Update README.md

README.md

@@ -3,33 +3,53 @@ license: apache-2.0
 base_model: distilbert-base-uncased
 tags:
 - generated_from_trainer
-model-index:
-- name: distilbert-base-uncased-fineweb-edu-llama3-annotations-512-vN
-  results: []
+datasets:
+- HuggingFaceFW/fineweb-edu-llama3-annotations
+language:
+- 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. -->
-
 [<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="200" height="32"/>](https://wandb.ai/pszemraj/eduscore-regression/runs/k6z0kenz)
-# distilbert-base-uncased-fineweb-edu-llama3-annotations-512-vN
+# distilbert-base-uncased: edu classifier
+
+> [!IMPORTANT]  
+> This is a (rare) encoder that supports flash attention 2!
+> Use `attn_implementation="flash_attention_2"` when loading w/ [FA2 installed](https://github.com/Dao-AILab/flash-attention?tab=readme-ov-file#installation-and-features)
+> for faster inference.
 
 This model is a fine-tuned version of [distilbert-base-uncased](https://huggingface.co/distilbert-base-uncased) on the HuggingFaceFW/fineweb-edu-llama3-annotations dataset.
 It achieves the following results on the evaluation set:
 - Loss: 0.2324
 - Mse: 0.2324
 
-## Model description
+## Usage 
 
-More information needed
+Note this is for CPU, for GPU you will need to make some (small) changes.
 
-## Intended uses & limitations
+```py
+from transformers import AutoTokenizer, AutoModelForSequenceClassification
 
-More information needed
+tokenizer = AutoTokenizer.from_pretrained("pszemraj/mpnet-base-edu-classifier")
+model = AutoModelForSequenceClassification.from_pretrained("pszemraj/mpnet-base-edu-classifier")
 
-## Training and evaluation data
+text = "This is a test sentence."
+inputs = tokenizer(text, return_tensors="pt", padding="longest", truncation=True)
+outputs = model(**inputs)
+logits = outputs.logits.squeeze(-1).float().detach().numpy()
+score = logits.item()
+result = {
+    "text": text,
+    "score": score,
+    "int_score": int(round(max(0, min(score, 5)))),
+}
+
+print(result)
+# {'text': 'This is a test sentence.', 'score': 0.3350256383419037, 'int_score': 0}
+```
+
+## Intended uses & limitations
 
-More information needed
+Refer to the hf classifier's [model card](https://huggingface.co/HuggingFaceFW/fineweb-edu-classifier#limitations) for more details
 
 ## Training procedure
 
@@ -46,50 +66,3 @@ The following hyperparameters were used during training:
 - lr_scheduler_type: linear
 - lr_scheduler_warmup_ratio: 0.05
 - num_epochs: 1.0
-
-### Training results
-
-| Training Loss | Epoch  | Step | Validation Loss | Mse    |
-|:-------------:|:------:|:----:|:---------------:|:------:|
-| 0.5361        | 0.0288 | 100  | 0.4934          | 0.4934 |
-| 0.3483        | 0.0576 | 200  | 0.3525          | 0.3525 |
-| 0.3238        | 0.0865 | 300  | 0.2931          | 0.2931 |
-| 0.2734        | 0.1153 | 400  | 0.3130          | 0.3130 |
-| 0.2891        | 0.1441 | 500  | 0.3298          | 0.3298 |
-| 0.2807        | 0.1729 | 600  | 0.2659          | 0.2659 |
-| 0.2727        | 0.2018 | 700  | 0.2690          | 0.2690 |
-| 0.2701        | 0.2306 | 800  | 0.2555          | 0.2555 |
-| 0.2954        | 0.2594 | 900  | 0.2501          | 0.2501 |
-| 0.2618        | 0.2882 | 1000 | 0.2483          | 0.2483 |
-| 0.3081        | 0.3171 | 1100 | 0.2456          | 0.2456 |
-| 0.2544        | 0.3459 | 1200 | 0.2370          | 0.2370 |
-| 0.2593        | 0.3747 | 1300 | 0.2349          | 0.2349 |
-| 0.2361        | 0.4035 | 1400 | 0.2406          | 0.2406 |
-| 0.2536        | 0.4324 | 1500 | 0.2453          | 0.2453 |
-| 0.26          | 0.4612 | 1600 | 0.2568          | 0.2568 |
-| 0.2897        | 0.4900 | 1700 | 0.2568          | 0.2568 |
-| 0.2597        | 0.5188 | 1800 | 0.2359          | 0.2359 |
-| 0.2489        | 0.5477 | 1900 | 0.2413          | 0.2413 |
-| 0.2376        | 0.5765 | 2000 | 0.2416          | 0.2416 |
-| 0.2424        | 0.6053 | 2100 | 0.2418          | 0.2418 |
-| 0.2798        | 0.6341 | 2200 | 0.2462          | 0.2462 |
-| 0.2523        | 0.6630 | 2300 | 0.2322          | 0.2322 |
-| 0.286         | 0.6918 | 2400 | 0.2432          | 0.2432 |
-| 0.247         | 0.7206 | 2500 | 0.2383          | 0.2383 |
-| 0.2856        | 0.7494 | 2600 | 0.2375          | 0.2375 |
-| 0.2216        | 0.7783 | 2700 | 0.2383          | 0.2383 |
-| 0.255         | 0.8071 | 2800 | 0.2367          | 0.2367 |
-| 0.2406        | 0.8359 | 2900 | 0.2345          | 0.2345 |
-| 0.2388        | 0.8647 | 3000 | 0.2282          | 0.2282 |
-| 0.2571        | 0.8936 | 3100 | 0.2331          | 0.2331 |
-| 0.2672        | 0.9224 | 3200 | 0.2336          | 0.2336 |
-| 0.2375        | 0.9512 | 3300 | 0.2337          | 0.2337 |
-| 0.2423        | 0.9800 | 3400 | 0.2324          | 0.2324 |
-
-
-### Framework versions
-
-- Transformers 4.42.3
-- Pytorch 2.3.1+cu121
-- Datasets 2.20.0
-- Tokenizers 0.19.1