ele-sage/mdeberta-v3-base-name-classifier-v2

README.md

@@ -4,9 +4,6 @@ license: mit
 base_model: microsoft/mdeberta-v3-base
 tags:
 - generated_from_trainer
-- name
-- person
-- company
 metrics:
 - accuracy
 - precision
@@ -15,64 +12,32 @@ metrics:
 model-index:
 - name: mdeberta-v3-base-name-classifier-v2
   results: []
-datasets:
-- ele-sage/person-company-names-classification
-language:
-- fr
-- en
 ---
 
+<!-- This model card has been generated automatically according to the information the Trainer had access to. You
+should probably proofread and complete it, then remove this comment. -->
 
 # mdeberta-v3-base-name-classifier-v2
 
-This model is a fine-tuned version of [microsoft/mdeberta-v3-base](https://huggingface.co/microsoft/mdeberta-v3-base) on [ele-sage/person-company-names-classification](https://huggingface.co/ele-sage/person-company-names-classification).
-
-
+This model is a fine-tuned version of [microsoft/mdeberta-v3-base](https://huggingface.co/microsoft/mdeberta-v3-base) on an unknown dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.0215
-- Accuracy: 0.9943
-- Precision: 0.9984
-- Recall: 0.9913
-- F1: 0.9949
+- Loss: 0.1317
+- Accuracy: 0.9946
+- Precision: 0.9989
+- Recall: 0.9914
+- F1: 0.9951
 
 ## Model description
 
-This model is a high-performance binary text classifier, fine-tuned from `mdeberta-v3-base`.
-Its purpose is to distinguish between a **person's name** and a **company/organization name** with high accuracy.
-
-### Direct Use
-
-This model is intended to be used for text classification. Given a string, it will return a label indicating whether the string is a `Person` or a `Company`.
-
-```python
-from transformers import pipeline
-
-classifier = pipeline("text-classification", model="ele-sage/mdeberta-v3-base-name-classifier-v2")
-
-results = classifier([
-    "Satya Nadella",
-    "Global Innovations Inc.",
-    "Martinez, Alonso"
-])
-
-for result in results:
-    print(f"Text: '{result['text']}', Prediction: {result['label']}, Score: {result['score']:.4f}")
-```
-
-### Downstream Use
-
-This model is a key component of a two-stage name processing pipeline. It is designed to be used as a fast, efficient "gatekeeper" to first identify person names before passing them to a more complex parsing model, such as `ele-sage/distilbert-base-uncased-name-splitter`.
-
-### Out-of-Scope Use
+More information needed
 
-- This model is not a general-purpose classifier. It is highly specialized for distinguishing persons from companies and will not perform well on other classification tasks (e.g., sentiment analysis).
+## Intended uses & limitations
 
-## Bias, Risks, and Limitations
+More information needed
 
-- **Geographic & Cultural Bias:** The training data is heavily biased towards North American (Canadian) person names and Quebec-based company names. The model will be less accurate when classifying names from other cultural or geographic origins.
-- **Ambiguity:** Certain names can legitimately be both a person's name and a company's name (e.g., "Ford"). In these cases, the model makes a statistical guess based on its training data, which may not always align with the specific context.
-- **Data Source:** The person name data is derived from a Facebook data leak and contains noise. While a rigorous cleaning process was applied, the model may have learned from some spurious data.
+## Training and evaluation data
 
+More information needed
 
 ## Training procedure
 
@@ -81,44 +46,31 @@ This model is a key component of a two-stage name processing pipeline. It is des
 The following hyperparameters were used during training:
 - learning_rate: 1e-05
 - train_batch_size: 128
-- eval_batch_size: 128
+- eval_batch_size: 512
 - seed: 42
 - optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
-- lr_scheduler_type: linear
-- lr_scheduler_warmup_steps: 2000
+- lr_scheduler_type: cosine
+- lr_scheduler_warmup_ratio: 0.05
 - num_epochs: 1
+- label_smoothing_factor: 0.05
 
