Phonon-2 card: about 2.1 bits per stored encoder weight
Browse files- README.md +1 -1
- SHA256SUMS.txt +1 -1
README.md
CHANGED
|
@@ -25,7 +25,7 @@ Phonon-2 is the most accurate open speech recognition model for English under 90
|
|
| 25 |
seven English sets it averages 5.21 % word error, and every open model that scores better is at least 5.8 times its size.
|
| 26 |
Set for set it holds the accuracy of its 2.5 GB full-precision teacher, reaching 100.8 % of the teacher's word accuracy on
|
| 27 |
parliamentary speech and beating it on meetings, from a download 15 times smaller. Its encoder holds each weight at one of
|
| 28 |
-
five learned levels in
|
| 29 |
|
| 30 |
It transcribes an hour of audio in about 20 seconds on an M5 MacBook Air (174x realtime), at 143x on eight Zen 5 cores
|
| 31 |
(16 vCPU) and at 6,680x on one H100 in batches of 128.
|
|
|
|
| 25 |
seven English sets it averages 5.21 % word error, and every open model that scores better is at least 5.8 times its size.
|
| 26 |
Set for set it holds the accuracy of its 2.5 GB full-precision teacher, reaching 100.8 % of the teacher's word accuracy on
|
| 27 |
parliamentary speech and beating it on meetings, from a download 15 times smaller. Its encoder holds each weight at one of
|
| 28 |
+
five learned levels in about 2.1 bits.
|
| 29 |
|
| 30 |
It transcribes an hour of audio in about 20 seconds on an M5 MacBook Air (174x realtime), at 143x on eight Zen 5 cores
|
| 31 |
(16 vCPU) and at 6,680x on one H100 in batches of 128.
|
SHA256SUMS.txt
CHANGED
|
@@ -4,6 +4,6 @@
|
|
| 4 |
00624a5043e7ce74029317b024132ca5286116d6fbf3191d226374bc8273789f NOTICE
|
| 5 |
6833410ae75725ec2a9392085c2138a6d4ffb87be62ff5fe86faff45413f8586 reference_transformers.py
|
| 6 |
5cf172e8cf6b313e951c4ef588ca913c7285b0aa3b92de413c3d7ec2b939c0e4 fermion_container.py
|
| 7 |
-
|
| 8 |
9ba9550ad48438d0836ddab3da480b3b69ffa0aac7b7878b5a0039e7ab429411 LICENSE-WEIGHTS-CC-BY-4.0.txt
|
| 9 |
cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30 LICENSE-CODE-Apache-2.0.txt
|
|
|
|
| 4 |
00624a5043e7ce74029317b024132ca5286116d6fbf3191d226374bc8273789f NOTICE
|
| 5 |
6833410ae75725ec2a9392085c2138a6d4ffb87be62ff5fe86faff45413f8586 reference_transformers.py
|
| 6 |
5cf172e8cf6b313e951c4ef588ca913c7285b0aa3b92de413c3d7ec2b939c0e4 fermion_container.py
|
| 7 |
+
0dad36aa73a9b80fd5165dc33bd226f874a4ff727b06c7bd04a289c69da58452 README.md
|
| 8 |
9ba9550ad48438d0836ddab3da480b3b69ffa0aac7b7878b5a0039e7ab429411 LICENSE-WEIGHTS-CC-BY-4.0.txt
|
| 9 |
cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30 LICENSE-CODE-Apache-2.0.txt
|