Hugging Face's logo

xiaomoguhzz
/
VisionEncoder

Transformers
Safetensors
vision-encoder
distillation
video-language
siglip2
dinov3
Model card Files
xet
 Community
xiaomoguhzz commited on
Commit
56b91c9
·
verified ·
1 Parent(s): 19cba99

Add vmllm_cached cached datasets (6.2G, 693 files)

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. vmllm_cached/qwen3vit/image_10pct/train/data-00000-of-00001.arrow +3 -0
  2. vmllm_cached/qwen3vit/image_10pct/train/dataset_info.json +65 -0
  3. vmllm_cached/qwen3vit/image_10pct/train/state.json +13 -0
  4. vmllm_cached/qwen3vit/image_full/train/data-00000-of-00002.arrow +3 -0
  5. vmllm_cached/qwen3vit/image_full/train/data-00001-of-00002.arrow +3 -0
  6. vmllm_cached/qwen3vit/image_full/train/dataset_info.json +65 -0
  7. vmllm_cached/qwen3vit/image_full/train/state.json +16 -0
  8. vmllm_cached/qwen3vit/video_10pct/train/data-00000-of-00001.arrow +3 -0
  9. vmllm_cached/qwen3vit/video_10pct/train/dataset_info.json +60 -0
  10. vmllm_cached/qwen3vit/video_10pct/train/state.json +13 -0
  11. vmllm_cached/qwen3vit/video_full/train/data-00000-of-00003.arrow +3 -0
  12. vmllm_cached/qwen3vit/video_full/train/data-00001-of-00003.arrow +3 -0
  13. vmllm_cached/qwen3vit/video_full/train/data-00002-of-00003.arrow +3 -0
  14. vmllm_cached/qwen3vit/video_full/train/dataset_info.json +60 -0
  15. vmllm_cached/qwen3vit/video_full/train/state.json +19 -0
  16. vmllm_cached/qwenvit_v4_1/image_10pct/train/data-00000-of-00001.arrow +3 -0
  17. vmllm_cached/qwenvit_v4_1/image_10pct/train/dataset_info.json +65 -0
  18. vmllm_cached/qwenvit_v4_1/image_10pct/train/state.json +1 -0
  19. vmllm_cached/qwenvit_v4_1/image_10pct/train/v3_patch_meta.json +11 -0
  20. vmllm_cached/qwenvit_v4_1/video_10pct/train/data-00000-of-00001.arrow +3 -0
  21. vmllm_cached/qwenvit_v4_1/video_10pct/train/dataset_info.json +60 -0
  22. vmllm_cached/qwenvit_v4_1/video_10pct/train/state.json +1 -0
  23. vmllm_cached/qwenvit_v4_1/video_10pct/train/v3_patch_meta.json +11 -0
  24. vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00000_of_00016.arrow +3 -0
  25. vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00001_of_00016.arrow +3 -0
  26. vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00002_of_00016.arrow +3 -0
  27. vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00003_of_00016.arrow +3 -0
  28. vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00004_of_00016.arrow +3 -0
  29. vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00005_of_00016.arrow +3 -0
  30. vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00006_of_00016.arrow +3 -0
  31. vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00007_of_00016.arrow +3 -0
  32. vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00008_of_00016.arrow +3 -0
  33. vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00009_of_00016.arrow +3 -0
  34. vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00010_of_00016.arrow +3 -0
  35. vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00011_of_00016.arrow +3 -0
  36. vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00012_of_00016.arrow +3 -0
  37. vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00013_of_00016.arrow +3 -0
  38. vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00014_of_00016.arrow +3 -0