 ### Training results
 
 | Training Loss | Epoch  | Step  | Validation Loss | Accuracy | Precision | Recall | F1     |
 |:-------------:|:------:|:-----:|:---------------:|:--------:|:---------:|:------:|:------:|
-| 0.0459        | 0.0359 | 2000  | 0.0425          | 0.9894   | 0.9981    | 0.9830 | 0.9905 |
-| 0.032         | 0.0718 | 4000  | 0.0343          | 0.9919   | 0.9960    | 0.9895 | 0.9927 |
-| 0.0322        | 0.1076 | 6000  | 0.0302          | 0.9926   | 0.9975    | 0.9893 | 0.9933 |
-| 0.0347        | 0.1435 | 8000  | 0.0262          | 0.9928   | 0.9968    | 0.9903 | 0.9936 |
-| 0.0292        | 0.1794 | 10000 | 0.0260          | 0.9931   | 0.9973    | 0.9903 | 0.9938 |
-| 0.029         | 0.2153 | 12000 | 0.0251          | 0.9933   | 0.9974    | 0.9905 | 0.9940 |
-| 0.0243        | 0.2511 | 14000 | 0.0254          | 0.9933   | 0.9975    | 0.9905 | 0.9940 |
-| 0.0309        | 0.2870 | 16000 | 0.0255          | 0.9935   | 0.9986    | 0.9898 | 0.9941 |
-| 0.0247        | 0.3229 | 18000 | 0.0242          | 0.9937   | 0.9983    | 0.9903 | 0.9943 |
-| 0.0232        | 0.3588 | 20000 | 0.0256          | 0.9935   | 0.9976    | 0.9908 | 0.9942 |
-| 0.0243        | 0.3946 | 22000 | 0.0238          | 0.9937   | 0.9979    | 0.9907 | 0.9943 |
-| 0.0235        | 0.4305 | 24000 | 0.0246          | 0.9935   | 0.9969    | 0.9914 | 0.9941 |
-| 0.0267        | 0.4664 | 26000 | 0.0235          | 0.9937   | 0.9975    | 0.9912 | 0.9943 |
-| 0.0228        | 0.5023 | 28000 | 0.0246          | 0.9937   | 0.9977    | 0.9910 | 0.9943 |
-| 0.025         | 0.5382 | 30000 | 0.0226          | 0.9938   | 0.9978    | 0.9912 | 0.9945 |
-| 0.0229        | 0.5740 | 32000 | 0.0229          | 0.9939   | 0.9982    | 0.9909 | 0.9945 |
-| 0.0234        | 0.6099 | 34000 | 0.0237          | 0.9940   | 0.9991    | 0.9900 | 0.9946 |
-| 0.0232        | 0.6458 | 36000 | 0.0230          | 0.9939   | 0.9975    | 0.9915 | 0.9945 |
-| 0.0257        | 0.6817 | 38000 | 0.0228          | 0.9942   | 0.9988    | 0.9907 | 0.9947 |
-| 0.0254        | 0.7175 | 40000 | 0.0221          | 0.9940   | 0.9979    | 0.9914 | 0.9946 |
-| 0.0222        | 0.7534 | 42000 | 0.0223          | 0.9941   | 0.9979    | 0.9915 | 0.9947 |
-| 0.0271        | 0.7893 | 44000 | 0.0219          | 0.9942   | 0.9981    | 0.9914 | 0.9948 |
-| 0.0231        | 0.8252 | 46000 | 0.0222          | 0.9940   | 0.9975    | 0.9917 | 0.9946 |
-| 0.0258        | 0.8610 | 48000 | 0.0214          | 0.9943   | 0.9986    | 0.9912 | 0.9949 |
-| 0.0217        | 0.8969 | 50000 | 0.0219          | 0.9943   | 0.9983    | 0.9914 | 0.9949 |
-| 0.0234        | 0.9328 | 52000 | 0.0215          | 0.9943   | 0.9983    | 0.9914 | 0.9948 |
-| 0.0237        | 0.9687 | 54000 | 0.0215          | 0.9943   | 0.9984    | 0.9913 | 0.9949 |
+| 0.1375        | 0.0718 | 4000  | 0.1431          | 0.9909   | 0.9991    | 0.9847 | 0.9918 |
+| 0.1391        | 0.1436 | 8000  | 0.1356          | 0.9930   | 0.9973    | 0.9902 | 0.9937 |
+| 0.1344        | 0.2154 | 12000 | 0.1361          | 0.9934   | 0.9983    | 0.9899 | 0.9941 |
+| 0.1387        | 0.2872 | 16000 | 0.1333          | 0.9937   | 0.9984    | 0.9903 | 0.9943 |
+| 0.1353        | 0.3590 | 20000 | 0.1340          | 0.9940   | 0.9985    | 0.9907 | 0.9946 |
+| 0.1337        | 0.4308 | 24000 | 0.1332          | 0.9939   | 0.9982    | 0.9909 | 0.9946 |
+| 0.1332        | 0.5026 | 28000 | 0.1332          | 0.9940   | 0.9977    | 0.9916 | 0.9946 |
+| 0.1359        | 0.5744 | 32000 | 0.1319          | 0.9943   | 0.9992    | 0.9907 | 0.9949 |
+| 0.1314        | 0.6462 | 36000 | 0.1326          | 0.9943   | 0.9984    | 0.9914 | 0.9949 |
+| 0.132         | 0.7180 | 40000 | 0.1318          | 0.9945   | 0.9990    | 0.9911 | 0.9950 |
+| 0.1309        | 0.7898 | 44000 | 0.1319          | 0.9945   | 0.9989    | 0.9913 | 0.9951 |
+| 0.1319        | 0.8616 | 48000 | 0.1318          | 0.9945   | 0.9988    | 0.9914 | 0.9951 |
+| 0.1288        | 0.9334 | 52000 | 0.1317          | 0.9946   | 0.9989    | 0.9914 | 0.9951 |
 
