Push model using huggingface_hub.
Browse files- README.md +74 -90
- config_setfit.json +2 -2
- model.safetensors +1 -1
- model_head.pkl +2 -2
README.md
CHANGED
|
@@ -10,12 +10,15 @@ tags:
|
|
| 10 |
- text-classification
|
| 11 |
- generated_from_setfit_trainer
|
| 12 |
widget:
|
| 13 |
-
- text:
|
| 14 |
-
|
| 15 |
-
- text:
|
| 16 |
-
|
| 17 |
-
- text: ل
|
| 18 |
-
- text:
|
|
|
|
|
|
|
|
|
|
| 19 |
inference: true
|
| 20 |
model-index:
|
| 21 |
- name: SetFit with akhooli/sbert_ar_nli_500k_norm
|
|
@@ -29,39 +32,14 @@ model-index:
|
|
| 29 |
split: test
|
| 30 |
metrics:
|
| 31 |
- type: accuracy
|
| 32 |
-
value: 0.
|
| 33 |
name: Accuracy
|
| 34 |
---
|
| 35 |
|
| 36 |
# SetFit with akhooli/sbert_ar_nli_500k_norm
|
| 37 |
|
| 38 |
-
This is a [SetFit](https://github.com/huggingface/setfit) model that can be used for Text Classification.
|
| 39 |
-
This SetFit model uses [akhooli/sbert_ar_nli_500k_norm](https://huggingface.co/akhooli/sbert_ar_nli_500k_norm) as the Sentence Transformer embedding model.
|
| 40 |
-
A [LogisticRegression](https://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LogisticRegression.html) instance is used for classification.
|
| 41 |
-
This model is trained with few shots using the [akhooli/ar_hs](https://huggingface.co/datasets/akhooli/ar_hs) dataset. The dataset uses LLM to generate labels.
|
| 42 |
-
Usage:
|
| 43 |
-
```python
|
| 44 |
-
pip install setfit
|
| 45 |
-
from setfit import SetFitModel
|
| 46 |
-
from unicodedata import normalize
|
| 47 |
|
| 48 |
-
# Download model from Hub
|
| 49 |
-
model = SetFitModel.from_pretrained("akhooli/setfit_ar_hs")
|
| 50 |
-
# Run inference
|
| 51 |
-
queries = [
|
| 52 |
-
"سكت دهراً و نطق كفراً",
|
| 53 |
-
"الخلاف ﻻ يفسد للود قضية.",
|
| 54 |
-
"أنت شخص منبوذ. احترم أسيادك.",
|
| 55 |
-
"دع المكارم ﻻ ترحل لبغيتها واقعد فإنك أنت الطاعم الكاسي",
|
| 56 |
-
]
|
| 57 |
-
queries_n = [normalize('NFKC', query) for query in queries]
|
| 58 |
-
preds = model.predict(queries_n)
|
| 59 |
-
print(preds)
|
| 60 |
-
# if you want to see the probabilities for each label
|
| 61 |
-
probas = model.predict_proba(queries_n)
|
| 62 |
-
print(probas)
|
| 63 |
-
```
|
| 64 |
-
The rest of this card is auto generated.
|
| 65 |
The model has been trained using an efficient few-shot learning technique that involves:
|
| 66 |
|
| 67 |
1. Fine-tuning a [Sentence Transformer](https://www.sbert.net) with contrastive learning.
|
|
@@ -74,7 +52,7 @@ The model has been trained using an efficient few-shot learning technique that i
|
|
| 74 |
- **Sentence Transformer body:** [akhooli/sbert_ar_nli_500k_norm](https://huggingface.co/akhooli/sbert_ar_nli_500k_norm)
|
| 75 |
- **Classification head:** a [LogisticRegression](https://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LogisticRegression.html) instance
|
| 76 |
- **Maximum Sequence Length:** 512 tokens
|
| 77 |
-
- **Number of Classes:**
|
| 78 |
<!-- - **Training Dataset:** [Unknown](https://huggingface.co/datasets/unknown) -->
|
| 79 |
<!-- - **Language:** Unknown -->
|
| 80 |
<!-- - **License:** Unknown -->
|
|
@@ -85,12 +63,18 @@ The model has been trained using an efficient few-shot learning technique that i
|
|
| 85 |
- **Paper:** [Efficient Few-Shot Learning Without Prompts](https://arxiv.org/abs/2209.11055)
|
| 86 |
- **Blogpost:** [SetFit: Efficient Few-Shot Learning Without Prompts](https://huggingface.co/blog/setfit)
|
| 87 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 88 |
## Evaluation
|
| 89 |
|
| 90 |
### Metrics
|
| 91 |
| Label | Accuracy |
|
| 92 |
|:--------|:---------|
|
| 93 |
-
| **all** | 0.
