diff --git a/examples/seld_spatialsoundqa/README.md b/examples/seld_spatialsoundqa/README.md
index 512d7490..2fbae126 100644
--- a/examples/seld_spatialsoundqa/README.md
+++ b/examples/seld_spatialsoundqa/README.md
@@ -1,39 +1,65 @@
# 
SELD_SpatialSoundQA
-This repo hosts the code and models of "[BAT: Learning to Reason about Spatial Sounds with Large Language Models](https://arxiv.org/abs/2402.01591)" [ICML 2024 [bib](https://github.com/zszheng147/Spatial-AST#citation)].
+This repo hosts the code and models of "[BAT: Learning to Reason about Spatial Sounds with Large Language Models](https://arxiv.org/abs/2402.01591)" [ICML 2024 [bib](https://github.com/X-LANCE/SLAM-LLM/tree/main/examples/seld_spatialsoundqa#citation)].
Checkout our [demo page](https://zhishengzheng.com/BAT/) and enjoy a QA game with spatial audio.
-## Performance and checkpoints
-Encoder | Projector | PEFT | LLM
-|---|---|---|---|
-[Spatial-AST](https://huggingface.co/zhisheng01/Bat/blob/main/spatial-ast.pth) | Q-Former | adapter |[llama-2-7b](https://huggingface.co/meta-llama/Llama-2-7b)
+## Performance evaluation on **SpatialSoundQA**
+We use [Spatial-AST](https://huggingface.co/datasets/zhisheng01/SpatialAudio/blob/main/SpatialAST/finetuned.pth) as audio encoder, [llama-2-7b](https://huggingface.co/meta-llama/Llama-2-7b) as LLM backbone. We finetune the model by adding Q-Former and LoRA. To calculate MAP, you can refer to [calculate_map.py](https://github.com/X-LANCE/SLAM-LLM/blob/main/examples/seld_spatialsoundqa/scripts/calculate_map.py)
+
+
+
+## Demo (Spatial Audio Inference)
+Try [`inference.ipynb`](https://github.com/X-LANCE/SLAM-LLM/blob/main/examples/seld_spatialsoundqa/inference.ipynb).
+
## Data preparation
You need to prepare the data jsonl in this format. Below is an example.
-You can download the SpatialSoundQA dataset from [huggingface](https://huggingface.co/datasets/zhisheng01/SpatialSoundQA).
-```
-{"audio_id": "eval/audio/YI-HlrcP6Qg4", "reverb_id": "q9vSo1VnCiC/0.npy", "audio_id2": null, "reverb_id2": null, "question_id": 0, "question_type": "CLASSIFICATION", "question": "Enumerate the sound occurrences in the audio clip.", "answer": "accelerating, revving, vroom; car; vehicle"}
+You can download the SpatialSoundQA dataset from [SpatialAudio](https://huggingface.co/datasets/zhisheng01/SpatialAudio).
+```json
+{
+ "audio_id": "eval/audio/YI-HlrcP6Qg4",
+ "reverb_id": "q9vSo1VnCiC/0.npy",
+ "audio_id2": null,
+ "reverb_id2": null,
+ "question_id": 0,
+ "question_type": "CLASSIFICATION",
+ "question": "Enumerate the sound occurrences in the audio clip.",
+ "answer": "accelerating, revving, vroom; car; vehicle"
+}
+
...
