Sentence Similarity
sentence-transformers
Safetensors
English
grt
geometric-rotation-transformer
feature-extraction
circular-pooling
rotational
qwen3
Instructions to use MatterWaveAI/grt-30m-embedding with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- sentence-transformers
How to use MatterWaveAI/grt-30m-embedding with sentence-transformers:
from sentence_transformers import SentenceTransformer model = SentenceTransformer("MatterWaveAI/grt-30m-embedding") sentences = [ "That is a happy person", "That is a happy dog", "That is a very happy person", "Today is a sunny day" ] embeddings = model.encode(sentences) similarities = model.similarity(embeddings, embeddings) print(similarities.shape) # [4, 4] - Notebooks
- Google Colab
- Kaggle
Update eval_results.json with 9-way CyclideAdam & Toroidal optimizer ablation
Browse files- eval_results.json +64 -0
eval_results.json
CHANGED
|
@@ -26,6 +26,70 @@
|
|
| 26 |
"stsb_test_spearman_after": 0.6471,
|
| 27 |
"stsb_test_pearson_after": 0.6494
|
| 28 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 29 |
"hessian_pair_adam_1x_coupling": {
|
| 30 |
"stsb_dev_spearman_after": 0.7242,
|
| 31 |
"stsb_dev_pearson_after": 0.6876,
|
|
|
|
| 26 |
"stsb_test_spearman_after": 0.6471,
|
| 27 |
"stsb_test_pearson_after": 0.6494
|
| 28 |
},
|
| 29 |
+
"cyclide_adam": {
|
| 30 |
+
"s1_dev_spearman": 0.6255,
|
| 31 |
+
"stsb_dev_spearman_after": 0.7368,
|
| 32 |
+
"stsb_dev_pearson_after": 0.7179,
|
| 33 |
+
"stsb_test_spearman_after": 0.6287,
|
| 34 |
+
"stsb_test_pearson_after": 0.6476,
|
| 35 |
+
"rocm_elapsed_sec": 380.42
|
| 36 |
+
},
|
| 37 |
+
"hessian_pair_adam_1x_coupling": {
|
| 38 |
+
"stsb_dev_spearman_after": 0.7242,
|
| 39 |
+
"stsb_dev_pearson_after": 0.6876,
|
| 40 |
+
"stsb_test_spearman_after": 0.6337,
|
| 41 |
+
"stsb_test_pearson_after": 0.6369
|
| 42 |
+
},
|
| 43 |
+
"rotor_adam": {
|
| 44 |
+
"stsb_dev_spearman_after": 0.7322,
|
| 45 |
+
"stsb_dev_pearson_after": 0.7099,
|
| 46 |
+
"stsb_test_spearman_after": 0.6329,
|
| 47 |
+
"stsb_test_pearson_after": 0.6309
|
| 48 |
+
},
|
| 49 |
+
"toroidal_lion": {
|
| 50 |
+
"stsb_dev_spearman_after": 0.7289,
|
| 51 |
+
"stsb_dev_pearson_after": 0.697,
|
| 52 |
+
"stsb_test_spearman_after": 0.6319,
|
| 53 |
+
"stsb_test_pearson_after": 0.6341
|
| 54 |
+
},
|
| 55 |
+
"adamw": {
|
| 56 |
+
"stsb_dev_spearman_after": 0.7289,
|
| 57 |
+
"stsb_test_spearman_after": 0.6317,
|
| 58 |
+
"stsb_test_pearson_after": 0.6372
|
| 59 |
+
},
|
| 60 |
+
"hessian_pair_adam_3x_coupling": {
|
| 61 |
+
"stsb_dev_spearman_after": 0.7278,
|
| 62 |
+
"stsb_dev_pearson_after": 0.6917,
|
| 63 |
+
"stsb_test_spearman_after": 0.6288,
|
| 64 |
+
"stsb_test_pearson_after": 0.6316
|
| 65 |
+
},
|
| 66 |
+
"pair_adam_3x_coupling": {
|
| 67 |
+
"stsb_dev_spearman_after": 0.7179,
|
| 68 |
+
"stsb_test_spearman_after": 0.6242,
|
| 69 |
+
"stsb_test_pearson_after": 0.633
|
| 70 |
+
},
|
| 71 |
+
"toroidal_muon": {
|
| 72 |
+
"stsb_dev_spearman_after": 0.7361,
|
| 73 |
+
"stsb_dev_pearson_after": 0.7105,
|
| 74 |
+
"stsb_test_spearman_after": 0.6197,
|
| 75 |
+
"stsb_test_pearson_after": 0.6304
|
| 76 |
+
}
|
| 77 |
+
},
|
| 78 |
+
"optimizer_ablations_9way": {
|
| 79 |
+
"polar_natural_adam": {
|
| 80 |
+
"stsb_dev_spearman_after": 0.7262,
|
| 81 |
+
"stsb_dev_pearson_after": 0.702,
|
| 82 |
+
"stsb_test_spearman_after": 0.6471,
|
| 83 |
+
"stsb_test_pearson_after": 0.6494
|
| 84 |
+
},
|
| 85 |
+
"cyclide_adam": {
|
| 86 |
+
"s1_dev_spearman": 0.6255,
|
| 87 |
+
"stsb_dev_spearman_after": 0.7368,
|
| 88 |
+
"stsb_dev_pearson_after": 0.7179,
|
| 89 |
+
"stsb_test_spearman_after": 0.6287,
|
| 90 |
+
"stsb_test_pearson_after": 0.6476,
|
| 91 |
+
"rocm_elapsed_sec": 380.42
|
| 92 |
+
},
|
| 93 |
"hessian_pair_adam_1x_coupling": {
|
| 94 |
"stsb_dev_spearman_after": 0.7242,
|
| 95 |
"stsb_dev_pearson_after": 0.6876,
|