|
| 94 |
|
| 95 |
## Uses
|
| 96 |
|
|
@@ -110,7 +94,7 @@ from setfit import SetFitModel
|
|
| 110 |
# Download from the 🤗 Hub
|
| 111 |
model = SetFitModel.from_pretrained("akhooli/setfit_ar_hs")
|
| 112 |
# Run inference
|
| 113 |
-
preds = model("وكل مي
|
| 114 |
```
|
| 115 |
|
| 116 |
<!--
|
|
@@ -140,9 +124,9 @@ preds = model("وكل مين بدو يشد على مشدو ")
|
|
| 140 |
## Training Details
|
| 141 |
|
| 142 |
### Training Set Metrics
|
| 143 |
-
| Training set | Min | Median
|
| 144 |
-
|:-------------|:----|:-------
|
| 145 |
-
| Word count | 1 | 12.
|
| 146 |
|
| 147 |
| Label | Training Sample Count |
|
| 148 |
|:---------|:----------------------|
|
|
@@ -164,64 +148,64 @@ preds = model("وكل مين بدو يشد على مشدو ")
|
|
| 164 |
- warmup_proportion: 0.1
|
| 165 |
- l2_weight: 0.01
|
| 166 |
- seed: 42
|
| 167 |
-
- run_name:
|
| 168 |
- eval_max_steps: -1
|
| 169 |
- load_best_model_at_end: False
|
| 170 |
|
| 171 |
### Training Results
|
| 172 |
| Epoch | Step | Training Loss | Validation Loss |
|
| 173 |
|:------:|:----:|:-------------:|:---------------:|
|
| 174 |
-
| 0.0004 | 1 | 0.
|
| 175 |
-
| 0.04 | 100 | 0.
|
| 176 |
-
| 0.08 | 200 | 0.
|
| 177 |
-
| 0.12 | 300 | 0.
|
| 178 |
-
| 0.16 | 400 | 0.
|
| 179 |
-
| 0.2 | 500 | 0.
|
| 180 |
-
| 0.24 | 600 | 0.
|
| 181 |
-
| 0.28 | 700 | 0.
|
| 182 |
-
| 0.32 | 800 | 0.
|
| 183 |
-
| 0.36 | 900 | 0.
|
| 184 |
-
| 0.4 | 1000 | 0.
|
| 185 |
-
| 0.44 | 1100 | 0.
|
| 186 |
-
| 0.48 | 1200 | 0.
|
| 187 |
-
| 0.52 | 1300 | 0.
|
| 188 |
-
| 0.56 | 1400 | 0.
|
| 189 |
-
| 0.6 | 1500 | 0.
|
| 190 |
-
| 0.64 | 1600 | 0.
|
| 191 |
-
| 0.68 | 1700 | 0.
|
| 192 |
-
| 0.72 | 1800 | 0.
|
| 193 |
-
| 0.76 | 1900 | 0.
|
| 194 |
-
| 0.8 | 2000 | 0.
|
| 195 |
-
| 0.84 | 2100 | 0.
|
| 196 |
-
| 0.88 | 2200 | 0.
|
| 197 |
-
| 0.92 | 2300 | 0.
|
| 198 |
-
| 0.96 | 2400 | 0.
|
| 199 |
-
| 1.0 | 2500 | 0.
|
| 200 |
-
| 1.04 | 2600 | 0.
|
| 201 |
-
| 1.08 | 2700 | 0.
|
| 202 |
-
| 1.12 | 2800 | 0.
|
| 203 |
-
| 1.16 | 2900 | 0.
|
| 204 |
| 1.2 | 3000 | 0.0039 | - |
|
| 205 |
-
| 1.24 | 3100 | 0.