-{"audio_id": "eval/audio/YZX2fVPmUidA", "reverb_id": "q9vSo1VnCiC/32.npy", "audio_id2": "eval/audio/YjNjUU01quLs", "reverb_id2": "q9vSo1VnCiC/31.npy", "question_id": 58, "question_type": "MIXUP_NONBINARY_DISTANCE", "question": "How far away is the sound of the banjo from the sound of the whack, thwack?", "answer": "2m"}
+
+{
+ "audio_id": "eval/audio/YZX2fVPmUidA",
+ "reverb_id": "q9vSo1VnCiC/32.npy",
+ "audio_id2": "eval/audio/YjNjUU01quLs",
+ "reverb_id2": "q9vSo1VnCiC/31.npy",
+ "question_id": 58,
+ "question_type": "MIXUP_NONBINARY_DISTANCE",
+ "question": "How far away is the sound of the banjo from the sound of the whack, thwack?",
+ "answer": "2m"
+}
```
## Train a new model
```bash
-bash examples/seld_spatialsoundqa/scripts/finetune_spatial-ast_qformer_llama_2_7b.sh
+cd examples/seld_spatialsoundqa/
+bash scripts/finetune_spatial-ast_qformer_llama_2_7b.sh
```
## Decoding with checkpoints
```bash
-bash examples/seld_spatialsoundqa/scripts/decode_spatial-ast_qformer_llama_2_7b.sh
+cd examples/seld_spatialsoundqa/
+bash scripts/decode_spatial-ast_qformer_llama_2_7b.sh
```
## TODO
- [x] Decode with checkpoints
- [x] Upload SpatialSoundQA dataset
-- [ ] Upload pretrained checkpoints
-- [ ] Update model performance
+- [x] Upload pretrained checkpoints
+- [x] Update model performance
## Citation
```
diff --git a/examples/seld_spatialsoundqa/assets/74.npy b/examples/seld_spatialsoundqa/assets/74.npy
new file mode 100644
index 00000000..cc16faaa
Binary files /dev/null and b/examples/seld_spatialsoundqa/assets/74.npy differ
diff --git a/examples/seld_spatialsoundqa/assets/75.npy b/examples/seld_spatialsoundqa/assets/75.npy
new file mode 100644
index 00000000..f1644629
Binary files /dev/null and b/examples/seld_spatialsoundqa/assets/75.npy differ
diff --git a/examples/seld_spatialsoundqa/assets/YCqvbWnTBfTk.wav b/examples/seld_spatialsoundqa/assets/YCqvbWnTBfTk.wav
new file mode 100644
index 00000000..e84270b7
Binary files /dev/null and b/examples/seld_spatialsoundqa/assets/YCqvbWnTBfTk.wav differ
diff --git a/examples/seld_spatialsoundqa/assets/Yq4Z8j3IalYs.wav b/examples/seld_spatialsoundqa/assets/Yq4Z8j3IalYs.wav
new file mode 100644
index 00000000..a5ec3f01
Binary files /dev/null and b/examples/seld_spatialsoundqa/assets/Yq4Z8j3IalYs.wav differ
diff --git a/examples/seld_spatialsoundqa/assets/performance.png b/examples/seld_spatialsoundqa/assets/performance.png
new file mode 100644
index 00000000..0393d3b9
Binary files /dev/null and b/examples/seld_spatialsoundqa/assets/performance.png differ
diff --git a/examples/seld_spatialsoundqa/dataset/spatial_audio_dataset.py b/examples/seld_spatialsoundqa/dataset/spatial_audio_dataset.py
index f986250f..c3906258 100644
--- a/examples/seld_spatialsoundqa/dataset/spatial_audio_dataset.py
+++ b/examples/seld_spatialsoundqa/dataset/spatial_audio_dataset.py
@@ -37,9 +37,8 @@ def __init__(
split,
):
super().__init__()
- dataset_path = os.path.join(dataset_config['qa_data_root'], dataset_config['stage'], split + '.jsonl')
- with open(dataset_path) as f:
- self.data = [json.loads(line) for line in f.readlines()]
+ dataset_path = os.path.join(dataset_config['qa_data_root'], dataset_config['stage'], split + '.json')
+ self.data = json.load(open(dataset_path))["data"]
self.anechoic_data_root = dataset_config['anechoic_data_root'] # which is AudioSet in this case
self.reverb_data_root = dataset_config['reverb_data_root']
diff --git a/examples/seld_spatialsoundqa/finetune_seld.py b/examples/seld_spatialsoundqa/finetune_seld.py
index 3b7a959d..605f3582 100644
--- a/examples/seld_spatialsoundqa/finetune_seld.py
+++ b/examples/seld_spatialsoundqa/finetune_seld.py
@@ -1,5 +1,6 @@
