Improve language tag
Browse filesHi! As the model is multilingual, this is a PR to add other languages than English to the language tag to improve the referencing. Note that 29 languages are announced in the README, but only 13 are explicitly listed. I was therefore only able to add these 13 languages.
README.md
CHANGED
@@ -1,87 +1,99 @@
|
|
1 |
-
---
|
2 |
-
library_name: transformers
|
3 |
-
license: mit
|
4 |
-
datasets:
|
5 |
-
- slprl/sTinyStories
|
6 |
-
language:
|
7 |
-
-
|
8 |
-
|
9 |
-
-
|
10 |
-
|
11 |
-
|
12 |
-
|
13 |
-
|
14 |
-
|
15 |
-
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
-
|
48 |
-
-
|
49 |
-
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
|
59 |
-
|
60 |
-
|
61 |
-
|
62 |
-
##
|
63 |
-
|
64 |
-
|
65 |
-
|
66 |
-
###
|
67 |
-
|
68 |
-
|
69 |
-
|
70 |
-
|
71 |
-
|
72 |
-
|
73 |
-
|
74 |
-
##
|
75 |
-
|
76 |
-
|
77 |
-
|
78 |
-
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
-
|
83 |
-
|
84 |
-
|
85 |
-
|
86 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
87 |
```
|
|
|
1 |
+
---
|
2 |
+
library_name: transformers
|
3 |
+
license: mit
|
4 |
+
datasets:
|
5 |
+
- slprl/sTinyStories
|
6 |
+
language:
|
7 |
+
- zho
|
8 |
+
- eng
|
9 |
+
- fra
|
10 |
+
- spa
|
11 |
+
- por
|
12 |
+
- deu
|
13 |
+
- ita
|
14 |
+
- rus
|
15 |
+
- jpn
|
16 |
+
- kor
|
17 |
+
- vie
|
18 |
+
- tha
|
19 |
+
- ara
|
20 |
+
base_model:
|
21 |
+
- Qwen/Qwen2.5-7B
|
22 |
+
pipeline_tag: audio-to-audio
|
23 |
+
---
|
24 |
+
|
25 |
+
# Scaling Analysis of Interleaved Speech-Text Language Models
|
26 |
+
|
27 |
+
The model was presented in the paper [Scaling Analysis of Interleaved Speech-Text Language Models](https://arxiv.org/abs/2504.02398).
|
28 |
+
|
29 |
+
# Paper abstract
|
30 |
+
Existing Speech Language Model (SLM) scaling analysis paints a bleak picture. They predict that SLMs require much more compute and data
|
31 |
+
compared to text, leading some to question the feasibility of training high-quality SLMs. However, modern SLMs are often initialised from
|
32 |
+
pre-trained TextLMs using speech-text interleaving to allow knowledge transfer. This raises the question - _Do interleaved SLMs scale more efficiently than textless-SLMs?_
|
33 |
+
In this paper we answer a resounding _yes!_ We conduct scaling analysis of interleaved SLMs by training several dozen and analysing the
|
34 |
+
scaling trends. We see that under this setup SLMs scale more efficiently with compute. Additionally, our results indicate that the
|
35 |
+
scaling-dynamics are significantly different than textless-SLMs, suggesting one should allocate notably more of the compute budget for
|
36 |
+
increasing model size over training tokens. We also study the role of synthetic data and TextLM model families in unlocking this potential.
|
37 |
+
Results suggest, that our scaled up model achieves comparable performance with leading models on speech semantic metrics while using less
|
38 |
+
compute and data than other approaches.
|
39 |
+
|
40 |
+
# Model Card for Model ID
|
41 |
+
This is a Speech Language Model (SLM) trained for generating speech or text continuations over discrete [Hubert tokens](https://huggingface.co/slprl/mhubert-base-25hz) given speech-text prompts.
|
42 |
+
|
43 |
+
|
44 |
+
## Model Details
|
45 |
+
|
46 |
+
### Model Description
|
47 |
+
This Speech Language Model, introduced in ["Scaling Analysis of Interleaved Speech-Text Language Models"](https://arxiv.org/abs/2504.02398), focuses on scaling analysis of interleaved speech-text SLMs.
|
48 |
+
It was fine-tuned from [Qwen/Qwen2.5-7B](https://huggingface.co/Qwen/Qwen2.5-7B) by extending its vocabulary with 500 speech tokens extracted from
|
49 |
+
the 11-th layer of [mhubert-25hz](https://huggingface.co/slprl/mhubert-base-25hz).
|
50 |
+
|
51 |
+
- **Developed by:** [SLP-RL](https://huggingface.co/slprl)
|
52 |
+
- **Model type:** SpeechLM
|
53 |
+
- **License:** MIT
|
54 |
+
- **Finetuned from model:** [Qwen/Qwen2.5-7B](https://huggingface.co/Qwen/Qwen2.5-7B)
|
55 |
+
|
56 |
+
### Model Sources
|
57 |
+
|
58 |
+
- **Repository:** [https://github.com/slp-rl/slamkit](https://github.com/slp-rl/slamkit)
|
59 |
+
- **Paper:** [https://arxiv.org/abs/2504.02398](https://arxiv.org/abs/2504.02398)
|
60 |
+
- **Demo:** [https://pages.cs.huji.ac.il/adiyoss-lab/sims/](https://pages.cs.huji.ac.il/adiyoss-lab/sims/)
|
61 |
+
|
62 |
+
## Uses
|
63 |
+
This base SpeechLM can be used to generate continuations for speech segments, or cross-modal e.g generate a text contiuation to a speech prompt, or as a base for further tuning. See the _SlamKit_
|
64 |
+
[codebase](https://github.com/slp-rl/slamkit) for more details on usage, and checkout the [demo page](https://pages.cs.huji.ac.il/adiyoss-lab/sims/) for some generation examples
|
65 |
+
|
66 |
+
### Out-of-Scope Use
|
67 |
+
This model was trained on diverse speech datasets, as such the outputs should not be treated as factual in any way.
|
68 |
+
|
69 |
+
|
70 |
+
## How to Get Started with the Model
|
71 |
+
We refer users to the official repository for full usage explanations - [github](https://github.com/slp-rl/slamkit).
|
72 |
+
|
73 |
+
|
74 |
+
## Training Details
|
75 |
+
We highly encourage users to read the full [paper](https://arxiv.org/abs/2504.02398), for full training details.
|
76 |
+
|
77 |
+
|
78 |
+
### Compute Infrastructure
|
79 |
+
#### Hardware
|
80 |
+
This model was trained using 8 Nvidia H100 GPUs.
|
81 |
+
|
82 |
+
#### Software
|
83 |
+
The model was trained using the [*SlamKit*](https://github.com/slp-rl/slamkit) codebase which builds upon 🤗transformers extending it to support
|
84 |
+
easy and efficient training of Speech Language Models.
|
85 |
+
|
86 |
+
## Citation
|
87 |
+
|
88 |
+
**BibTeX:**
|
89 |
+
```
|
90 |
+
@misc{maimon2025scaling,
|
91 |
+
title={Scaling Analysis of Interleaved Speech-Text Language Models},
|
92 |
+
author={Gallil Maimon and Michael Hassid and Amit Roth and Yossi Adi},
|
93 |
+
year={2025},
|
94 |
+
eprint={2504.02398},
|
95 |
+
archivePrefix={arXiv},
|
96 |
+
primaryClass={cs.CL},
|
97 |
+
url={https://arxiv.org/abs/2504.02398},
|
98 |
+
}
|
99 |
```
|