Hugging Face's logo

AXERA-TECH
/
Qwen3-VL-8B-Instruct-GPTQ-Int4

Image-Text-to-Text
Transformers
Safetensors
English
Chinese
Qwen3-VL
Qwen3-VL-2B-Instruct
Qwen3-VL-4B-Instruct
Qwen3-VL-8B-Instruct
Int4
VLM
GPTQ
Model card Files
xet
Community
lihongjie commited on
Commit
457735b
·
1 Parent(s): e2ec202

first commit

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. .gitattributes +59 -0
  2. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/Qwen3-VL-8B-Instruct_vision.axmodel +3 -0
  3. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/model.embed_tokens.weight.bfloat16.bin +3 -0
  4. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l0_together.axmodel +3 -0
  5. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l10_together.axmodel +3 -0
  6. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l11_together.axmodel +3 -0
  7. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l12_together.axmodel +3 -0
  8. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l13_together.axmodel +3 -0
  9. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l14_together.axmodel +3 -0
  10. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l15_together.axmodel +3 -0
  11. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l16_together.axmodel +3 -0
  12. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l17_together.axmodel +3 -0
  13. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l18_together.axmodel +3 -0
  14. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l19_together.axmodel +3 -0
  15. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l1_together.axmodel +3 -0
  16. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l20_together.axmodel +3 -0
  17. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l21_together.axmodel +3 -0
  18. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l22_together.axmodel +3 -0
  19. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l23_together.axmodel +3 -0
  20. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l24_together.axmodel +3 -0
  21. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l25_together.axmodel +3 -0
  22. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l26_together.axmodel +3 -0
  23. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l27_together.axmodel +3 -0
  24. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l28_together.axmodel +3 -0
  25. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l29_together.axmodel +3 -0
  26. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l2_together.axmodel +3 -0
  27. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l30_together.axmodel +3 -0
  28. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l31_together.axmodel +3 -0
  29. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l32_together.axmodel +3 -0
  30. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l33_together.axmodel +3 -0
  31. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l34_together.axmodel +3 -0
  32. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l35_together.axmodel +3 -0
  33. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l3_together.axmodel +3 -0
  34. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l4_together.axmodel +3 -0
  35. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l5_together.axmodel +3 -0
  36. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l6_together.axmodel +3 -0
  37. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l7_together.axmodel +3 -0
  38. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l8_together.axmodel +3 -0
  39. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l9_together.axmodel +3 -0
  40. Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_post.axmodel +3 -0
  41. README.md +233 -3
  42. config.json +0 -0
  43. images/demo.jpg +3 -0
  44. images/demo1.jpg +3 -0
  45. images/demo_720p.jpg +3 -0
  46. images/recoAll_attractions_1.jpg +3 -0
  47. images/recoAll_attractions_2.jpg +3 -0
  48. images/recoAll_attractions_3.jpg +3 -0
  49. images/recoAll_attractions_4.jpg +3 -0
  50. images/ssd_car.jpg +3 -0
.gitattributes CHANGED
@@ -33,3 +33,62 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/Qwen3-VL-8B-Instruct_vision.axmodel filter=lfs diff=lfs merge=lfs -text
37
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l15_together.axmodel filter=lfs diff=lfs merge=lfs -text
38
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l16_together.axmodel filter=lfs diff=lfs merge=lfs -text
39
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l31_together.axmodel filter=lfs diff=lfs merge=lfs -text
40