import hydra
import logging
+from typing import Optional
from dataclasses import dataclass, field
from omegaconf import DictConfig, ListConfig, OmegaConf
@@ -16,32 +17,20 @@ class RunConfig:
peft_config: PeftConfig = field(default_factory=PeftConfig)
debug: bool = field(default=False, metadata={"help": "Use pdb when true"})
metric: str = field(default="acc", metadata={"help": "The metric for evaluation"})
- ckpt_path: str = field(
- default="output/model.pt", metadata={"help": "The path to projector checkpoint"}
+ ckpt_path: Optional[str] = field(
+ default=None, metadata={"help": "The path to projector checkpoint"}
)
@hydra.main(config_name=None, version_base=None)
def main_hydra(cfg: DictConfig):
run_config = RunConfig()
cfg = OmegaConf.merge(run_config, cfg)
- def to_plain_list(cfg_item):
- if isinstance(cfg_item, ListConfig):
- return OmegaConf.to_container(cfg_item, resolve=True)
- elif isinstance(cfg_item, DictConfig):
- return {k: to_plain_list(v) for k, v in cfg_item.items()}
- else:
- return cfg_item
-
- # kwargs = to_plain_list(cfg)
- kwargs = cfg
- log_level = getattr(logging, kwargs.get("log_level", "INFO").upper())
+ cfg.train_config.peft_config = cfg.peft_config
+
+ log_level = getattr(logging, cfg.get("log_level", "INFO").upper())
logging.basicConfig(level=log_level)
-
- if kwargs.get("debug", False):
- import pdb;
- pdb.set_trace()
- train(kwargs)
+ train(cfg)
if __name__ == "__main__":
diff --git a/examples/seld_spatialsoundqa/inference.ipynb b/examples/seld_spatialsoundqa/inference.ipynb
new file mode 100644
index 00000000..03e29bbe
--- /dev/null
+++ b/examples/seld_spatialsoundqa/inference.ipynb
@@ -0,0 +1,786 @@
+{
+ "cells": [
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "# Let's dive into a spatial sound world."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 1,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "import os\n",
+ "\n",
+ "import numpy as np\n",
+ "import soundfile as sf\n",
+ "from scipy import signal\n",
+ "from IPython.display import Audio\n",
+ "\n",
+ "import torch\n",
+ "\n",
+ "from dataset.spatial_audio_dataset import format_prompt, SpatialAudioDatasetJsonl"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 2,
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "Let's load and listen to anechoic audio...\n",
+ "Audio 1: Drum; Percussion\n"
+ ]
+ },
+ {
+ "data": {
+ "text/html": [
+ "\n",
+ " \n",
+ " "
+ ],
+ "text/plain": [
+ ""
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "Audio 2: Emergency vehicle; Fire engine, fire truck (siren); Siren\n"
+ ]
+ },
+ {
+ "data": {
+ "text/html": [
+ "\n",
+ " \n",
+ " "
+ ],
+ "text/plain": [
+ ""
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ }
+ ],
+ "source": [
+ "audio_path1 = \"./assets/YCqvbWnTBfTk.wav\"\n",
+ "audio_path2 = \"./assets/Yq4Z8j3IalYs.wav\"\n",
+ "\n",
+ "print(\"Let's load and listen to anechoic audio...\")\n",
+ "print(\"Audio 1: Drum; Percussion\")\n",
+ "display(Audio(audio_path1))\n",
+ "\n",
+ "print(\"Audio 2: Emergency vehicle; Fire engine, fire truck (siren); Siren\")\n",
+ "display(Audio(audio_path2))"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 3,
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "Let's load and listen to reverb audio 1 (w/o mixup)...\n"
+ ]
+ },
+ {
+ "data": {
+ "text/html": [
+ "\n",
+ " \n",
+ " "
+ ],
+ "text/plain": [
+ ""
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "Let's load and listen to reverb audio 2 (w/o mixup)...\n"
+ ]
+ },
+ {
+ "data": {
+ "text/html": [
+ "\n",
+ " \n",
+ " "
+ ],
+ "text/plain": [
+ ""
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ }
+ ],
+ "source": [