|
| 206 |
-
| 1.28 | 3200 | 0.
|
| 207 |
-
| 1.32 | 3300 | 0.
|
| 208 |
-
| 1.3600 | 3400 | 0.
|
| 209 |
-
| 1.4 | 3500 | 0.
|
| 210 |
-
| 1.44 | 3600 | 0.
|
| 211 |
-
| 1.48 | 3700 | 0.
|
| 212 |
-
| 1.52 | 3800 | 0.
|
| 213 |
-
| 1.56 | 3900 | 0.
|
| 214 |
-
| 1.6 | 4000 | 0.
|
| 215 |
-
| 1.6400 | 4100 | 0.
|
| 216 |
-
| 1.6800 | 4200 | 0.
|
| 217 |
-
| 1.72 | 4300 | 0.
|
| 218 |
-
| 1.76 | 4400 | 0.
|
| 219 |
-
| 1.8 | 4500 | 0.
|
| 220 |
-
| 1.8400 | 4600 | 0.
|
| 221 |
-
| 1.88 | 4700 | 0.
|
| 222 |
-
| 1.92 | 4800 | 0.
|
| 223 |
-
| 1.96 | 4900 | 0.
|
| 224 |
-
| 2.0 | 5000 | 0.
|
| 225 |
|
| 226 |
### Framework Versions
|
| 227 |
- Python: 3.10.14
|
|
|
|
| 10 |
- text-classification
|
| 11 |
- generated_from_setfit_trainer
|
| 12 |
widget:
|
| 13 |
+
- text: عزيزي جبران باسيل بدك تعرف كتييير منيح انو مش شغلتنا نحفظ امن اسرائيل يلي
|
| 14 |
+
ما منعترف ولن نعترف ب وجودها ابدا
|
| 15 |
+
- text: 'يجب على هؤلاك المجرمون الارهابيون وكل من دس فتنة انا يتحاسبو حساب مؤلم لكن
|
| 16 |
+
سؤال من سيحاسبهن '
|
| 17 |
+
- text: شيل عينك عن لبنان انت و كل كلب متلك حكايتك و غير هيك انشالله بتنباع بالعزى
|
| 18 |
+
- text: لسه بصرعوا طيزنا بدكن نصير متل العراق وليبيا يا حمير تجاوزناهن بأشواط، هلق
|
| 19 |
+
لو نصير متل العراق وليبيا تحسن كبير جدا
|
| 20 |
+
- text: كول هوا خسرتو بأرضك وبين جمهورك بعد ما منعت القطريين من تشجيع جمهورهم انتو
|
| 21 |
+
فاشلين في كل شئ وهم متفوقين عليكم في...
|
| 22 |
inference: true
|
| 23 |
model-index:
|
| 24 |
- name: SetFit with akhooli/sbert_ar_nli_500k_norm
|
|
|
|
| 32 |
split: test
|
| 33 |
metrics:
|
| 34 |
- type: accuracy
|
| 35 |
+
value: 0.8452520515826495
|
| 36 |
name: Accuracy
|
| 37 |
---
|
| 38 |
|
| 39 |
# SetFit with akhooli/sbert_ar_nli_500k_norm
|
| 40 |
|
| 41 |
+
This is a [SetFit](https://github.com/huggingface/setfit) model that can be used for Text Classification. This SetFit model uses [akhooli/sbert_ar_nli_500k_norm](https://huggingface.co/akhooli/sbert_ar_nli_500k_norm) as the Sentence Transformer embedding model. A [LogisticRegression](https://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LogisticRegression.html) instance is used for classification.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 42 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 43 |
The model has been trained using an efficient few-shot learning technique that involves:
