)
-torchaudio.save("answer1.wav", waveform.cpu(), 24_000)
+sf.write("answer1.wav", waveform.cpu()[0], 24_000)
 
 # Append newly generated tokens to chat history
 chat.append(
@@ -132,7 +139,7 @@ for t in model.generate_interleaved(**chat, max_new_tokens=512, audio_temperatur
 # Detokenize second turn audio, removing the last "end-of-audio" codes
 audio_codes = torch.stack(audio_out[:-1], 1).unsqueeze(0)
 waveform = processor.decode(audio_codes)
-torchaudio.save("answer2.wav", waveform.cpu(), 24_000)
+sf.write("answer2.wav", waveform.cpu()[0], 24_000)
 ```
 
 
@@ -151,7 +158,7 @@ https://github.com/user-attachments/assets/b3cc017f-363d-49f3-8e7d-f6db9556900e
 
 ```python
 import torch
-import torchaudio
+import soundfile as sf
 from liquid_audio import LFM2AudioModel, LFM2AudioProcessor, ChatState, LFMModality
 
 # Load models
@@ -168,7 +175,8 @@ chat.add_text("Perform ASR.")
 chat.end_turn()
 
 chat.new_turn("user")
-wav, sampling_rate = torchaudio.load("assets/asr.wav")
+wav, sampling_rate = sf.read("assets/asr.wav", dtype="float32")
+wav = torch.from_numpy(wav).unsqueeze(0)
 chat.add_audio(wav, sampling_rate)
 chat.end_turn()
 
@@ -207,7 +215,7 @@ https://github.com/user-attachments/assets/8d57c184-b92e-4e1a-983b-d1f9d16d0d92
 
 ```python
 import torch
-import torchaudio
+import soundfile as sf
 from liquid_audio import LFM2AudioModel, LFM2AudioProcessor, ChatState, LFMModality
 
 # Load models
@@ -238,7 +246,7 @@ for t in model.generate_sequential(**chat, max_new_tokens=512, audio_temperature
 # Detokenize audio
 audio_codes = torch.stack(audio_out[:-1], 1).unsqueeze(0)
 waveform = processor.decode(audio_codes)
-torchaudio.save("tts.wav", waveform.cpu(), 24_000)
+sf.write("tts.wav", waveform.cpu()[0], 24_000)
 ```
 
 ## Finetuning
@@ -249,12 +257,6 @@ To finetune on your own data, make use of the `ChatMessage` interface. This requ
 2. use the [`LFM2AudioChatMapper`](src/liquid_audio/data/mapper.py) to create a preprocessed dataset
 3. train a model from the preprocessed dataset with `LFM2DataLoader`
 
-First, install project dependencies:
-
-```bash
-uv sync
-```
-
 ### Preprocess
 
 Before training, convert dataset into our preprocessed training format.
@@ -274,7 +276,7 @@ See [examples/preprocess_jenny_tts.py](examples/preprocess_jenny_tts.py) for an
 Run preprocessing with:
 
 ```bash
-python -m examples.preprocess_jenny_tts
+python examples/preprocess_jenny_tts.py
 ```
 
 This writes a preprocessed dataset to `data/jenny_tts/train`.
@@ -287,7 +289,7 @@ For example, to finetune a model on the [Jenny TTS Dataset](https://huggingface.
 using the preprocessed dataset from before, run:
 
 ```bash
-python -m examples.train
+python examples/train.py
 ```
 
 

pyproject.toml

@@ -17,7 +17,6 @@ dependencies = [
     "sentencepiece>=0.2.1",
     "torch>=2.8.0",
     "torchaudio>=2.8.0",
-    "torchcodec>=0.9.1",
     "transformers>=4.55.4",
 ]
 keywords = ["Liquid AI", "LFM", "LFM2", "Audio", "Speech-to-Speech"]

src/liquid_audio/data/mapper.py

@@ -2,6 +2,7 @@
 
 import io
 
+import soundfile
 import torch
 import torchaudio
 
@@ -233,7 +234,8 @@ def _encode_audio_out(self, *, wav: torch.Tensor, sampling_rate: int) -> torch.T
     @staticmethod
     def _load_audio_bytes(audio: bytes) -> tuple[torch.Tensor, int]:
         with io.BytesIO(audio) as stream:
-            wav, sampling_rate = torchaudio.load(stream)
+            data, sampling_rate = soundfile.read(stream, dtype="float32", always_2d=True)
+        wav = torch.from_numpy(data.T.copy())
         if wav.shape[0] > 1:
             wav = wav.mean(dim=0, keepdim=True)
         return wav, sampling_rate