Upload ENVISION eye imaging classifier v1.0

1_Pooling/config.json

@@ -1,7 +1,7 @@
 {
-  "word_embedding_dimension": 1024,
-  "pooling_mode_cls_token": true,
-  "pooling_mode_mean_tokens": false,
+  "word_embedding_dimension": 768,
+  "pooling_mode_cls_token": false,
+  "pooling_mode_mean_tokens": true,
   "pooling_mode_max_tokens": false,
   "pooling_mode_mean_sqrt_len_tokens": false,
   "pooling_mode_weightedmean_tokens": false,

README.md

@@ -19,7 +19,7 @@ SetFit few-shot classifier for identifying eye imaging datasets from scientific
 
 ## Model Description
 
-Uses `Alibaba-NLP/gte-large-en-v1.5` as backbone with 4-class classification:
+Uses `sentence-transformers/all-mpnet-base-v2` as backbone with 4-class classification:
 
 - **EYE_IMAGING (3)**: Actual ophthalmic imaging datasets (fundus, OCT, OCTA, cornea)
 - **EYE_SOFTWARE (2)**: Code, tools, models for eye imaging
@@ -32,18 +32,18 @@ Tested on 515 Zenodo datasets (filtered to `resource_type=dataset` only):
 
 | Class | Count |
 |-------|-------|
-| EYE_IMAGING | 120 |
-| EYE_SOFTWARE | 66 |
-| EDGE_CASE | 3 |
-| NEGATIVE | 325 |
+| EYE_IMAGING | 127 |
+| EYE_SOFTWARE | 24 |
+| EDGE_CASE | 32 |
+| NEGATIVE | 331 |
 
 ### Confidence Distribution (EYE_IMAGING)
 
 | Confidence | Count | % |
 |------------|-------|---|
-| High (≥0.95) | 117 | 97.5% |
-| Medium (0.80-0.95) | 2 | 1.7% |
-| Lower (<0.80) | 1 | 0.8% |
+| High (>=0.95) | 49 | 38.6% |
+| Medium (0.80-0.95) | 70 | 55.1% |
+| Lower (<0.80) | 8 | 6.3% |
 
 ### Data Pipeline
 
@@ -53,17 +53,43 @@ Tested on 515 Zenodo datasets (filtered to `resource_type=dataset` only):
 
 ## Training
 
-- **Examples**: 452 (99 positive, 30 software, 90 edge case, 233 negative)
+- **Base Model**: `sentence-transformers/all-mpnet-base-v2` (768-dimensional embeddings)
+- **Examples**: 474 (77 EYE_IMAGING, 48 EYE_SOFTWARE, 79 EDGE_CASE, 270 NEGATIVE)
 - **Epochs**: 2
 - **Batch Size**: 16
 