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l30_together.axmodel filter=lfs diff=lfs merge=lfs -text
41
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l35_together.axmodel filter=lfs diff=lfs merge=lfs -text
42
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l6_together.axmodel filter=lfs diff=lfs merge=lfs -text
43
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l9_together.axmodel filter=lfs diff=lfs merge=lfs -text
44
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_post.axmodel filter=lfs diff=lfs merge=lfs -text
45
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l14_together.axmodel filter=lfs diff=lfs merge=lfs -text
46
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l19_together.axmodel filter=lfs diff=lfs merge=lfs -text
47
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l23_together.axmodel filter=lfs diff=lfs merge=lfs -text
48
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l25_together.axmodel filter=lfs diff=lfs merge=lfs -text
49
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l20_together.axmodel filter=lfs diff=lfs merge=lfs -text
50
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l27_together.axmodel filter=lfs diff=lfs merge=lfs -text
51
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l3_together.axmodel filter=lfs diff=lfs merge=lfs -text
52
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l5_together.axmodel filter=lfs diff=lfs merge=lfs -text
53
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l12_together.axmodel filter=lfs diff=lfs merge=lfs -text
54
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l21_together.axmodel filter=lfs diff=lfs merge=lfs -text
55
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l2_together.axmodel filter=lfs diff=lfs merge=lfs -text
56
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/model.embed_tokens.weight.bfloat16.bin filter=lfs diff=lfs merge=lfs -text
57
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l13_together.axmodel filter=lfs diff=lfs merge=lfs -text
58
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l33_together.axmodel filter=lfs diff=lfs merge=lfs -text
59
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l4_together.axmodel filter=lfs diff=lfs merge=lfs -text
60
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l28_together.axmodel filter=lfs diff=lfs merge=lfs -text
61
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l29_together.axmodel filter=lfs diff=lfs merge=lfs -text
62
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l32_together.axmodel filter=lfs diff=lfs merge=lfs -text
63
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l10_together.axmodel filter=lfs diff=lfs merge=lfs -text
64
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l11_together.axmodel filter=lfs diff=lfs merge=lfs -text
65
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l17_together.axmodel filter=lfs diff=lfs merge=lfs -text
66
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l18_together.axmodel filter=lfs diff=lfs merge=lfs -text
67
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l26_together.axmodel filter=lfs diff=lfs merge=lfs -text
68
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l34_together.axmodel filter=lfs diff=lfs merge=lfs -text
69
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l7_together.axmodel filter=lfs diff=lfs merge=lfs -text
70
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l8_together.axmodel filter=lfs diff=lfs merge=lfs -text
71
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l0_together.axmodel filter=lfs diff=lfs merge=lfs -text
72
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l1_together.axmodel filter=lfs diff=lfs merge=lfs -text
73
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l22_together.axmodel filter=lfs diff=lfs merge=lfs -text
74
+ Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l24_together.axmodel filter=lfs diff=lfs merge=lfs -text
75
+ images/recoAll_attractions_4.jpg filter=lfs diff=lfs merge=lfs -text
76
+ video/frame_0008.jpg filter=lfs diff=lfs merge=lfs -text
77
+ video/frame_0056.jpg filter=lfs diff=lfs merge=lfs -text
78
+ images/recoAll_attractions_3.jpg filter=lfs diff=lfs merge=lfs -text
79
+ images/demo_720p.jpg filter=lfs diff=lfs merge=lfs -text
80
+ images/recoAll_attractions_2.jpg filter=lfs diff=lfs merge=lfs -text
81
+ images/ssd_horse.jpg filter=lfs diff=lfs merge=lfs -text
82
+ video/frame_0024.jpg filter=lfs diff=lfs merge=lfs -text
83
+ video/frame_0040.jpg filter=lfs diff=lfs merge=lfs -text
84
+ video/frame_0048.jpg filter=lfs diff=lfs merge=lfs -text
85