  39. vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00015_of_00016.arrow +3 -0
  40. vmllm_cached/siglip2/image_10pct/train/cache-2b2dbbebf1aceb79_00000_of_00016.arrow +3 -0
  41. vmllm_cached/siglip2/image_10pct/train/cache-2b2dbbebf1aceb79_00001_of_00016.arrow +3 -0
  42. vmllm_cached/siglip2/image_10pct/train/cache-2b2dbbebf1aceb79_00002_of_00016.arrow +3 -0
  43. vmllm_cached/siglip2/image_10pct/train/cache-2b2dbbebf1aceb79_00003_of_00016.arrow +3 -0
  44. vmllm_cached/siglip2/image_10pct/train/cache-2b2dbbebf1aceb79_00004_of_00016.arrow +3 -0
  45. vmllm_cached/siglip2/image_10pct/train/cache-2b2dbbebf1aceb79_00005_of_00016.arrow +3 -0
  46. vmllm_cached/siglip2/image_10pct/train/cache-2b2dbbebf1aceb79_00006_of_00016.arrow +3 -0
  47. vmllm_cached/siglip2/image_10pct/train/cache-2b2dbbebf1aceb79_00007_of_00016.arrow +3 -0
  48. vmllm_cached/siglip2/image_10pct/train/cache-2b2dbbebf1aceb79_00008_of_00016.arrow +3 -0
  49. vmllm_cached/siglip2/image_10pct/train/cache-2b2dbbebf1aceb79_00009_of_00016.arrow +3 -0
  50. vmllm_cached/siglip2/image_10pct/train/cache-2b2dbbebf1aceb79_00010_of_00016.arrow +3 -0
vmllm_cached/qwen3vit/image_10pct/train/data-00000-of-00001.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dde6c7cd2ce6c46b02dec2761a3aad5040dc692cce2c7f97f505767f4c53a880
3
+ size 69352320
vmllm_cached/qwen3vit/image_10pct/train/dataset_info.json ADDED
@@ -0,0 +1,65 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "builder_name": "json",
3
+ "citation": "",
4
+ "config_name": "default",
5
+ "dataset_name": "json",
6
+ "dataset_size": 66130519,
7
+ "description": "",
8
+ "download_checksums": {
9
+ "/share/m2v_intern_v3/wangjunjie09/model_cache/VisionEncoder/ms-swift-data/image_sft_small_10pct_sharegpt.json": {
10
+ "num_bytes": 85235535,
11
+ "checksum": null
12
+ }
13
+ },
14
+ "download_size": 85235535,
15
+ "features": {
16
+ "messages": [
17
+ {
18
+ "content": {
19
+ "dtype": "string",
20
+ "_type": "Value"
21
+ },
22
+ "role": {
23
+ "dtype": "string",
24
+ "_type": "Value"
25
+ }
26
+ }
27
+ ],
28
+ "images": [
29
+ {
30
+ "bytes": {
31
+ "dtype": "null",
32
+ "_type": "Value"
33
+ },
34
+ "path": {
35
+ "dtype": "string",
36
+ "_type": "Value"
37
+ }
38
+ }
39
+ ],
40
+ "lengths": {
41
+ "feature": {
42
+ "dtype": "int64",
43
+ "_type": "Value"
44
+ },
45
+ "_type": "Sequence"
46
+ }
47
+ },
48
+ "homepage": "",
49
+ "license": "",
50
+ "size_in_bytes": 151366054,
51
+ "splits": {
52
+ "train": {
53
+ "name": "train",
54
+ "num_bytes": 66130519,
55
+ "num_examples": 73859,
56
+ "dataset_name": "json"
57
+ }
58
+ },
59
+ "version": {
60
+ "version_str": "0.0.0",
61
+ "major": 0,
62
+ "minor": 0,
63
+ "patch": 0
64
+ }
65
+ }
vmllm_cached/qwen3vit/image_10pct/train/state.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_data_files": [
3
+ {
4
+ "filename": "data-00000-of-00001.arrow"
5
+ }
6
+ ],
7
+ "_fingerprint": "7cd6dba47f471540",
8
+ "_format_columns": null,
9
+ "_format_kwargs": {},
10
+ "_format_type": null,
11
+ "_output_all_columns": false,
12
+ "_split": null
13
+ }
vmllm_cached/qwen3vit/image_full/train/data-00000-of-00002.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:242b7e98f62eb717313586119a386c5911a1d42d7910f6b0db37e549ba97b86b
3
+ size 347798136
