Mirroristas/whisper
2
0
Fork 0
mirror of https://github.com/openai/whisper.git synced 2025-11-23 22:15:58 +00:00
Code Issues Packages Projects Releases Wiki Activity

large-v3 checkpoint and evals

Browse Source
This commit is contained in:
Jong Wook Kim 2023-11-06 04:35:37 -08:00
parent b7ac4888a2
commit 72fc6ef0ee
4 changed files with 6949 additions and 2062 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8998
language-breakdown.svg
View File

File diff suppressed because it is too large Load Diff
Side by Side Swipe Overlay

Before

Width:  |  Height:  |  Size: 100 KiB

After

Width:  |  Height:  |  Size: 272 KiB

4
model-card.md
View File

@ -17,12 +17,12 @@ The Whisper models are trained for speech recognition and translation tasks, cap
| medium | 769 M | ✓ | ✓ | | medium | 769 M | ✓ | ✓ |
| large | 1550 M | | ✓ | | large | 1550 M | | ✓ |
In December 2022, we [released an improved large model named `large-v2`](https://github.com/openai/whisper/discussions/661). In December 2022, we [released an improved large model named `large-v2`](https://github.com/openai/whisper/discussions/661), and `large-v3` in November 2023.
### Release date ### Release date
September 2022 (original series) and December 2022 (`large-v2`) September 2022 (original series), December 2022 (`large-v2`), and November 2023 (`large-v3`)
### Model type ### Model type

6
whisper/__init__.py
View File

@ -25,7 +25,8 @@ _MODELS = {
"medium": "https://openaipublic.azureedge.net/main/whisper/models/345ae4da62f9b3d59415adc60127b97c714f32e89e936602e85993674d08dcb1/medium.pt", "medium": "https://openaipublic.azureedge.net/main/whisper/models/345ae4da62f9b3d59415adc60127b97c714f32e89e936602e85993674d08dcb1/medium.pt",
"large-v1": "https://openaipublic.azureedge.net/main/whisper/models/e4b87e7e0bf463eb8e6956e646f1e277e901512310def2c24bf0e11bd3c28e9a/large-v1.pt", "large-v1": "https://openaipublic.azureedge.net/main/whisper/models/e4b87e7e0bf463eb8e6956e646f1e277e901512310def2c24bf0e11bd3c28e9a/large-v1.pt",
"large-v2": "https://openaipublic.azureedge.net/main/whisper/models/81f7c96c852ee8fc832187b0132e569d6c3065a3252ed18e56effd0b6a73e524/large-v2.pt", "large-v2": "https://openaipublic.azureedge.net/main/whisper/models/81f7c96c852ee8fc832187b0132e569d6c3065a3252ed18e56effd0b6a73e524/large-v2.pt",
"large": "https://openaipublic.azureedge.net/main/whisper/models/81f7c96c852ee8fc832187b0132e569d6c3065a3252ed18e56effd0b6a73e524/large-v2.pt", "large-v3": "https://openaipublic.azureedge.net/main/whisper/models/e5b1a55b89c1367dacf97e3e19bfd829a01529dbfdeefa8caeb59b3f1b81dadb/large-v3.pt",
"large": "https://openaipublic.azureedge.net/main/whisper/models/e5b1a55b89c1367dacf97e3e19bfd829a01529dbfdeefa8caeb59b3f1b81dadb/large-v3.pt",
} }
# base85-encoded (n_layers, n_heads) boolean arrays indicating the cross-attention heads that are # base85-encoded (n_layers, n_heads) boolean arrays indicating the cross-attention heads that are
@ -41,7 +42,8 @@ _ALIGNMENT_HEADS = {
"medium": b"ABzY8B0Jh+0{>%R7}kK1fFL7w6%<-Pf*t^=N)Qr&0RR9", "medium": b"ABzY8B0Jh+0{>%R7}kK1fFL7w6%<-Pf*t^=N)Qr&0RR9",
"large-v1": b"ABzY8r9j$a0{>%R7#4sLmoOs{s)o3~84-RPdcFk!JR<kSfC2yj", "large-v1": b"ABzY8r9j$a0{>%R7#4sLmoOs{s)o3~84-RPdcFk!JR<kSfC2yj",
"large-v2": b"ABzY8zd+h!0{>%R7=D0pU<_bnWW*tkYAhobTNnu$jnkEkXqp)j;w1Tzk)UH3X%SZd&fFZ2fC2yj", "large-v2": b"ABzY8zd+h!0{>%R7=D0pU<_bnWW*tkYAhobTNnu$jnkEkXqp)j;w1Tzk)UH3X%SZd&fFZ2fC2yj",
"large": b"ABzY8zd+h!0{>%R7=D0pU<_bnWW*tkYAhobTNnu$jnkEkXqp)j;w1Tzk)UH3X%SZd&fFZ2fC2yj", "large-v3": b'ABzY8gWO1E0{>%R7(9S+Kn!D~%ngiGaR?*L!iJG9p-nab0JQ=-{D1-g00',
"large": b'ABzY8gWO1E0{>%R7(9S+Kn!D~%ngiGaR?*L!iJG9p-nab0JQ=-{D1-g00',
} }

3
whisper/model.py
View File

@ -236,7 +236,8 @@ class Whisper(nn.Module):
self.dims.n_text_head, self.dims.n_text_head,
self.dims.n_text_layer, self.dims.n_text_layer,
) )
# use the last half layers for alignment by default; see `set_alignment_heads()` below # use the last half among the decoder layers for time alignment by default;
# to use a specific set of heads, see `set_alignment_heads()` below.
all_heads = torch.zeros( all_heads = torch.zeros(
self.dims.n_text_layer, self.dims.n_text_head, dtype=torch.bool self.dims.n_text_layer, self.dims.n_text_head, dtype=torch.bool
) )