+ images/demo1.jpg filter=lfs diff=lfs merge=lfs -text
86
+ images/ssd_car.jpg filter=lfs diff=lfs merge=lfs -text
87
+ video/frame_0000.jpg filter=lfs diff=lfs merge=lfs -text
88
+ video/frame_0032.jpg filter=lfs diff=lfs merge=lfs -text
89
+ images/demo.jpg filter=lfs diff=lfs merge=lfs -text
90
+ video/frame_0016.jpg filter=lfs diff=lfs merge=lfs -text
91
+ images/recoAll_attractions_1.jpg filter=lfs diff=lfs merge=lfs -text
92
+ main_ax650 filter=lfs diff=lfs merge=lfs -text
93
+ main_axcl_aarch64 filter=lfs diff=lfs merge=lfs -text
94
+ main_axcl_x86 filter=lfs diff=lfs merge=lfs -text
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/Qwen3-VL-8B-Instruct_vision.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dd6d98cb2bfb3f65992135f353226d3973040584083146142c78c0861cc1d0b9
3
+ size 650854262
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/model.embed_tokens.weight.bfloat16.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d6d8b73339dcddb4ecf93c0741c319af21e7f1c0f2a224b31a7cc7b0d42045c0
3
+ size 1244659712
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l0_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:843aba18020bc242bb45359070ee12c0901985e6e3cee1db347c1a2c61d16941
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l10_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:46fec0abbf0d9c2aae13cf76bd2a5cc81fcbd49004bf823b0d54de4a38654978
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l11_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f1969d07f6124801386f507e1a87fbe7cda98e5016b3dcbdc74047c895408b66
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l12_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:638fef60af71245a84fcc9ee3e0532b108e3deb87159a4218fa8c2235d9eef50
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l13_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:23d602a1e952d6f462e6265493d681f2b9ead24cbccf5d451b407464069f3f9f
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l14_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0e190549f788c538c7f1ae1287862ebf1f148721816629e57f2b48d6e6eed486
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l15_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:53dceff05902e2b8fc376c6fd9dc51c5166ebc5c2395974d0fe7177d6b4c947a
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l16_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1bd5acd49e86121a06de4e4ef0737c7291fa917239cc86964edb22150646ac78
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l17_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3556f46c5a54610ad1ec4bde19b7f8d3a34fd4ad8c2bc04f382d011a0e60e82f
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l18_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d7c693af46df214116c47f5291ab804367dc214583b205acfc7761d3a48e822
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l19_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:05753c2a53845407b33f238dd4e6a16a4aec84e2c6ccae9b39d0a0e25128be3e
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l1_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c7a84d3784633cefa3e5a990fe5aebea01da5d496b6b0c4c29b34d70e786c243
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l20_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:786b99dcfaaf304b7f10e812f95500bd01494786acd3be7d8cac06a55ad3328c
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l21_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:54ef71f1591cf758e378fd86cab4c66cae944234a63a1c0b07ae47563d562797
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l22_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7583f27994defe5dded092bbe0e1d52920e4678cd8f63f5d1e4cd7fed9adf984
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l23_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:84057fb752f65b050bd52c6fab3dde19feade03074a3e0758fbb3f0d129aa294
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l24_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1a2d3688e2274c4baf532f9b7831c2fb81d1eb873ac51e5322de687bcd15985f
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l25_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:291c3fe0900dfb5bd148615e8329476226efe6d5ca2bff4b0a1ea31c1a651649
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l26_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:01ce20e98516cd907d64b535c00a7baa91a9ce6cb75366a5e235276b1ed2afa8
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l27_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1a8276974573e202b4436d343fa4a960a42254836d74ba4e9dc6f1b3fa305dcb