+ "def reverb_waveform(audio_path, reverb_path):\n",
+ " waveform, sr = sf.read(audio_path)\n",
+ " if len(waveform.shape) > 1:\n",
+ " waveform = waveform[:, 0]\n",
+ " if sr != 32000: #! Please make sure the audio is 32000 Hz\n",
+ " waveform = signal.resample_poly(waveform, 32000, sr)\n",
+ " sr = 32000\n",
+ " waveform = SpatialAudioDatasetJsonl.normalize_audio(waveform, -14.0).reshape(1, -1)\n",
+ " reverb = np.load(reverb_path)\n",
+ " waveform = signal.fftconvolve(waveform, reverb, mode='full')\n",
+ " return waveform, sr\n",
+ "\n",
+ "reverb_path1 = \"./assets/74.npy\"\n",
+ "reverb_path2 = \"./assets/75.npy\"\n",
+ "\"\"\"\n",
+ "{\"fname\": \"q9vSo1VnCiC/74.npy\", \"agent_position\": \"-12.8775,0.0801,8.415\", \"sensor_position\": \"-12.8775,1.5801,8.415\", \"source_position\": \"-13.4677,1.2183,8.6525\",},\n",
+ "{\"fname\": \"q9vSo1VnCiC/75.npy\", \"agent_position\": \"-12.8775,0.0801,8.415\", \"sensor_position\": \"-12.8775,1.5801,8.415\", \"source_position\": \"-11.8976,1.0163,8.9789\",}\n",
+ "\"\"\"\n",
+ "\n",
+ "print(\"Let's load and listen to reverb audio 1 (w/o mixup)...\")\n",
+ "waveform1, _ = reverb_waveform(audio_path1, reverb_path1)\n",
+ "display(Audio(waveform1, rate=32000))\n",
+ "\n",
+ "print(\"Let's load and listen to reverb audio 2 (w/o mixup)...\")\n",
+ "waveform2, _ = reverb_waveform(audio_path2, reverb_path2)\n",
+ "display(Audio(waveform2, rate=32000))"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 4,
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "Now let's mix them up!\n"
+ ]
+ },
+ {
+ "data": {
+ "text/html": [
+ "\n",
+ " \n",
+ " "
+ ],
+ "text/plain": [
+ ""
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ }
+ ],
+ "source": [
+ "print(\"Now let's mix them up!\")\n",
+ "if waveform1.shape[1] < waveform2.shape[1]:\n",
+ " waveform2 = waveform2[:, :waveform1.shape[1]]\n",
+ "else:\n",
+ " waveform1 = waveform1[:, :waveform2.shape[1]]\n",
+ "waveform = (waveform1 + waveform2) / 2\n",
+ "display(Audio(waveform, rate=32000))"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 5,
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "[\"Based on the audio you've heard, refer to the instruction and provide a response.\\n\\n### Instruction:\\nWhat is the distance between the sound of the drum and the sound of the siren?\\n\\n### Response:\", \"Based on the audio you've heard, refer to the instruction and provide a response.\\n\\n### Instruction:\\nWhat is the sound on the right side of the sound of the drum?\\n\\n### Response:\", \"Based on the audio you've heard, refer to the instruction and provide a response.\\n\\n### Instruction:\\nAre you able to detect the percussion's sound coming from the left and the emergency vehicle's sounds from the right?\\n\\n### Response:\"]\n"
+ ]
+ }
+ ],
+ "source": [
+ "prompts = [\n",
+ " \"What is the distance between the sound of the drum and the sound of the siren?\",\n",
+ " \"What is the sound on the right side of the sound of the drum?\",\n",
+ " \"Are you able to detect the percussion's sound coming from the left and the emergency vehicle's sounds from the right?\",\n",
+ "]\n",
+ "\n",
+ "gt_answers = [\n",
+ " \"1.5m\",\n",
+ " \"emergency vehicle; fire engine, fire truck (siren); siren\",\n",
+ " \"Yes\"\n",
+ "]\n",
+ "\n",
+ "prompts = [format_prompt(prompt) for prompt in prompts]\n",
+ "print(prompts)"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 6,
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "/home/zhisheng/miniconda3/envs/speech/lib/python3.8/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
+ " from .autonotebook import tqdm as notebook_tqdm\n",
+ "Loading checkpoint shards: 0%| | 0/2 [00:00