Add DeepSeek-OCR-2 4-bit ONNX model

.gitattributes

@@ -43,3 +43,7 @@ deepseek-ocr2-fp32/decoder.onnx.data filter=lfs diff=lfs merge=lfs -text
 deepseek-ocr2-fp32/vision_encoder.onnx.data filter=lfs diff=lfs merge=lfs -text
 deepseek-ocr2-fp32/vision_sam_1024.onnx.data filter=lfs diff=lfs merge=lfs -text
 deepseek-ocr2-fp32/vision_sam_768.onnx.data filter=lfs diff=lfs merge=lfs -text
+deepseek-ocr2-int4/decoder.onnx.data filter=lfs diff=lfs merge=lfs -text
+deepseek-ocr2-int4/vision_encoder.onnx.data filter=lfs diff=lfs merge=lfs -text
+deepseek-ocr2-int4/vision_sam_1024.onnx.data filter=lfs diff=lfs merge=lfs -text
+deepseek-ocr2-int4/vision_sam_768.onnx.data filter=lfs diff=lfs merge=lfs -text

deepseek-ocr2-int4/config.json

@@ -0,0 +1,52 @@
+{
+ "model": "deepseek-ai/DeepSeek-OCR-2",
+ "dtype": "fp32",
+ "num_layers": 12,
+ "num_heads": 10,
+ "head_dim": 128,
+ "hidden_size": 1280,
+ "vocab_size": 129280,
+ "bos_token_id": 0,
+ "eos_token_id": 1,
+ "image_token_id": 128815,
+ "base_size": 1024,
+ "crop_size": 768,
+ "base_tokens": 256,
+ "crop_tokens": 144,
+ "min_crops": 2,
+ "max_crops": 6,
+ "prompts": {
+  "markdown": {
+   "text": "<image>\n<|grounding|>Convert the document to markdown.",
+   "prefix_ids": [],
+   "suffix_ids": [
+    201,
+    128820,
+    21842,
+    270,
+    4940,
+    304,
+    2121,
+    7919,
+    16
+   ]
+  },
+  "ocr": {
+   "text": "<image>\nFree OCR.",
+   "prefix_ids": [],
+   "suffix_ids": [
+    201,
+    21431,
+    126041,
+    16
+   ]
+  }
+ },
+ "quantization": {
+  "bits": {
+   "decoder": 4,
+   "vision_encoder": 8
+  },
+  "block_size": 32
+ }
+}

deepseek-ocr2-int4/decoder.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0fd80374b2e3a8dac331940e9073d0a3125827fd613075a3efde3cc29f2e375e
+size 266021

deepseek-ocr2-int4/decoder.onnx.data

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab22d5dd3cb9410f6fe21fc32fa4538b4f1440243198117b9097de9b554916b5
+size 1880176640

deepseek-ocr2-int4/vision_encoder.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:147066af2e918a7764f7d6cd5e4602ab72234a430d4c2ca6caf1e4e1dac1777d
+size 983635

deepseek-ocr2-int4/vision_encoder.onnx.data

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:42ffddcbd15d391289c34ca30cda542b9e035a61fb2ed69b0c0c58dc0319e1ad
+size 416763392

deepseek-ocr2-int4/vision_sam_1024.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b0e939b213cfcf8372b271d6a14a54effd1b7886ffa2217eb31af80f76a765f4
+size 596897

deepseek-ocr2-int4/vision_sam_1024.onnx.data

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0c540ba1296e7463170f0b4b0d6af22ad566254176bea145aea9c85e8eda283c
+size 379917312

deepseek-ocr2-int4/vision_sam_768.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f8b99a22ce5aed9b852fff1dafda96ece7f5e05f70fb3bd67dab37addfbd46ae
+size 603589

deepseek-ocr2-int4/vision_sam_768.onnx.data

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:95a07aec108afb8e3192d5a66f59625b89ae1ca471bfc072ac4377aad65076d4
+size 374412288