Hugging Face's logo

OpenASR
/
moonshine-tiny

Automatic Speech Recognition
OpenASR
speech-to-text
oasr
moonshine
Model card Files
xet
 Community
OpenASR commited on
Commit
cffc886
·
verified ·
1 Parent(s): 06c21be

Publish moonshine-tiny OpenASR packs (fp16/q8_0/q4_k)

Browse files
Files changed (3) hide show
  1. README.md +5 -5
  2. moonshine-tiny-fp16.oasr +2 -2
  3. moonshine-tiny-q8_0.oasr +2 -2
README.md CHANGED
@@ -42,7 +42,7 @@ engineered for peak performance on CPU & GPU, **no Python at inference time**.
42
  ```bash
43
  # 1. Install the OpenASR CLI · https://openasr.org
44
  # 2. Pull a build (pick a quant — see the table below)
45
- openasr pull moonshine:tiny-q8
46
 
47
  # 3. Transcribe
48
  openasr transcribe audio.wav --model moonshine-tiny
@@ -51,16 +51,16 @@ openasr transcribe audio.wav --model moonshine-tiny
51
  All builds for this model:
52
 
53
  ```bash
54
- openasr pull moonshine:tiny-fp16
55
- openasr pull moonshine:tiny-q8
56
  ```
57
 
58
  ## 📦 Available builds
59
 
60
  | Quant | File (`.oasr`) | Size | RAM peak | RTF · M1 CPU | RTF · M1 GPU | JFK ΔWER vs fp16 |
61
  |:------|:---------------|-----:|---------:|-------------:|-------------:|-----------------:|
62
- | fp16 | `moonshine-tiny-fp16.oasr` | 109 MB | 322 MB | 0.02× | 0.03× | 0.0% |
63
- | q8_0 | `moonshine-tiny-q8_0.oasr` | 34 MB | 305 MB | 0.02× | 0.03× | 0.0% |
64
 
65
  <sub>RTF = real-time factor on the fixed 11s JFK clip (**lower is faster**); RAM peak measured per pack
66
  in an isolated subprocess. JFK ΔWER compares each quantized build's JFK transcript to this model's
 
42
  ```bash
43
  # 1. Install the OpenASR CLI · https://openasr.org
44
  # 2. Pull a build (pick a quant — see the table below)
45
+ openasr pull moonshine-tiny:q8
46
 
47
  # 3. Transcribe
48
  openasr transcribe audio.wav --model moonshine-tiny
 
51
  All builds for this model:
52
 
53
  ```bash
54
+ openasr pull moonshine-tiny:fp16
55
+ openasr pull moonshine-tiny:q8
56
  ```
57
 
58
  ## 📦 Available builds
59
 
60
  | Quant | File (`.oasr`) | Size | RAM peak | RTF · M1 CPU | RTF · M1 GPU | JFK ΔWER vs fp16 |
61
  |:------|:---------------|-----:|---------:|-------------:|-------------:|-----------------:|
62
+ | fp16 | `moonshine-tiny-fp16.oasr` | 109 MB | 323 MB | 0.04× | 0.03× | 0.0% |
63
+ | q8_0 | `moonshine-tiny-q8_0.oasr` | 34 MB | 306 MB | 0.03× | 0.03× | 0.0% |
64
 
65
  <sub>RTF = real-time factor on the fixed 11s JFK clip (**lower is faster**); RAM peak measured per pack
66
  in an isolated subprocess. JFK ΔWER compares each quantized build's JFK transcript to this model's
moonshine-tiny-fp16.oasr CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:08dc692fcfaec26ce597bb21189406052a68d6c428ea5c38e72a4e73afd3a21d
3
- size 108862976
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2f024ab25f82fd06c6c213df0a2db6ec542e6463acd8bb5e799ed158468d6258
3
+ size 108863040
moonshine-tiny-q8_0.oasr CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:5df858db11aca1aca8d4bb8ba1230e9e683da8292a46137f0993981b89c655a3
3
- size 34360832
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:80ec6caed0a99abb27aa6775c9616c94a397417b6696466dfe59ae43c16f3e77
3
+ size 34360896