+## Validation
+
+### Held-out Test Set
+
+| Metric | Value |
+|--------|-------|
+| Accuracy | 0.937 |
+| Macro F1 | 0.902 |
+
+### Spot-Check Validation (33 records)
+
+| Metric | Value |
+|--------|-------|
+| Accuracy | 29/33 (87.9%) |
+| Macro F1 | 0.828 |
+
+#### Per-Class F1 (Spot-Check)
+
+| Class | F1 |
+|-------|-----|
+| EYE_IMAGING | 0.947 |
+| EDGE_CASE | 0.889 |
+| NEGATIVE | 0.903 |
+| EYE_SOFTWARE | 0.571 |
+
 ## Usage
 
 ```python
 from sentence_transformers import SentenceTransformer
 import joblib
 
-model = SentenceTransformer("jimnoneill/envision-eye-imaging-classifier", trust_remote_code=True)
+model = SentenceTransformer("fairdataihub/envision-eye-imaging-classifier")
 head = joblib.load("model_head.pkl")
 
 embeddings = model.encode(["Retinal OCT dataset for diabetic retinopathy"])
@@ -74,9 +100,8 @@ predictions = head.predict(embeddings)
 
 - EyeACT Envision project
 - FAIR Data Innovations Hub (fairdataihub.org)
-- Alibaba-NLP/gte-large-en-v1.5
+- sentence-transformers/all-mpnet-base-v2
 
 ## Contact
 
 EyeACT team: [eyeactstudy.org](https://eyeactstudy.org)
-

config.json

@@ -1,43 +1,23 @@
 {
   "architectures": [
-    "NewModel"
+    "MPNetModel"
   ],
-  "attention_probs_dropout_prob": 0.0,
-  "auto_map": {
-    "AutoConfig": "configuration.NewConfig",
-    "AutoModel": "modeling.NewModel",
-    "AutoModelForMaskedLM": "Alibaba-NLP/new-impl--modeling.NewForMaskedLM",
-    "AutoModelForMultipleChoice": "Alibaba-NLP/new-impl--modeling.NewForMultipleChoice",
-    "AutoModelForQuestionAnswering": "Alibaba-NLP/new-impl--modeling.NewForQuestionAnswering",
-    "AutoModelForSequenceClassification": "Alibaba-NLP/new-impl--modeling.NewForSequenceClassification",
-    "AutoModelForTokenClassification": "Alibaba-NLP/new-impl--modeling.NewForTokenClassification"
-  },
-  "classifier_dropout": null,
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": 0,
   "dtype": "float32",
+  "eos_token_id": 2,
   "hidden_act": "gelu",
   "hidden_dropout_prob": 0.1,
-  "hidden_size": 1024,
+  "hidden_size": 768,
   "initializer_range": 0.02,
-  "intermediate_size": 4096,
-  "layer_norm_eps": 1e-12,
-  "layer_norm_type": "layer_norm",
-  "logn_attention_clip1": false,
-  "logn_attention_scale": false,
-  "max_position_embeddings": 8192,
-  "model_type": "new",
-  "num_attention_heads": 16,
-  "num_hidden_layers": 24,
-  "pack_qkv": true,
-  "pad_token_id": 0,
-  "position_embedding_type": "rope",
-  "rope_scaling": {
-    "factor": 2.0,
-    "type": "ntk"
-  },
-  "rope_theta": 160000,
+  "intermediate_size": 3072,
+  "layer_norm_eps": 1e-05,
+  "max_position_embeddings": 514,
+  "model_type": "mpnet",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 1,
+  "relative_attention_num_buckets": 32,
   "transformers_version": "4.57.1",
-  "type_vocab_size": 2,
-  "unpad_inputs": false,
-  "use_memory_efficient_attention": false,
-  "vocab_size": 30528
+  "vocab_size": 30527
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8926bf7bdf17bf066a4365684d38d9fd0a0fcefff070d1cbb2b5147082fb42f8
-size 1736585680
+oid sha256:0b0538e5a8c2eb88301ca85f98f4e049d0ccb3e5269f9d63fb853449d5f60ad5
+size 437967672

model_head.pkl

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ead33c08eab190ebc8ab22952cc33fe8c100b3615f0306134cc41ee83a190c22
-size 33671
+oid sha256:4cd66021541f6dc6f6da7ff44f660b2c1abdc762837c30bfc7ae7652cc812e5c
+size 25479

modules.json

@@ -10,5 +10,11 @@
     "name": "1",
     "path": "1_Pooling",
     "type": "sentence_transformers.models.Pooling"
+  },
+  {
+    "idx": 2,
+    "name": "2",
+    "path": "2_Normalize",
+    "type": "sentence_transformers.models.Normalize"
   }
 ]

sentence_bert_config.json

@@ -1,4 +1,4 @@
 {
-  "max_seq_length": 8192,
+  "max_seq_length": 384,
   "do_lower_case": false
 }

special_tokens_map.json

@@ -1,27 +1,41 @@
 {
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
   "cls_token": {
-    "content": "[CLS]",
+    "content": "<s>",
     "lstrip": false,
     "normalized": false,
     "rstrip": false,
     "single_word": false
   },
-  "mask_token": {
-    "content": "[MASK]",
+  "eos_token": {
+    "content": "</s>",
     "lstrip": false,
     "normalized": false,
     "rstrip": false,
     "single_word": false
   },
+  "mask_token": {
+    "content": "<mask>",
+    "lstrip": true,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
   "pad_token": {
-    "content": "[PAD]",
+    "content": "<pad>",
     "lstrip": false,
     "normalized": false,
     "rstrip": false,
     "single_word": false
   },
   "sep_token": {
-    "content": "[SEP]",
+    "content": "</s>",
     "lstrip": false,
     "normalized": false,
     "rstrip": false,

tokenizer_config.json

@@ -1,62 +1,72 @@
 {
   "added_tokens_decoder": {
     "0": {
-      "content": "[PAD]",
+      "content": "<s>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "100": {
-      "content": "[UNK]",
+    "1": {
+      "content": "<pad>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "101": {
-      "content": "[CLS]",
+    "2": {
+      "content": "</s>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "102": {
-      "content": "[SEP]",
+    "3": {
+      "content": "<unk>",
       "lstrip": false,
-      "normalized": false,
+      "normalized": true,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "103": {
-      "content": "[MASK]",
+    "104": {
+      "content": "[UNK]",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
+    },
+    "30526": {
+      "content": "<mask>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
     }
   },
-  "clean_up_tokenization_spaces": true,
-  "cls_token": "[CLS]",
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": false,
+  "cls_token": "<s>",
   "do_lower_case": true,
+  "eos_token": "</s>",
   "extra_special_tokens": {},
-  "mask_token": "[MASK]",
-  "max_length": 8000,
-  "model_max_length": 8192,
+  "mask_token": "<mask>",
+  "max_length": 128,
+  "model_max_length": 384,
   "pad_to_multiple_of": null,
-  "pad_token": "[PAD]",
+  "pad_token": "<pad>",
   "pad_token_type_id": 0,
   "padding_side": "right",
-  "sep_token": "[SEP]",
+  "sep_token": "</s>",
   "stride": 0,
   "strip_accents": null,
   "tokenize_chinese_chars": true,
-  "tokenizer_class": "BertTokenizer",
+  "tokenizer_class": "MPNetTokenizer",
   "truncation_side": "right",
   "truncation_strategy": "longest_first",
   "unk_token": "[UNK]"

vocab.txt

@@ -1,3 +1,7 @@
+<s>
+<pad>
+</s>
+<unk>
 [PAD]
 [unused0]
 [unused1]
@@ -30520,3 +30524,4 @@ necessitated
 ##：
 ##？
 ##～
+<mask>