Sentence Similarity
PyTorch
English
taylor-m1
constructai commited on
Commit
5156aa7
·
verified ·
1 Parent(s): 1da9674

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +12 -0
README.md CHANGED
@@ -12,11 +12,15 @@ tags:
12
 
13
  # constructai/taylor-m1-38m-128
14
 
 
 
15
  This model is a fine-tuned version of a custom BERT-like encoder (hidden_size=384, 6 layers) trained on MS MARCO passage ranking dataset.
16
  It uses **triplet hard negatives** from [`sentence-transformers/msmarco-msmarco-distilbert-base-v3`](https://huggingface.co/datasets/sentence-transformers/msmarco-msmarco-distilbert-base-v3) (Apache-2.0 license). The base MS MARCO data is subject to [Microsoft Research License](https://github.com/microsoft/MSMARCO-Passage-Ranking/blob/master/LICENSE).
17
 
18
  The model produces **384-dimensional** embeddings (CLS token) optimized for cosine similarity.
19
 
 
 
20
  # Training details
21
 
22
  * ~38.3M parameters
@@ -31,6 +35,8 @@ The model produces **384-dimensional** embeddings (CLS token) optimized for cosi
31
 
32
  * Data: ~500k triplets from MS MARCO
33
 
 
 
34
  # Evaluation
35
 
36
  On a small test set, the model achieves:
@@ -41,6 +47,8 @@ On a small test set, the model achieves:
41
 
42
  * Margin: 0.44
43
 
 
 
44
  # Usage
45
 
46
  **Option 1: Via custom Python package (recommended)**
@@ -91,6 +99,8 @@ print("Similarity matrix:")
91
  print(similarities)
92
  ```
93
 
 
 
94
  **Requirements:**
95
 
96
  * Python ≥ 3.9
@@ -99,6 +109,8 @@ print(similarities)
99
  * huggingface_hub ≥ 0.20.0
100
  * numpy
101
 
 
 
102
  # License
103
 
104
  This model is released under the **Apache 2.0** License.
 
12
 
13
  # constructai/taylor-m1-38m-128
14
 
15
+ ---
16
+
17
  This model is a fine-tuned version of a custom BERT-like encoder (hidden_size=384, 6 layers) trained on MS MARCO passage ranking dataset.
18
  It uses **triplet hard negatives** from [`sentence-transformers/msmarco-msmarco-distilbert-base-v3`](https://huggingface.co/datasets/sentence-transformers/msmarco-msmarco-distilbert-base-v3) (Apache-2.0 license). The base MS MARCO data is subject to [Microsoft Research License](https://github.com/microsoft/MSMARCO-Passage-Ranking/blob/master/LICENSE).
19
 
20
  The model produces **384-dimensional** embeddings (CLS token) optimized for cosine similarity.
21
 
22
+ ---
23
+
24
  # Training details
25
 
26
  * ~38.3M parameters
 
35
 
36
  * Data: ~500k triplets from MS MARCO
37
 
38
+ ---
39
+
40
  # Evaluation
41
 
42
  On a small test set, the model achieves:
 
47
 
48
  * Margin: 0.44
49
 
50
+ ---
51
+
52
  # Usage
53
 
54
  **Option 1: Via custom Python package (recommended)**
 
99
  print(similarities)
100
  ```
101
 
102
+ ---
103
+
104
  **Requirements:**
105
 
106
  * Python ≥ 3.9
 
109
  * huggingface_hub ≥ 0.20.0
110
  * numpy
111
 
112
+ ---
113
+
114
  # License
115
 
116
  This model is released under the **Apache 2.0** License.