Trimmed MetaCLIP-2 text vocab for Esperanto

README.md

@@ -14,41 +14,13 @@ datasets:
 
 # metaclip-2-worldwide-m16-384-epo-32768
 
-This model is a 82.6% smaller version of [facebook/metaclip-2-worldwide-m16-384](https://huggingface.co/facebook/metaclip-2-worldwide-m16-384) 
-optimized for Esperanto language via **text vocabulary trimming** mined on 
-[Lumberjackk/fineweb-2-trimming](https://huggingface.co/datasets/Lumberjackk/fineweb-2-trimming).
-
-Only the **text encoder** vocabulary and embeddings are trimmed.  
-The **vision encoder** is kept identical to the original.
+This model is a 82.6% smaller version of [facebook/metaclip-2-worldwide-m16-384](https://huggingface.co/facebook/metaclip-2-worldwide-m16-384)
+optimized for Esperanto via text vocabulary trimming.
 
 ## Model Statistics
-- **Original text vocabulary size:** 901,629 tokens
-- **Trimmed text vocabulary size:** 32,767 tokens
+- **Original vocab size:** 901,629
+- **Trimmed vocab size:** 32,768
 - **Vocabulary reduction:** 96.4%
-- **Original model size:** 538,547,201 parameters
-- **Trimmed model size:** 93,690,881 parameters
+- **Original params:** 538,547,201
+- **Trimmed params:** 93,690,369
 - **Size reduction:** 82.6%
-
-## Usage
-
-```python
-from transformers import AutoProcessor, AutoModel
-import torch
-
-model = AutoModel.from_pretrained("provisoirement/metaclip-2-worldwide-m16-384-epo-32768")
-processor = AutoProcessor.from_pretrained("provisoirement/metaclip-2-worldwide-m16-384-epo-32768")
-
-# Text-only encoding
-texts = ["a photo of a cat", "a photo of a dog"]
-inputs = processor(text=texts, return_tensors="pt", padding=True)
-with torch.inference_mode():
-    text_features = model.get_text_features(**inputs)
-
-# Image + text (standard CLIP usage)
-# from PIL import Image
-# image = Image.open("image.jpg")
-# inputs = processor(images=image, text=texts, return_tensors="pt", padding=True)
-# with torch.inference_mode():
-#     outputs = model(**inputs)
-#     logits_per_image = outputs.logits_per_image
-```

config.json

@@ -21,7 +21,7 @@
     "num_attention_heads": 8,
     "num_hidden_layers": 12,
     "projection_dim": 512,
-    "vocab_size": 32769
+    "vocab_size": 32768
   },
   "transformers_version": "4.57.1",
   "vision_config": {

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:af993a6c6b6f3301dc506d8a40531c54ed588f112f476623bc33398c7ea5081d
-size 374810868
+oid sha256:acc9e60667f62287c900ca8bbdc5431e398f7b5b1ad9c937141f07cb45403441
+size 374808820

tokenizer.json

@@ -61,15 +61,6 @@
       "rstrip": false,
       "normalized": false,
       "special": true
-    },
-    {
-      "id": 32768,
-      "content": "<image_soft_token>",
-      "single_word": false,
-      "lstrip": false,
-      "rstrip": false,
-      "normalized": true,
-      "special": false
     }
   ],
   "normalizer": {