Update README.md
Browse files
README.md
CHANGED
|
@@ -9,6 +9,8 @@ tags:
|
|
| 9 |
- experiment
|
| 10 |
---
|
| 11 |
|
|
|
|
|
|
|
| 12 |
A first experiment to test and convert clip-vit-base-patch32 into a geometric model by using only a classification head.
|
| 13 |
|
| 14 |
Below is GPT 5's auto-generated dictation based on the notebook. I'll include the full notebook in a moment here.
|
|
@@ -17,6 +19,12 @@ The question was simple; can linear layers learn geometric?
|
|
| 17 |
|
| 18 |
The answer is... maybe. More research required.
|
| 19 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 20 |
# Notebook-6 · Crystal-CLIP CIFAR-100
|
| 21 |
|
| 22 |
One-vector image embeddings (HF CLIP) + pentachora vocabulary anchors → cosine-similarity classifier for CIFAR-100.
|
|
|
|
| 9 |
- experiment
|
| 10 |
---
|
| 11 |
|
| 12 |
+
# Preface
|
| 13 |
+
|
| 14 |
A first experiment to test and convert clip-vit-base-patch32 into a geometric model by using only a classification head.
|
| 15 |
|
| 16 |
Below is GPT 5's auto-generated dictation based on the notebook. I'll include the full notebook in a moment here.
|
|
|
|
| 19 |
|
| 20 |
The answer is... maybe. More research required.
|
| 21 |
|
| 22 |
+
# Reasoning
|
| 23 |
+
|
| 24 |
+
I used the 32 dim; as it seemed to be the weakest with flow-match euler-discreet to test the hypothesis that a small dimensional geometry could in fact be used in substitution of a high-geometric variation.
|
| 25 |
+
|
| 26 |
+
The output model is much larger than I wanted; which defeats the purpose of the overall structure - but it's paired directly at the knee with clip-vit-base-patch32, so I'll prepare a decoupled version here in a bit.
|
| 27 |
+
|
| 28 |
# Notebook-6 · Crystal-CLIP CIFAR-100
|
| 29 |
|
| 30 |
One-vector image embeddings (HF CLIP) + pentachora vocabulary anchors → cosine-similarity classifier for CIFAR-100.
|