reach-vb HF staff commited on
Commit
cb4a4fc
1 Parent(s): 76a062d

Update Model URLs + Notification about PyTorch versions (#2)

Browse files

- Update Model URLs + Notification about PyTorch versions (814ee4dacc231a8f9e05666876014841ce123ffd)

Files changed (1) hide show
  1. README.md +9 -7
README.md CHANGED
@@ -3,6 +3,7 @@ inference: false
3
  tags:
4
  - SeamlessM4T
5
  license: cc-by-nc-4.0
 
6
  ---
7
 
8
  # SeamlessM4T - On-Device
@@ -10,20 +11,21 @@ SeamlessM4T is designed to provide high quality translation, allowing people fro
10
 
11
  SeamlessM4T covers:
12
  - 📥 101 languages for speech input
13
- - ⌨️ 96 Languages for text input/output
14
- - 🗣️ 35 languages for speech output.
15
 
16
  Apart from [SeamlessM4T-LARGE (2.3B)](https://huggingface.co/facebook/seamless-m4t-large) and [SeamlessM4T-MEDIUM (1.2B)](https://huggingface.co/facebook/seamless-m4t-medium) models, we are also developing a small model (281M) targeting for on-device inference.
17
  [This folder](https://huggingface.co/facebook/seamless-m4t-unity-small) contains an example to run an exported small model covering most tasks (ASR/S2TT/S2ST). The model could be executed on popular mobile devices with Pytorch Mobile (https://pytorch.org/mobile/home/).
18
 
19
  ## Overview
 
 
 
 
20
 
21
- | Model | Disk Size | Supported Tasks | Supported Languages|
22
- |---------|----------------------|-------------------------|-------------------------|
23
- | [UnitY-Small](https://huggingface.co/facebook/seamless-m4t-unity-small/resolve/main/unity_on_device.ptl) | 862MB | S2ST, S2TT, ASR |eng, fra, hin, por, spa|
24
- | [UnitY-Small-S2T](https://huggingface.co/facebook/seamless-m4t-unity-small-s2t/resolve/main/unity_on_device_s2t.ptl) | 637MB | S2TT, ASR |eng, fra, hin, por, spa|
25
 
26
- [UnitY-Small-S2T](https://huggingface.co/facebook/seamless-m4t-unity-small-s2t) is a pruned version of [UnitY-Small](https://huggingface.co/facebook/seamless-m4t-unity-small) without 2nd pass unit decoding. Unlike [UnitY-Small](https://huggingface.co/facebook/seamless-m4t-unity-small), it can only be used for ASR and S2TT tasks.
27
 
28
  ## Inference
29
  To use exported model, users don't need seamless_communication or fairseq2 dependency.
 
3
  tags:
4
  - SeamlessM4T
5
  license: cc-by-nc-4.0
6
+ library_name: fairseq2
7
  ---
8
 
9
  # SeamlessM4T - On-Device
 
11
 
12
  SeamlessM4T covers:
13
  - 📥 101 languages for speech input
14
+ - ⌨️ 96 Languages for text input/output
15
+ - 🗣️ 35 languages for speech output.
16
 
17
  Apart from [SeamlessM4T-LARGE (2.3B)](https://huggingface.co/facebook/seamless-m4t-large) and [SeamlessM4T-MEDIUM (1.2B)](https://huggingface.co/facebook/seamless-m4t-medium) models, we are also developing a small model (281M) targeting for on-device inference.
18
  [This folder](https://huggingface.co/facebook/seamless-m4t-unity-small) contains an example to run an exported small model covering most tasks (ASR/S2TT/S2ST). The model could be executed on popular mobile devices with Pytorch Mobile (https://pytorch.org/mobile/home/).
19
 
20
  ## Overview
21
+ | Model | Checkpoint | Num Params | Disk Size | Supported Tasks | Supported Languages|
22
+ |---------|------------|----------|-------------|------------|-------------------------|
23
+ | UnitY-Small|[🤗 Model card](https://huggingface.co/facebook/seamless-m4t-unity-small) - [checkpoint](https://huggingface.co/facebook/seamless-m4t-unity-small/resolve/main/unity_on_device.ptl) | 281M | 862MB | S2ST, S2TT, ASR |eng, fra, hin, por, spa|
24
+ | UnitY-Small-S2T |[🤗 Model card](https://huggingface.co/facebook/seamless-m4t-unity-small-s2t) - [checkpoint](https://huggingface.co/facebook/seamless-m4t-unity-small-s2t/resolve/main/unity_on_device_s2t.ptl) | 235M | 637MB | S2TT, ASR |eng, fra,hin, por, spa|
25
 
26
+ UnitY-Small-S2T is a pruned version of UnitY-Small without 2nd pass unit decoding.
 
 
 
27
 
28
+ Note: If using pytorch runtime in python, only **pytorch<=1.11.0** is supported for **UnitY-Small(281M)**. We tested UnitY-Small-S2T(235M), it works with later versions.
29
 
30
  ## Inference
31
  To use exported model, users don't need seamless_communication or fairseq2 dependency.