vmllm_cached/qwen3vit/image_full/train/data-00001-of-00002.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:66b7932b167a90d96a6ec07e778c0d0a8995a92a7f8b3a7d10db420daa35b08c
3
+ size 348273776
vmllm_cached/qwen3vit/image_full/train/dataset_info.json ADDED
@@ -0,0 +1,65 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "builder_name": "json",
3
+ "citation": "",
4
+ "config_name": "default",
5
+ "dataset_name": "json",
6
+ "dataset_size": 663687137,
7
+ "description": "",
8
+ "download_checksums": {
9
+ "/share/m2v_intern_v3/wangjunjie09/model_cache/VisionEncoder/ms-swift-data/image_sft_full_v800k_sharegpt.json": {
10
+ "num_bytes": 856061267,
11
+ "checksum": null
12
+ }
13
+ },
14
+ "download_size": 856061267,
15
+ "features": {
16
+ "messages": [
17
+ {
18
+ "content": {
19
+ "dtype": "string",
20
+ "_type": "Value"
21
+ },
22
+ "role": {
23
+ "dtype": "string",
24
+ "_type": "Value"
25
+ }
26
+ }
27
+ ],
28
+ "images": [
29
+ {
30
+ "bytes": {
31
+ "dtype": "null",
32
+ "_type": "Value"
33
+ },
34
+ "path": {
35
+ "dtype": "string",
36
+ "_type": "Value"
37
+ }
38
+ }
39
+ ],
40
+ "lengths": {
41
+ "feature": {
42
+ "dtype": "int64",
43
+ "_type": "Value"
44
+ },
45
+ "_type": "Sequence"
46
+ }
47
+ },
48
+ "homepage": "",
49
+ "license": "",
50
+ "size_in_bytes": 1519748404,
51
+ "splits": {
52
+ "train": {
53
+ "name": "train",
54
+ "num_bytes": 663687137,
55
+ "num_examples": 738590,
56
+ "dataset_name": "json"
57
+ }
58
+ },
59
+ "version": {
60
+ "version_str": "0.0.0",
61
+ "major": 0,
62
+ "minor": 0,
63
+ "patch": 0
64
+ }
65
+ }
vmllm_cached/qwen3vit/image_full/train/state.json ADDED
@@ -0,0 +1,16 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_data_files": [
3
+ {
4
+ "filename": "data-00000-of-00002.arrow"
5
+ },
6
+ {
7
+ "filename": "data-00001-of-00002.arrow"
8
+ }
9
+ ],
10
+ "_fingerprint": "8897283c77da70ab",
11
+ "_format_columns": null,
12
+ "_format_kwargs": {},
13
+ "_format_type": null,
14
+ "_output_all_columns": false,
15
+ "_split": null
16
+ }
vmllm_cached/qwen3vit/video_10pct/train/data-00000-of-00001.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cb3ccc9ed7d1ce2b304f7d545abb611b12457518489aad44732008430eb33e83
3
+ size 157175192
vmllm_cached/qwen3vit/video_10pct/train/dataset_info.json ADDED
@@ -0,0 +1,60 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "builder_name": "json",
3
+ "citation": "",
4
+ "config_name": "default",
5
+ "dataset_name": "json",
6
+ "dataset_size": 153484834,
7
+ "description": "",
8
+ "download_checksums": {
9
+ "/share/m2v_intern_v3/wangjunjie09/model_cache/VisionEncoder/ms-swift-data/video_sft_small_10pct_sharegpt.json": {
10
+ "num_bytes": 173305720,
11
+ "checksum": null
12
+ }
13
+ },
14
+ "download_size": 173305720,
15
+ "features": {
16
+ "messages": [
17
+ {
18
+ "content": {
19
+ "dtype": "string",
20
+ "_type": "Value"
21
+ },
22
+ "role": {
23
+ "dtype": "string",
24
+ "_type": "Value"
25
+ }
26
+ }
27
+ ],
28
+ "videos": {
29
+ "feature": {
30
+ "dtype": "string",
31
+ "_type": "Value"
32
+ },
33
+ "_type": "Sequence"
34
+ },
35
+ "lengths": {
36
+ "feature": {
37
+ "dtype": "int64",
38
+ "_type": "Value"
39
+ },
40
+ "_type": "Sequence"
41
+ }
42
+ },
43
+ "homepage": "",
44
+ "license": "",
45
+ "size_in_bytes": 326790554,
46
+ "splits": {
47
+ "train": {
48
+ "name": "train",
49
+ "num_bytes": 153484834,
