Wangyou Zhang
commited on
Commit
•
e8da9a0
1
Parent(s):
a863afc
Add models
Browse files- README.md +13 -0
- WavLM-Base+_cpt.pt +3 -0
- WavLM-Base_cpt.pt +3 -0
- wavlm_large_cpt.pt +3 -0
README.md
CHANGED
@@ -1,3 +1,16 @@
|
|
1 |
---
|
2 |
license: mit
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: mit
|
3 |
---
|
4 |
+
|
5 |
+
|
6 |
+
# Note
|
7 |
+
|
8 |
+
The provided checkpoints are adapated from the official WavLM models released at https://github.com/microsoft/unilm/blob/master/wavlm/README.md#pre-trained-models to be used for fine-tuning in fairseq.
|
9 |
+
|
10 |
+
They should be used together with the specialized fairseq repository https://github.com/Emrys365/fairseq/tree/wavlm where the support for WavLM is added.
|
11 |
+
|
12 |
+
> NOTE: The adapted checkpoints are only guaranteed to have the same `model configuration` and `model parameters` as the official released models. The other parameters (such as `task hyperparameters` and `optimizer hyperparameters`) are just a template.
|
13 |
+
>
|
14 |
+
> Therefore, these adapted checkpoints are only suitable to be used as an initialization for fine-tuning in the downstream tasks, **NOT** for continuing pre-training.
|
15 |
+
|
16 |
+
Please refer to https://github.com/Emrys365/fairseq/blob/wavlm/examples/wavlm/README.md#load-adapted-checkpoints-for-fairseq-finetuning for more information.
|
WavLM-Base+_cpt.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:2305c553c301822c4c06522d1e5e7f2575645a08f2b2e8f9a8ef80cf2d96f3a2
|
3 |
+
size 1135291829
|
WavLM-Base_cpt.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:154aa69ff4ec5399f5e2dfce95fc630d81d809d9b5ec322d5cb7f36d68461035
|
3 |
+
size 1135291565
|
wavlm_large_cpt.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:40a12c5cc5e7d61f7f29961e7810b89aa6352894cb7aafd9ed922b10d86f9b3a
|
3 |
+
size 2019658005
|