Instructions to use WariHima/DanMatsu-KilosizeJapaneseTextEmbedModel with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use WariHima/DanMatsu-KilosizeJapaneseTextEmbedModel with Transformers:
# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("WariHima/DanMatsu-KilosizeJapaneseTextEmbedModel", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Update README.md
Browse files
README.md
CHANGED
|
@@ -8,4 +8,92 @@ library_name: transformers
|
|
| 8 |
tags:
|
| 9 |
- rwkv7
|
| 10 |
- text-embed
|
| 11 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 8 |
tags:
|
| 9 |
- rwkv7
|
| 10 |
- text-embed
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
## 断松 kilo-size-japanese-text-embedding-models
|
| 14 |
+
|
| 15 |
+
### 20k model
|
| 16 |
+
#### architecture
|
| 17 |
+
rwkv7
|
| 18 |
+
hidden size 16dim
|
| 19 |
+
2 layers
|
| 20 |
+
chr + bpe hybrid tokenizer (vocab size 512)
|
| 21 |
+
#### scores
|
| 22 |
+
##### JSTS
|
| 23 |
+
"pearson": 0.589115,
|
| 24 |
+
"spearman": 0.579923,
|
| 25 |
+
"cosine_pearson": 0.589115,
|
| 26 |
+
"cosine_spearman": 0.579923,
|
| 27 |
+
"manhattan_pearson": 0.601691,
|
| 28 |
+
"manhattan_spearman": 0.571704,
|
| 29 |
+
"euclidean_pearson": 0.60794,
|
| 30 |
+
"euclidean_spearman": 0.578109,
|
| 31 |
+
"main_score": 0.579923,
|
| 32 |
+
##### JaQuADRetrieval
|
| 33 |
+
"hit_rate_at_1": 0.07422,
|
| 34 |
+
"hit_rate_at_3": 0.14893,
|
| 35 |
+
"hit_rate_at_5": 0.2002,
|
| 36 |
+
"hit_rate_at_10": 0.27441,
|
| 37 |
+
"hit_rate_at_20": 0.3877,
|
| 38 |
+
"hit_rate_at_100": 0.67139,
|
| 39 |
+
"hit_rate_at_1000": 0.979,
|
| 40 |
+
"main_score": 0.10268,
|
| 41 |
+
|
| 42 |
+
---
|
| 43 |
+
|
| 44 |
+
### 40k model
|
| 45 |
+
#### architecture
|
| 46 |
+
rwkv7
|
| 47 |
+
hidden size 16dim
|
| 48 |
+
2 layers
|
| 49 |
+
chr + bpe hybrid tokenizer (vocab size 512)
|
| 50 |
+
#### scores
|
| 51 |
+
##### JSTS
|
| 52 |
+
"pearson": 0.622386,
|
| 53 |
+
"spearman": 0.601339,
|
| 54 |
+
"cosine_pearson": 0.622386,
|
| 55 |
+
"cosine_spearman": 0.601339,
|
| 56 |
+
"manhattan_pearson": 0.626993,
|
| 57 |
+
"manhattan_spearman": 0.594981,
|
| 58 |
+
"euclidean_pearson": 0.634518,
|
| 59 |
+
"euclidean_spearman": 0.601892,
|
| 60 |
+
"main_score": 0.601339,
|
| 61 |
+
##### JaQuADRetrieval
|
| 62 |
+
"hit_rate_at_1": 0.12988,
|
| 63 |
+
"hit_rate_at_3": 0.23926,
|
| 64 |
+
"hit_rate_at_5": 0.30713,
|
| 65 |
+
"hit_rate_at_10": 0.41113,
|
| 66 |
+
"hit_rate_at_20": 0.51318,
|
| 67 |
+
"hit_rate_at_100": 0.74707,
|
| 68 |
+
"hit_rate_at_1000": 0.98096,
|
| 69 |
+
"main_score": 0.16319,
|
| 70 |
+
|
| 71 |
+
---
|
| 72 |
+
|
| 73 |
+
### 60k model
|
| 74 |
+
#### architecture
|
| 75 |
+
rwkv7
|
| 76 |
+
hidden size 32dim
|
| 77 |
+
2 layers
|
| 78 |
+
chr + bpe hybrid tokenizer (vocab size 512)
|
| 79 |
+
#### scores
|
| 80 |
+
##### JSTS
|
| 81 |
+
"pearson": 0.685757,
|
| 82 |
+
"spearman": 0.656014,
|
| 83 |
+
"cosine_pearson": 0.685757,
|
| 84 |
+
"cosine_spearman": 0.656014,
|
| 85 |
+
"manhattan_pearson": 0.679063,
|
| 86 |
+
"manhattan_spearman": 0.652057,
|
| 87 |
+
"euclidean_pearson": 0.682095,
|
| 88 |
+
"euclidean_spearman": 0.655796,
|
| 89 |
+
"main_score": 0.656014,
|
| 90 |
+
##### JaQuADRetrieval
|
| 91 |
+
"hit_rate_at_1": 0.1792,
|
| 92 |
+
"hit_rate_at_3": 0.31494,
|
| 93 |
+
"hit_rate_at_5": 0.39404,
|
| 94 |
+
"hit_rate_at_10": 0.49805,
|
| 95 |
+
"hit_rate_at_20": 0.60254,
|
| 96 |
+
"hit_rate_at_100": 0.79932,
|
| 97 |
+
"hit_rate_at_1000": 0.98682,
|
| 98 |
+
"main_score": 0.20629,
|
| 99 |
+
|