50
+ "num_examples": 113615,
51
+ "dataset_name": "json"
52
+ }
53
+ },
54
+ "version": {
55
+ "version_str": "0.0.0",
56
+ "major": 0,
57
+ "minor": 0,
58
+ "patch": 0
59
+ }
60
+ }
vmllm_cached/qwen3vit/video_10pct/train/state.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_data_files": [
3
+ {
4
+ "filename": "data-00000-of-00001.arrow"
5
+ }
6
+ ],
7
+ "_fingerprint": "2700a2518587c720",
8
+ "_format_columns": null,
9
+ "_format_kwargs": {},
10
+ "_format_type": null,
11
+ "_output_all_columns": false,
12
+ "_split": null
13
+ }
vmllm_cached/qwen3vit/video_full/train/data-00000-of-00003.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1b88beff00c56b12f30267dba6013ff4a41c5b3a0b5cf0bc2192a2958e315fbf
3
+ size 354565576
vmllm_cached/qwen3vit/video_full/train/data-00001-of-00003.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aea108e61e829dbb4b25d191ac19b52f5e5fbea5bfc5c9669a88ef249d9f325c
3
+ size 354296560
vmllm_cached/qwen3vit/video_full/train/data-00002-of-00003.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0f12a9db322e5c1237684efe23b4612a2044b3ef2dbd74f96a3a447de8b24991
3
+ size 354842288
vmllm_cached/qwen3vit/video_full/train/dataset_info.json ADDED
@@ -0,0 +1,60 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "builder_name": "json",
3
+ "citation": "",
4
+ "config_name": "default",
5
+ "dataset_name": "json",
6
+ "dataset_size": 1037696374,
7
+ "description": "",
8
+ "download_checksums": {
9
+ "/share/m2v_intern_v3/wangjunjie09/model_cache/VisionEncoder/ms-swift-data/video_sft_full_v800k_sharegpt.json": {
10
+ "num_bytes": 1177312903,
11
+ "checksum": null
12
+ }
13
+ },
14
+ "download_size": 1177312903,
15
+ "features": {
16
+ "messages": [
17
+ {
18
+ "content": {
19
+ "dtype": "string",
20
+ "_type": "Value"
21
+ },
22
+ "role": {
23
+ "dtype": "string",
24
+ "_type": "Value"
25
+ }
26
+ }
27
+ ],
28
+ "videos": {
29
+ "feature": {
30
+ "dtype": "string",
31
+ "_type": "Value"
32
+ },
33
+ "_type": "Sequence"
34
+ },
35
+ "lengths": {
36
+ "feature": {
37
+ "dtype": "int64",
38
+ "_type": "Value"
39
+ },
40
+ "_type": "Sequence"
41
+ }
42
+ },
43
+ "homepage": "",
44
+ "license": "",
45
+ "size_in_bytes": 2215009277,
46
+ "splits": {
47
+ "train": {
48
+ "name": "train",
49
+ "num_bytes": 1037696374,
50
+ "num_examples": 800001,
51
+ "dataset_name": "json"
52
+ }
53
+ },
54
+ "version": {
55
+ "version_str": "0.0.0",
56
+ "major": 0,
57
+ "minor": 0,
58
+ "patch": 0
59
+ }
60
+ }
vmllm_cached/qwen3vit/video_full/train/state.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_data_files": [
3
+ {
4
+ "filename": "data-00000-of-00003.arrow"
5
+ },
6
+ {
7
+ "filename": "data-00001-of-00003.arrow"
8
+ },
9
+ {
10
+ "filename": "data-00002-of-00003.arrow"
11
+ }
12
+ ],
13
+ "_fingerprint": "4c41ef3382480e1b",
14
+ "_format_columns": null,
15
+ "_format_kwargs": {},
16
+ "_format_type": null,
17
+ "_output_all_columns": false,
18
+ "_split": null
19
+ }
vmllm_cached/qwenvit_v4_1/image_10pct/train/data-00000-of-00001.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4b30ec6ac4ecd10f6af680df3ddbba35278bd2a1834899de7aed1dbf5fd229ee
3
+ size 69352320
vmllm_cached/qwenvit_v4_1/image_10pct/train/dataset_info.json ADDED
@@ -0,0 +1,65 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "builder_name": "json",
3
