AbstractPhil
/

geolip-deep-embedding-analysis

Model card Files Files and versions

AbstractPhil commited on Apr 5

Commit

1014585

·

verified ·

1 Parent(s): 9c5e7b8

Update README.md

Files changed (1) hide show

README.md +1 -1

README.md CHANGED Viewed

@@ -6,7 +6,7 @@ This is a repo specifically dedicated to analyzing embedding spaces for perfect
 We're going to get to the bottom of why embeddings are so symmetrically sample-sturdy and can be assessed with CV deterministically on random,
 and with that compare the deconstructive nature of collapsing embeddings into more unified compacted spaces for MHA heads.
-The bands show; D16 and D38 are directly CV sample-capable for volume validity, 64 dim being the utmost upper bounds for distance before
 full degredation and breakdown using LINALG.det
 Generally the rule of thumb for MHA head counts is 64 dims per head and scale from there. I'm thinking there may be a direct causal response

 We're going to get to the bottom of why embeddings are so symmetrically sample-sturdy and can be assessed with CV deterministically on random,
 and with that compare the deconstructive nature of collapsing embeddings into more unified compacted spaces for MHA heads.
+The bands show; D16 and D36 are directly CV sample-capable for volume validity, 64 dim being the utmost upper bounds for distance before
 full degredation and breakdown using LINALG.det
 Generally the rule of thumb for MHA head counts is 64 dims per head and scale from there. I'm thinking there may be a direct causal response