Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add small-scaled model for audio tagging #1604

Merged
merged 1 commit into from
Apr 24, 2024
Merged
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
51 changes: 51 additions & 0 deletions egs/audioset/AT/RESULTS.md
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,8 @@ See <https://github.com/k2-fsa/icefall/pull/1421> for more details

[zipformer](./zipformer)

#### normal-scaled model, number of model parameters: 65549011, i.e., 65.55 M

You can find a pretrained model, training logs, decoding logs, and decoding results at:
<https://huggingface.co/marcoyang/icefall-audio-tagging-audioset-zipformer-2024-03-12#/>

Expand Down Expand Up @@ -42,3 +44,52 @@ python zipformer/evaluate.py \
--exp-dir zipformer/exp_at_as_full \
--max-duration 500
```


#### small-scaled model, number of model parameters: 22125218, i.e., 22.13 M

You can find a pretrained model, training logs, decoding logs, and decoding results at:
<https://huggingface.co/marcoyang/icefall-audio-tagging-audioset-zipformer-small-2024-04-23#/>

The model achieves the following mean averaged precision on AudioSet:

| Model | mAP |
| ------ | ------- |
| Zipformer-S-AT | 45.1 |

The training command is:

```bash
export CUDA_VISIBLE_DEVICES="4,5,6,7"
subset=full

python zipformer/train.py \
--world-size 4 \
--num-epochs 50 \
--exp-dir zipformer/exp_small_at_as_${subset} \
--start-epoch 1 \
--use-fp16 1 \
--num-events 527 \
--num-encoder-layers 2,2,2,2,2,2 \
--feedforward-dim 512,768,768,768,768,768 \
--encoder-dim 192,256,256,256,256,256 \
--encoder-unmasked-dim 192,192,192,192,192,192 \
--audioset-subset $subset \
--max-duration 1200 \
--enable-musan True \
--master-port 13455
```

The evaluation command is:

```bash
python zipformer/evaluate.py \
--epoch 31 \
--avg 4 \
--num-encoder-layers 2,2,2,2,2,2 \
--feedforward-dim 512,768,768,768,768,768 \
--encoder-dim 192,256,256,256,256,256 \
--encoder-unmasked-dim 192,192,192,192,192,192 \
--exp-dir zipformer/exp_small_at_as_full \
--max-duration 500
```
Loading