 
 ### Framework versions

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5516afa1505f1ad513eb3ef356b92e6d4b5bfe20da25320bbc5cb57e0fc20b2d
+oid sha256:60aa9aa53a0e36b245c5593dce8d7310a3f52c28f31c6edd7c2c5eabf260bcb5
 size 1115268200

runs/Dec07_20-33-59_elesage-pc/events.out.tfevents.1765157729.elesage-pc.222739.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e18ad5fbfff1f448bdecc47ab7759f4c3eed7970aabc680c25c37d6e3f4e1da6
+size 17824

runs/Dec07_20-50-30_elesage-pc/events.out.tfevents.1765158719.elesage-pc.229949.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2879a8efb1d54e147154736614053ac2063e5bc6071b49cce370793598480850
+size 10431

runs/Dec07_21-02-51_elesage-pc/events.out.tfevents.1765159462.elesage-pc.236720.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92105e4dd6cf323869b8b6f92da44ceed8f751069d4a8ba737d505d3064a8833
+size 10858

runs/Dec07_21-07-50_elesage-pc/events.out.tfevents.1765159760.elesage-pc.239632.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0beedb0686f58cb6302686e967459dba8220de99c716b6fd626196b02e188f6a
+size 10225

runs/Dec07_21-14-10_elesage-pc/events.out.tfevents.1765160138.elesage-pc.241974.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:afb0e68fc3008a45ec6cceb0bdc9ad81980c185c71d237d6595fccbfd0aa5fb9
+size 20029

runs/Dec07_21-23-19_elesage-pc/events.out.tfevents.1765160688.elesage-pc.246729.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cf671854f23a0161592d75254ce15ff9e9ca83ad177cbfd224227998332a2165
+size 15543

runs/Dec07_21-28-39_elesage-pc/events.out.tfevents.1765161008.elesage-pc.249831.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1425ba64feb3b6d5b4e3ab98460f322f84ac573425dc3f404eda5c31dbb148a
+size 40652

runs/Dec07_21-38-56_elesage-pc/events.out.tfevents.1765161626.elesage-pc.255681.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:36f4cbd783e9edc1399d009352f332b731e6265558950a50d0d95370d9d976bd
+size 250095

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2b06cc3ee3f4b528ae0bd95797d97625a573b809454df55eb19fb9f727fd3018
+oid sha256:61a02fb851b1134ce920c19e3f3a46c6f21e569e6068dc37c5228abe0d993582
 size 5841