pshashid commited on
Commit
0667e40
·
verified ·
1 Parent(s): 70f63db

Add/update smolvlm2-256m

Browse files
smolvlm2-256m/meta.json CHANGED
@@ -4,7 +4,7 @@
4
  "method": "vlm-onnx",
5
  "note": null,
6
  "revision": "067788b187b95ebe7b2e040b3e4299e342e5b8fd",
7
- "layout": "smolvlm-official",
8
  "bundle_files": {
9
  "added_tokens.json": "74135b8664b56088c0006f1c8e848d79a8eba003411f72ebf1dc2ee96227be3a",
10
  "chat_template.json": "b585e3598909a5687f9f9d738d35223724dedef256b9b274e1cbfb32b13c74bf",
@@ -13,7 +13,7 @@
13
  "merges.txt": "0b54e8aa4e53d5383e2e4bc635a56b43f9647f7b13832d5d9ecd8f82dac4f510",
14
  "onnx/decoder_model_merged.onnx": "1563e4240f154ba95f5b24e6746d4116b9b43a1567aaf9781841f62be220983d",
15
  "onnx/embed_tokens.onnx": "05c2c945e08b8f87088a1e855fc363782c8e65f1f7283c30382d93b568d2bbf8",
16
- "onnx/vision_encoder.onnx": "9f2fa81b77ed18b8372da99923ed846db8b04de8bd62bd4d32352a1f1b725f49",
17
  "preprocessor_config.json": "149e315d9410368e5491455bb06e0f763426e9e56cca731c13b24404a29b6374",
18
  "processor_config.json": "f3ad45028447b3562b4752be0d5916d6806c1ef589091a469608dcf0faa1737c",
19
  "special_tokens_map.json": "2dfea2a426162316ff1567c82bc6d36d9690cd9f90455f075c77daca78b45c60",
@@ -29,7 +29,17 @@
29
  "embed_tokens.onnx": "05c2c945e08b8f87088a1e855fc363782c8e65f1f7283c30382d93b568d2bbf8",
30
  "decoder_model_merged.onnx": "2ff9605f7493cd2283bad9d9c14c5145f5c8b5f6dff8a197d05b247f5871d378"
31
  },
32
- "normalization": "experimental RMSNorm expanded to standard ONNX"
 
 
 
 
 
 
 
 
 
 
33
  },
34
  "opset_requested": 17,
35
  "files": {
@@ -799,32 +809,34 @@
799
  "opset": 13
800
  },
801
  "onnx/vision_encoder.onnx": {
802
- "sha256": "9f2fa81b77ed18b8372da99923ed846db8b04de8bd62bd4d32352a1f1b725f49",
803
- "bytes": 374316454,
804
  "inputs": {
805
  "pixel_values": [
806
- "batch_size",
807
  "num_images",
808
  3,
809
  512,
810
  512
811
  ],
812
- "pixel_attention_mask": [
813
- "batch_size",
814
  "num_images",
815
- 512,
816
- 512
 
 
 
817
  ]
818
  },
819
  "outputs": {
820
  "image_features": [
821
- "NonZero_7_o0__d1",
822
- 64,
823
  576
824
  ]
825
  },
826
- "opset": 14
827
  }
828
  },
829
- "exported_at": "2026-11-12T21:28:58.333297+00:00"
830
  }
 
4
  "method": "vlm-onnx",
5
  "note": null,
6
  "revision": "067788b187b95ebe7b2e040b3e4299e342e5b8fd",
7
+ "layout": "smolvlm-source-vision-v1",
8
  "bundle_files": {
9
  "added_tokens.json": "74135b8664b56088c0006f1c8e848d79a8eba003411f72ebf1dc2ee96227be3a",
10
  "chat_template.json": "b585e3598909a5687f9f9d738d35223724dedef256b9b274e1cbfb32b13c74bf",
 
13
  "merges.txt": "0b54e8aa4e53d5383e2e4bc635a56b43f9647f7b13832d5d9ecd8f82dac4f510",
14
  "onnx/decoder_model_merged.onnx": "1563e4240f154ba95f5b24e6746d4116b9b43a1567aaf9781841f62be220983d",
15
  "onnx/embed_tokens.onnx": "05c2c945e08b8f87088a1e855fc363782c8e65f1f7283c30382d93b568d2bbf8",
16
+ "onnx/vision_encoder.onnx": "1e03130f4c92ebaba796a03aaeae7fb6d29512d3bf489a2c06c106d1a9ef52eb",
17
  "preprocessor_config.json": "149e315d9410368e5491455bb06e0f763426e9e56cca731c13b24404a29b6374",
18
  "processor_config.json": "f3ad45028447b3562b4752be0d5916d6806c1ef589091a469608dcf0faa1737c",
19
  "special_tokens_map.json": "2dfea2a426162316ff1567c82bc6d36d9690cd9f90455f075c77daca78b45c60",
 
29
  "embed_tokens.onnx": "05c2c945e08b8f87088a1e855fc363782c8e65f1f7283c30382d93b568d2bbf8",
30
  "decoder_model_merged.onnx": "2ff9605f7493cd2283bad9d9c14c5145f5c8b5f6dff8a197d05b247f5871d378"
31
  },
32
+ "normalization": "experimental RMSNorm expanded to standard ONNX",
33
+ "vision_reexport": {
34
+ "method": "pinned source vision export; ORT vision and end-to-end prefill parity",
35
+ "max_abs_error": {
36
+ "vision_1_images": 0.0001621246337890625,
37
+ "full_prefill_1_images": 3.528594970703125e-05,
38
+ "vision_2_images": 6.866455078125e-05,
39
+ "full_prefill_2_images": 3.0994415283203125e-05
40
+ },
41
+ "official_vision_rejected": true
42
+ }
43
  },
44
  "opset_requested": 17,
45
  "files": {
 
809
  "opset": 13
810
  },
811
  "onnx/vision_encoder.onnx": {
812
+ "sha256": "1e03130f4c92ebaba796a03aaeae7fb6d29512d3bf489a2c06c106d1a9ef52eb",
813
+ "bytes": 374203453,
814
  "inputs": {
815
  "pixel_values": [
816
+ 1,
817
  "num_images",
818
  3,
819
  512,
820
  512
821
  ],
822
+ "vision_position_ids": [
 
823
  "num_images",
824
+ 1024
825
+ ],
826
+ "patch_attention_mask": [
827
+ "num_images",
828
+ 1024
829
  ]
830
  },
831
  "outputs": {
832
  "image_features": [
833
+ "num_images",
834
+ "MatMulimage_features_dim_1",
835
  576
836
  ]
837
  },
838
+ "opset": 17
839
  }
840
  },
841
+ "exported_at": "2026-11-12T22:09:44.587042+00:00"
842
  }
smolvlm2-256m/onnx/vision_encoder.onnx CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:9f2fa81b77ed18b8372da99923ed846db8b04de8bd62bd4d32352a1f1b725f49
3
- size 374316454
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1e03130f4c92ebaba796a03aaeae7fb6d29512d3bf489a2c06c106d1a9ef52eb
3
+ size 374203453