|
| 44 |
|
| 45 |
1. Fine-tuning a [Sentence Transformer](https://www.sbert.net) with contrastive learning.
|
|
|
|
| 52 |
- **Sentence Transformer body:** [akhooli/sbert_ar_nli_500k_norm](https://huggingface.co/akhooli/sbert_ar_nli_500k_norm)
|
| 53 |
- **Classification head:** a [LogisticRegression](https://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LogisticRegression.html) instance
|
| 54 |
- **Maximum Sequence Length:** 512 tokens
|
| 55 |
+
- **Number of Classes:** 2 classes
|
| 56 |
<!-- - **Training Dataset:** [Unknown](https://huggingface.co/datasets/unknown) -->
|
| 57 |
<!-- - **Language:** Unknown -->
|
| 58 |
<!-- - **License:** Unknown -->
|
|
|
|
| 63 |
- **Paper:** [Efficient Few-Shot Learning Without Prompts](https://arxiv.org/abs/2209.11055)
|
| 64 |
- **Blogpost:** [SetFit: Efficient Few-Shot Learning Without Prompts](https://huggingface.co/blog/setfit)
|
| 65 |
|
| 66 |
+
### Model Labels
|
| 67 |
+
| Label | Examples |
|
| 68 |
+
|:---------|:---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
|
| 69 |
+
| negative | <ul><li>'يا ريت بيمنعوا الأرغيلة بلبنان، لأن غير هيك ما منعمل ثورة '</li><li>'أصلا جبران عندو طيارة وعندو قصر بأوروبا ومحيط الهادىء الى اسهم فيه وتم اكتشاف كوكب جديد مثل زحل وجوبيتير تم شرائه ك...'</li><li>'اكره البرازيل بس لا تقوليلي خلاص كلشي انتهى بليز'</li></ul> |
|
| 70 |
+
| positive | <ul><li>'السيد والرئيس وليش عم تشددددد دخلك كل حجمك أرنب عند معلمك بالقرداحة'</li><li>'العوني اذا تمدن متل الجحش اذا تكدن بعمرك شفت عوني بيفهم'</li><li>'لا بس الوطن بدو تكنيس من ل متلك '</li></ul> |
|
| 71 |
+
|
| 72 |
## Evaluation
|
| 73 |
|
| 74 |
### Metrics
|
| 75 |
| Label | Accuracy |
|
| 76 |
|:--------|:---------|
|
| 77 |
+
| **all** | 0.8453 |
|
| 78 |
|
| 79 |
## Uses
|
| 80 |
|
|
|
|
| 94 |
# Download from the 🤗 Hub
|
| 95 |
model = SetFitModel.from_pretrained("akhooli/setfit_ar_hs")
|
| 96 |
# Run inference
|
| 97 |
+
preds = model("شيل عينك عن لبنان انت و كل كلب متلك حكايتك و غير هيك انشالله بتنباع بالعزى")
|
| 98 |
```
|
| 99 |
|
| 100 |
<!--
|
|
|
|
| 124 |
## Training Details
|
| 125 |
|
| 126 |
### Training Set Metrics
|
| 127 |
+
| Training set | Min | Median | Max |
|
| 128 |
+
|:-------------|:----|:-------|:----|
|
| 129 |
+
| Word count | 1 | 12.809 | 52 |
|
| 130 |
|
| 131 |
| Label | Training Sample Count |
|
| 132 |
|:---------|:----------------------|
|
|
|
|
| 148 |
- warmup_proportion: 0.1
|
| 149 |
- l2_weight: 0.01
|
| 150 |
- seed: 42
|
| 151 |
+
- run_name: setfit_hate_2kv
|
| 152 |
- eval_max_steps: -1
|
| 153 |
- load_best_model_at_end: False
|
| 154 |
|
| 155 |
### Training Results
|
| 156 |
| Epoch | Step | Training Loss | Validation Loss |
|
| 157 |
|:------:|:----:|:-------------:|:---------------:|
|
| 158 |
+
| 0.0004 | 1 | 0.3239 | - |
|
| 159 |
+