+ "citation": "",
4
+ "config_name": "default",
5
+ "dataset_name": "json",
6
+ "dataset_size": 66130519,
7
+ "description": "",
8
+ "download_checksums": {
9
+ "/share/m2v_intern_v3/wangjunjie09/model_cache/VisionEncoder/ms-swift-data/image_sft_small_10pct_sharegpt.json": {
10
+ "num_bytes": 85235535,
11
+ "checksum": null
12
+ }
13
+ },
14
+ "download_size": 85235535,
15
+ "features": {
16
+ "messages": [
17
+ {
18
+ "content": {
19
+ "dtype": "string",
20
+ "_type": "Value"
21
+ },
22
+ "role": {
23
+ "dtype": "string",
24
+ "_type": "Value"
25
+ }
26
+ }
27
+ ],
28
+ "images": [
29
+ {
30
+ "bytes": {
31
+ "dtype": "null",
32
+ "_type": "Value"
33
+ },
34
+ "path": {
35
+ "dtype": "string",
36
+ "_type": "Value"
37
+ }
38
+ }
39
+ ],
40
+ "lengths": {
41
+ "feature": {
42
+ "dtype": "int64",
43
+ "_type": "Value"
44
+ },
45
+ "_type": "Sequence"
46
+ }
47
+ },
48
+ "homepage": "",
49
+ "license": "",
50
+ "size_in_bytes": 151366054,
51
+ "splits": {
52
+ "train": {
53
+ "name": "train",
54
+ "num_bytes": 66130519,
55
+ "num_examples": 73859,
56
+ "dataset_name": "json"
57
+ }
58
+ },
59
+ "version": {
60
+ "version_str": "0.0.0",
61
+ "major": 0,
62
+ "minor": 0,
63
+ "patch": 0
64
+ }
65
+ }
vmllm_cached/qwenvit_v4_1/image_10pct/train/state.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"_data_files": [{"filename": "data-00000-of-00001.arrow"}], "_fingerprint": "1ee927233338c5f1", "_format_columns": null, "_format_kwargs": {}, "_format_type": null, "_output_all_columns": false, "_split": null}
vmllm_cached/qwenvit_v4_1/image_10pct/train/v3_patch_meta.json ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "v3_patch_version": "v4_fast",
3
+ "mode": "image",
4
+ "from": "qwen3vit (Qwen3 ViT base)",
5
+ "delta_formula": "image: lengths[0] += 1 * len(images); video: lengths[0] += 16 (= 16 frames \u00d7 1) * len(videos)",
6
+ "input": "/share/m2v_intern_v3/wangjunjie09/VisionEncoder/data/vmllm_cached/qwen3vit/image_10pct/train",
7
+ "rows": 73859,
8
+ "lengths_sample_mean_old": 2624.35,
9
+ "lengths_sample_mean_new": 2625.35,
10
+ "timestamp": "2026-05-06T09:22:26"
11
+ }
vmllm_cached/qwenvit_v4_1/video_10pct/train/data-00000-of-00001.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ec2d9f8240d1319817225d0697d1a2e43eed9c1f022ddd38cd73fa0e802d86a9
3
+ size 157175192
vmllm_cached/qwenvit_v4_1/video_10pct/train/dataset_info.json ADDED
@@ -0,0 +1,60 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "builder_name": "json",
3
+ "citation": "",
4
+ "config_name": "default",
5
+ "dataset_name": "json",
6
+ "dataset_size": 153484834,
7
+ "description": "",
8
+ "download_checksums": {
9
+ "/share/m2v_intern_v3/wangjunjie09/model_cache/VisionEncoder/ms-swift-data/video_sft_small_10pct_sharegpt.json": {
10
+ "num_bytes": 173305720,
11
+ "checksum": null
12
+ }
13
+ },
14
+ "download_size": 173305720,
15
+ "features": {
16
+ "messages": [
17
+ {
18
+ "content": {
19
+ "dtype": "string",
20
+ "_type": "Value"
21
+ },
22
+ "role": {
23
+ "dtype": "string",
24
+ "_type": "Value"
25
+ }
26
+ }
27
+ ],
28
+ "videos": {
29
+ "feature": {
30
+ "dtype": "string",
31
+ "_type": "Value"
32
+ },
33
+ "_type": "Sequence"
34
+ },
35
+ "lengths": {
36
+ "feature": {
37
+ "dtype": "int64",
38
+ "_type": "Value"
39
+ },
40
+ "_type": "Sequence"
41
+ }
42
+ },
43
