Upload folder using huggingface_hub

README.md

@@ -0,0 +1,90 @@
+---
+license: apache-2.0
+language:
+- en
+library_name: mlx
+tags:
+- sentence-transformers
+- sentence-similarity
+- feature-extraction
+- food
+- embeddings
+- mlx
+- apple-silicon
+base_model: thenlper/gte-large
+pipeline_tag: sentence-similarity
+---
+
+# FoodMapper GTE-Large (MLX Format)
+
+This is [thenlper/gte-large](https://huggingface.co/thenlper/gte-large) converted to MLX-Swift safetensors format for use with the FoodMapper macOS application.
+
+## Model Description
+
+GTE-Large is a 335M parameter text embedding model that maps sentences to 1024-dimensional dense vectors. It excels at semantic similarity tasks, making it ideal for matching food names across different databases and nomenclatures.
+
+This conversion is optimized for Apple Silicon GPUs via [MLX-Swift](https://github.com/ml-explore/mlx-swift).
+
+## Intended Use
+
+- Semantic food name matching (e.g., matching "granny smith apple" to "Apple, raw, with skin")
+- Food database harmonization between USDA FoodData Central, FooDB, and custom datasets
+- General text similarity on Apple Silicon Macs
+
+## Model Details
+
+| Property | Value |
+|----------|-------|
+| Parameters | 335M |
+| Embedding Dimension | 1024 |
+| Max Sequence Length | 512 |
+| Architecture | BERT |
+| Precision | float16 |
+| Format | safetensors |
+
+## Files
+
+- `gte-large.safetensors` - Model weights in safetensors format (~670MB)
+- `config.json` - Model architecture configuration
+- `tokenizer.json` - Tokenizer vocabulary and settings
+- `tokenizer_config.json` - Tokenizer configuration
+- `vocab.txt` - WordPiece vocabulary
+- `special_tokens_map.json` - Special token mappings
+
+## Usage with FoodMapper
+
+This model is automatically downloaded by the FoodMapper macOS app when first launched. No manual setup required.
+
+## Usage with MLX-Swift
+
+```swift
+import MLX
+import MLXNN
+
+// Load weights
+let weights = try loadArrays(url: modelURL)
+let parameters = ModuleParameters.unflattened(weights)
+try model.update(parameters: parameters, verify: .none)
+```
+
+## Pooling
+
+GTE models use **mean pooling** over token embeddings (not CLS token pooling). The attention mask should be applied before averaging:
+
+```swift
+func meanPooling(_ hiddenState: MLXArray, attentionMask: MLXArray) -> MLXArray {
+    let maskExpanded = attentionMask.expandedDimensions(axis: -1)
+        .asType(hiddenState.dtype)
+    let sumEmbeddings = (hiddenState * maskExpanded).sum(axis: 1)
+    let sumMask = MLX.maximum(maskExpanded.sum(axis: 1), MLXArray(1e-9))
+    return sumEmbeddings / sumMask
+}
+```
+
+## Original Model
+
+Based on [thenlper/gte-large](https://huggingface.co/thenlper/gte-large) by Alibaba DAMO Academy.
+
+## License
+
+Apache 2.0 (same as original GTE-Large)

config.json

@@ -0,0 +1,25 @@
+{
+  "architectures": [
+    "BertModel"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "classifier_dropout": null,
+  "gradient_checkpointing": false,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 24,
+  "pad_token_id": 0,
+  "position_embedding_type": "absolute",
+  "torch_dtype": "float16",
+  "transformers_version": "4.28.1",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 30522
+}

gte-large.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f917f334b6e38e966519983a6b567a5a86d90065932c780f6b4ad72e6bf3a90b
+size 670326040

special_tokens_map.json

@@ -0,0 +1,7 @@
+{
+  "cls_token": "[CLS]",
+  "mask_token": "[MASK]",
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "unk_token": "[UNK]"
+}

test_baseline.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cf14068025e2b0831a3e45f202db9bfef9a48e164877a1f73d98ffbc19d82f5c
+size 20608

test_texts.json

@@ -0,0 +1,7 @@
+[
+  "red wine, table",
+  "chicken breast, roasted, skinless",
+  "apple, fresh, raw, with skin",
+  "milk, whole, 3.25% fat",
+  "bread, white, commercially prepared"
+]

tokenizer_config.json

@@ -0,0 +1,13 @@
+{
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "[CLS]",
+  "do_lower_case": true,
+  "mask_token": "[MASK]",
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "strip_accents": null,
+  "tokenize_chinese_chars": true,
+  "tokenizer_class": "BertTokenizer",
+  "unk_token": "[UNK]"
+}