| 0.04 | 100 | 0.277 | - |
|
| 160 |
+
| 0.08 | 200 | 0.2406 | - |
|
| 161 |
+
| 0.12 | 300 | 0.1737 | - |
|
| 162 |
+
| 0.16 | 400 | 0.1259 | - |
|
| 163 |
+
| 0.2 | 500 | 0.0701 | - |
|
| 164 |
+
| 0.24 | 600 | 0.0473 | - |
|
| 165 |
+
| 0.28 | 700 | 0.0298 | - |
|
| 166 |
+
| 0.32 | 800 | 0.0239 | - |
|
| 167 |
+
| 0.36 | 900 | 0.02 | - |
|
| 168 |
+
| 0.4 | 1000 | 0.0151 | - |
|
| 169 |
+
| 0.44 | 1100 | 0.0143 | - |
|
| 170 |
+
| 0.48 | 1200 | 0.0126 | - |
|
| 171 |
+
| 0.52 | 1300 | 0.0121 | - |
|
| 172 |
+
| 0.56 | 1400 | 0.0078 | - |
|
| 173 |
+
| 0.6 | 1500 | 0.0111 | - |
|
| 174 |
+
| 0.64 | 1600 | 0.0099 | - |
|
| 175 |
+
| 0.68 | 1700 | 0.0091 | - |
|
| 176 |
+
| 0.72 | 1800 | 0.0064 | - |
|
| 177 |
+
| 0.76 | 1900 | 0.0101 | - |
|
| 178 |
+
| 0.8 | 2000 | 0.0073 | - |
|
| 179 |
+
| 0.84 | 2100 | 0.0042 | - |
|
| 180 |
+
| 0.88 | 2200 | 0.0038 | - |
|
| 181 |
+
| 0.92 | 2300 | 0.0058 | - |
|
| 182 |
+
| 0.96 | 2400 | 0.0041 | - |
|
| 183 |
+
| 1.0 | 2500 | 0.0026 | - |
|
| 184 |
+
| 1.04 | 2600 | 0.0037 | - |
|
| 185 |
+
| 1.08 | 2700 | 0.0035 | - |
|
| 186 |
+
| 1.12 | 2800 | 0.0045 | - |
|
| 187 |
+
| 1.16 | 2900 | 0.0038 | - |
|
| 188 |
| 1.2 | 3000 | 0.0039 | - |
|
| 189 |
+
| 1.24 | 3100 | 0.0018 | - |
|
| 190 |
+
| 1.28 | 3200 | 0.003 | - |
|
| 191 |
+
| 1.32 | 3300 | 0.0028 | - |
|
| 192 |
+
| 1.3600 | 3400 | 0.0023 | - |
|
| 193 |
+
| 1.4 | 3500 | 0.0022 | - |
|
| 194 |
+
| 1.44 | 3600 | 0.0032 | - |
|
| 195 |
+
| 1.48 | 3700 | 0.0028 | - |
|
| 196 |
+
| 1.52 | 3800 | 0.0022 | - |
|
| 197 |
+
| 1.56 | 3900 | 0.0024 | - |
|
| 198 |
+
| 1.6 | 4000 | 0.0021 | - |
|
| 199 |
+
| 1.6400 | 4100 | 0.0032 | - |
|
| 200 |
+
| 1.6800 | 4200 | 0.0026 | - |
|
| 201 |
+
| 1.72 | 4300 | 0.0025 | - |
|
| 202 |
+
| 1.76 | 4400 | 0.003 | - |
|
| 203 |
+
| 1.8 | 4500 | 0.0028 | - |
|
| 204 |
+
| 1.8400 | 4600 | 0.003 | - |
|
| 205 |
+
| 1.88 | 4700 | 0.0028 | - |
|
| 206 |
+
| 1.92 | 4800 | 0.0033 | - |
|
| 207 |
+
| 1.96 | 4900 | 0.0019 | - |
|
| 208 |
+
| 2.0 | 5000 | 0.0023 | - |
|
| 209 |
|
| 210 |
### Framework Versions
|
| 211 |
- Python: 3.10.14
|
config_setfit.json
CHANGED
|
@@ -1,7 +1,7 @@
|
|
| 1 |
{
|
| 2 |
-
"normalize_embeddings": false,
|
| 3 |
"labels": [
|
| 4 |
"negative",
|
| 5 |
"positive"
|
| 6 |
-
]
|
|
|
|
| 7 |
}
|
|
|
|
| 1 |
{
|
|
|
|
| 2 |
"labels": [
|
| 3 |
"negative",
|
| 4 |
"positive"
|
| 5 |
+
],
|
| 6 |
+
"normalize_embeddings": false
|
| 7 |
}
|
model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 540795752
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:aa207876d4a89ac428c7260c57c75272051dfb17bbf88ee51b56bc87c54f9a67
|
| 3 |
size 540795752
|
model_head.pkl
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:49f3e09533da336510f66c9419d4d76468ed0ad3e8378107f08645838e801645
|
| 3 |
+
size 7007
|