+ "homepage": "",
44
+ "license": "",
45
+ "size_in_bytes": 326790554,
46
+ "splits": {
47
+ "train": {
48
+ "name": "train",
49
+ "num_bytes": 153484834,
50
+ "num_examples": 113615,
51
+ "dataset_name": "json"
52
+ }
53
+ },
54
+ "version": {
55
+ "version_str": "0.0.0",
56
+ "major": 0,
57
+ "minor": 0,
58
+ "patch": 0
59
+ }
60
+ }
vmllm_cached/qwenvit_v4_1/video_10pct/train/state.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"_data_files": [{"filename": "data-00000-of-00001.arrow"}], "_fingerprint": "9c3dccfcc84579a5", "_format_columns": null, "_format_kwargs": {}, "_format_type": null, "_output_all_columns": false, "_split": null}
vmllm_cached/qwenvit_v4_1/video_10pct/train/v3_patch_meta.json ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "v3_patch_version": "v4_fast",
3
+ "mode": "video",
4
+ "from": "qwen3vit (Qwen3 ViT base)",
5
+ "delta_formula": "image: lengths[0] += 1 * len(images); video: lengths[0] += 16 (= 16 frames \u00d7 1) * len(videos)",
6
+ "input": "/share/m2v_intern_v3/wangjunjie09/VisionEncoder/data/vmllm_cached/qwen3vit/video_10pct/train",
7
+ "rows": 113615,
8
+ "lengths_sample_mean_old": 2568.1,
9
+ "lengths_sample_mean_new": 2584.1,
10
+ "timestamp": "2026-05-06T09:22:27"
11
+ }
vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00000_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6196ec385d9b4e7345034bf80c853e33850054fddca0993796946cbb6511d4ce
3
+ size 4348624
vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00001_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2abec3e42edcea0486bd90f8b09427102fa26553e897a556b289f981c071f482
3
+ size 4397128
vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00002_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a10416252759d5bed49dbacb73833a08a7bf3d4582feef0312d52b53f34f5316
3
+ size 4337304
vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00003_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6c249788d8710e351652e7d7b56c11ff6e843d06361d40da024b95fe431d6f79
3
+ size 4294744
vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00004_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1ae9e64e766bd13e39a466fecdd6fb29d85177df30cd3de0582061f037530ea8
3
+ size 4317096
vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00005_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:771c2d9c3297ff6fcd68224d32c93d6d953648963bec19146e11f839f6737ab6
3
+ size 4366920
vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00006_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2615d1d49510237a6b5668ebb113e3a5fcd3a1c4aa7fd47ee518eb0d60d8d32b
3
+ size 4356176
vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00007_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d65edab8224984e90875355653bbbc1ec5956d5968cc124131c446cb2ffde792
3
+ size 4407040
vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00008_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d3e76a89f1f7c158e7385f2993a90791d6177626649273984a6a085e0dd63bac
3
+ size 4347048
vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00009_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:602f2c58b4c39ca17d7ca19622e1971f218bd54177cbf15c45a9daf1725ed63c
3
+ size 4314704
vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00010_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f8d0fc9c580d736e3d17c41009f855aeb4c3bb81a3ceaef3ce33ee832bfc5c9e
