Darayut commited on
Commit
ac9e7d9
·
verified ·
1 Parent(s): 2e120f6

Split the architecture figure from the pruning figure

Browse files

The model is now its own figure (assets/architecture.png), with one transformer layer opened up; assets/pruning.png shows only the before-and-after pruning comparison.

Files changed (4) hide show
  1. .gitattributes +1 -0
  2. README.md +5 -1
  3. assets/architecture.png +3 -0
  4. assets/pruning.png +2 -2
.gitattributes CHANGED
@@ -35,3 +35,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
  assets/tradeoff.png filter=lfs diff=lfs merge=lfs -text
37
  assets/pruning.png filter=lfs diff=lfs merge=lfs -text
 
 
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
  assets/tradeoff.png filter=lfs diff=lfs merge=lfs -text
37
  assets/pruning.png filter=lfs diff=lfs merge=lfs -text
38
+ assets/architecture.png filter=lfs diff=lfs merge=lfs -text
README.md CHANGED
@@ -42,7 +42,11 @@ Test sets; greedy CTC decoding, no language model. CER for Khmer, WER for Englis
42
 
43
  ## Architecture
44
 
45
- <p align="center"><img src="https://huggingface.co/Darayut/sota-base/resolve/main/assets/pruning.png" width="100%" alt="Sota's architecture with one transformer layer opened up, beside the fine-tuned omniASR CTC-1B before and sota-large after structured pruning"></p>
 
 
 
 
46
 
47
  *What is pruned*, shown for sota-large: whole attention heads (MHA) and FFN units are removed from every transformer layer, and the CNN front end is kept. Each cell is a quarter of a layer's heads or units.
48
 
 
42
 
43
  ## Architecture
44
 
45
+ <p align="center"><img src="https://huggingface.co/Darayut/sota-base/resolve/main/assets/architecture.png" width="70%" alt="Sota's architecture, reading upwards from 16 kHz audio to the CTC head, with one transformer layer opened up"></p>
46
+
47
+ *The model*, reading upwards, with one transformer layer opened up: 16 kHz audio, a feature extractor of 7 convolutional layers giving one frame every 20 ms, a convolutional position encoder added back to the frames, 48 pre-LayerNorm transformer layers of width 1,280, a final LayerNorm and a CTC head over 9,812 characters. Every attention head and every FFN unit has its own Hard Concrete gate z. The cells in the opened layer are layer 12 of sota-large: 8 of 16 heads and 3,355 of 5,120 FFN units kept; which ones are kept is schematic.
48
+
49
+ <p align="center"><img src="https://huggingface.co/Darayut/sota-base/resolve/main/assets/pruning.png" width="100%" alt="The fine-tuned omniASR CTC-1B before structured pruning and sota-large after: CNN and CTC head kept, attention heads and FFN units removed"></p>
50
 
51
  *What is pruned*, shown for sota-large: whole attention heads (MHA) and FFN units are removed from every transformer layer, and the CNN front end is kept. Each cell is a quarter of a layer's heads or units.
52
 
assets/architecture.png ADDED

Git LFS Details

  • SHA256: ce2104d870326f05bd31dda3fda263fe509b165d01d183121d7306d785117f94
  • Pointer size: 131 Bytes
  • Size of remote file: 228 kB
assets/pruning.png CHANGED

Git LFS Details

  • SHA256: df965173ab43e8552775f8c48931d7e44d2112b22cb77d7b5f2596f9683ced61
  • Pointer size: 131 Bytes
  • Size of remote file: 229 kB

Git LFS Details

  • SHA256: ec7f903754cdc10d127172f0781f1b9bba3ff67de914c621b59dd2d112a328ca
  • Pointer size: 130 Bytes
  • Size of remote file: 98 kB