alvanli
commited on
Commit
•
8034c9c
1
Parent(s):
b14b2d8
cer 7.93
Browse files- README.md +5 -4
- config.json +1 -1
- model.safetensors +1 -1
- optimizer.pt +1 -1
README.md
CHANGED
@@ -23,7 +23,7 @@ model-index:
|
|
23 |
metrics:
|
24 |
- name: Normalized CER
|
25 |
type: cer
|
26 |
-
value:
|
27 |
---
|
28 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
29 |
should probably proofread and complete it, then remove this comment. -->
|
@@ -41,17 +41,18 @@ For training,
|
|
41 |
|--|--|
|
42 |
|Common Voice 16.0 zh-HK Train|138|
|
43 |
|Common Voice 16.0 yue Train|85|
|
|
|
44 |
|Cantonese-ASR|72|
|
45 |
|CantoMap|23|
|
46 |
|[Pseudo-Labelled YouTube Data](https://huggingface.co/datasets/alvanlii/cantonese-youtube-pseudo-transcription)|438|
|
47 |
-
|Total|756|
|
48 |
|
49 |
|
50 |
For evaluation, Common Voice 16.0 yue Test set is used.
|
51 |
|
52 |
## Results
|
53 |
-
- CER (lower is better): 0.
|
54 |
-
- down from 0.1581 in the previous
|
|
|
55 |
- GPU Inference with Fast Attention (example below): 0.055s/sample
|
56 |
- Note all GPU evaluations are done on RTX 3090 GPU
|
57 |
- GPU Inference: 0.308s/sample
|
|
|
23 |
metrics:
|
24 |
- name: Normalized CER
|
25 |
type: cer
|
26 |
+
value: 7.93
|
27 |
---
|
28 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
29 |
should probably proofread and complete it, then remove this comment. -->
|
|
|
41 |
|--|--|
|
42 |
|Common Voice 16.0 zh-HK Train|138|
|
43 |
|Common Voice 16.0 yue Train|85|
|
44 |
+
|Common Voice 17.0 yue Train|178|
|
45 |
|Cantonese-ASR|72|
|
46 |
|CantoMap|23|
|
47 |
|[Pseudo-Labelled YouTube Data](https://huggingface.co/datasets/alvanlii/cantonese-youtube-pseudo-transcription)|438|
|
|
|
48 |
|
49 |
|
50 |
For evaluation, Common Voice 16.0 yue Test set is used.
|
51 |
|
52 |
## Results
|
53 |
+
- CER (lower is better): 0.0972
|
54 |
+
- down from 0.1073, 0.1581 in the previous versions
|
55 |
+
- CER (punctuations removed): 0.0793
|
56 |
- GPU Inference with Fast Attention (example below): 0.055s/sample
|
57 |
- Note all GPU evaluations are done on RTX 3090 GPU
|
58 |
- GPU Inference: 0.308s/sample
|
config.json
CHANGED
@@ -45,7 +45,7 @@
|
|
45 |
"scale_embedding": false,
|
46 |
"suppress_tokens": [],
|
47 |
"torch_dtype": "float32",
|
48 |
-
"transformers_version": "4.
|
49 |
"use_cache": false,
|
50 |
"use_weighted_layer_sum": false,
|
51 |
"vocab_size": 51865
|
|
|
45 |
"scale_embedding": false,
|
46 |
"suppress_tokens": [],
|
47 |
"torch_dtype": "float32",
|
48 |
+
"transformers_version": "4.39.3",
|
49 |
"use_cache": false,
|
50 |
"use_weighted_layer_sum": false,
|
51 |
"vocab_size": 51865
|
model.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 966995080
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:fde842c011b43b2a7b988ad9b9f1a082731bc29c3d58b0cdffa3e71252ef7256
|
3 |
size 966995080
|
optimizer.pt
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 1925064044
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:52d3d75f188b0e83816ddfe2c081aa5811fef9f12f90bcf95efbb6f22d4cb03a
|
3 |
size 1925064044
|