3
+ size 4371096
vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00011_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b35f857402134fe8a314a3f61b698fd1c76db54529f934af00f101620f62b2f9
3
+ size 4369816
vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00012_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7bed79f4d3726d77649dc3ae1ba46c2cea36fe17d087b76df889e3f829a123bd
3
+ size 4378360
vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00013_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3b977b503d2d0c8730947f76b284daa697a3cfc31508c959e0f79ee48b77b5b7
3
+ size 4362504
vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00014_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a19a02543444662722d5a6971f2ba9a04995e9a82efe79cd8622ec4d9c48f4fb
3
+ size 4452368
vmllm_cached/siglip2/image_10pct/train/cache-293d2d32c7066ab2_00015_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:690cdb0c532447190c7ab6baf54e28f886cc84c7fb764ca1bb2c8bf5d3440c8e
3
+ size 4372144
vmllm_cached/siglip2/image_10pct/train/cache-2b2dbbebf1aceb79_00000_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2bdbbea67ca3da6665cbed5aafc75cffa0fc2489bfb3147bb33ab9eab3685a76
3
+ size 4348624
vmllm_cached/siglip2/image_10pct/train/cache-2b2dbbebf1aceb79_00001_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9947ccb4fbd7e237a47c62afdcba358fd1f70ccdafa2b75e0901adc04d54f110
3
+ size 4397128
vmllm_cached/siglip2/image_10pct/train/cache-2b2dbbebf1aceb79_00002_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:56ff06c350c38b02e83f34543cd2a4c4c9112a2e020bbca185ad8536f40411a6
3
+ size 4337304
vmllm_cached/siglip2/image_10pct/train/cache-2b2dbbebf1aceb79_00003_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9a02edb69066ef0eb0e3763102886a2c635a8d4cae4349c152f8f9e0a9742840
3
+ size 4294744
vmllm_cached/siglip2/image_10pct/train/cache-2b2dbbebf1aceb79_00004_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3933f3f36ec0a176758430e57d3cfa356d4e06aa0c50de03b94a71d591148fd0
3
+ size 4317096
vmllm_cached/siglip2/image_10pct/train/cache-2b2dbbebf1aceb79_00005_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9baa6891e3b6a8254d8a897694984b5c1314ca9cd403f0063c384d52b2a84307
3
+ size 4366920
vmllm_cached/siglip2/image_10pct/train/cache-2b2dbbebf1aceb79_00006_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6b67cbee268f6706ddec761821f9c4cf4639125fe2386920c6a19af06244546a
3
+ size 4356176
vmllm_cached/siglip2/image_10pct/train/cache-2b2dbbebf1aceb79_00007_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9b11634a8f510b85116d2b629a251588b5d8d79ba4ca33d49f9703bfa1d49f61
3
+ size 4407040
vmllm_cached/siglip2/image_10pct/train/cache-2b2dbbebf1aceb79_00008_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:98b7b1b323d4af133985be1062d316eb0a2cfdf975d8ebaf6a6f85c485a62c87
3
+ size 4347048
vmllm_cached/siglip2/image_10pct/train/cache-2b2dbbebf1aceb79_00009_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6e9e828b05b7a7d3aaf2e8a432003aefa9a9785ac82473de01d160dc63dd59ba
3
+ size 4314704
vmllm_cached/siglip2/image_10pct/train/cache-2b2dbbebf1aceb79_00010_of_00016.arrow ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e2bdd212e2bbb3add2c91f167be38f6a8014fecbb0b1b9bf77e056d4723f6077
3
+ size 4371096