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l28_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eb8d8712aa696b754613bf650b4466e3079ec5633db910c01a577a4a23a06b49
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l29_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:502686793f056cd25191c275267e7002ec29d8b173db8804ee574c87f7da2be1
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l2_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9c926cb0a0ad538d76478aa412c999aa33f3081d2cac26aa5312faa7aa6df208
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l30_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe67cdefa9d3fb38d01ab79bd04e7c478d463292dd2870b570bcfdfef288d973
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l31_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:13fba1148e4f3ffe3d0afc485457e35a2386b142561f72cb7dd4eaad9b814ac1
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l32_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e8e42c7dbb3bd7ba16f0bdbc0b210553f98a2d608e3d4d61f5656a2aca7546ac
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l33_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0f2963270d6f15d961bea8ad3a98d1c2bcd9b00bf6bf3b0ae533518c56800048
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l34_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:afa240ea9dd8584ef82f4f35ee0ed11a793b18ce9e492fe473445cf0ab33e60f
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l35_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:82ed4581df5a1f30422498c6fc071207324335fe26e78c452363376b612a0679
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l3_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0f5ea9f5a2d598123e9b6579514ea90fe9f29f97f97ed39099f68cf1ad92673b
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l4_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bfb1f23a5373ee93ba526b27c4ae2d9ae38e58157bf8a9f960b105ec97d65aee
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l5_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e38d17a17969cb3a8fd52e67e3d3d385e49c8b5c0466a4ff357c4de2a6275d62
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l6_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:43fe3585cd081e72aa3736879d03e6f836359e0bf91eb3e945ab522dc402a586
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l7_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ba6dda4f8eb84af4d54a9fa4fea2eee5e0a547a00c4cfd0237afac1cc26e3db8
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l8_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d7b74e7aa3431e629d7b5ce6cbcd27a7b16d478bba1eaa0def5d4fd43c5e191c
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_p128_l9_together.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5a24cf32e166085c47aab1a98649bbe910b25b811b84aba62f54ea2b9a3c115e
3
+ size 139406559
Qwen3-VL-8B-Instruct-AX650-c128_p1152-int4/qwen3_vl_text_post.axmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:94f37c2ba1d950f97fb791103f80b2817eb6e6ee50798597841cc23f6669badd
3
+ size 678989672
README.md CHANGED
@@ -1,3 +1,233 @@
1
- ---
2
- license: mit
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: mit
3
+ language:
4
+ - en
5
+ - zh
6
+ base_model:
7
+ - Qwen/Qwen3-VL-2B-Instruct
8
+ - Qwen/Qwen3-VL-4B-Instruct
9
+ - Qwen/Qwen3-VL-8B-Instruct
10
+ pipeline_tag: image-text-to-text
11
+ library_name: transformers
12
+ tags:
13
+ - Qwen3-VL
14
+ - Qwen3-VL-2B-Instruct
15
+ - Qwen3-VL-4B-Instruct
16
+ - Qwen3-VL-8B-Instruct
17
+ - Int4
18
+ - VLM
19
+ - GPTQ
20
+ ---
21
+
22
+ # Qwen3-VL
23
+
24
+ This version of Qwen3-VL-2B-Instruct has been converted to run on the Axera NPU using **w4a16** quantization.
25
+
26
+ Compatible with Pulsar2 version: 5.0
27
+
28
+ ## Convert tools links:
29
+
30
+ For those who are interested in model conversion, you can try to export axmodel through the original repo :
31
+
32
+ - https://huggingface.co/Qwen/Qwen3-VL-2B-Instruct
33
+ - https://huggingface.co/Qwen/Qwen3-VL-4B-Instruct
34
+
35
+ [Pulsar2 Link, How to Convert LLM from Huggingface to axmodel](https://pulsar2-docs.readthedocs.io/en/latest/appendix/build_llm.html)
36
+
37
+ [AXera NPU HOST LLM Runtime](https://github.com/AXERA-TECH/Qwen3-VL.AXERA)
38
+
39
+
40
+ ## Support Platform
41
+
42
+ - AX650
43
+ - AX650N DEMO Board
44
+ - [M4N-Dock(爱芯派Pro)](https://wiki.sipeed.com/hardware/zh/maixIV/m4ndock/m4ndock.html)
45
+ - [M.2 Accelerator card](https://axcl-docs.readthedocs.io/zh-cn/latest/doc_guide_hardware.html)
46
+
47
+ **Image Process**
48
+ |Chips| input size | image num | image encoder | ttft(168 tokens) | w4a16 | CMM | Flash |
49
+ |--|--|--|--|--|--|--|--|
50
+ |AX650| 384*384 | 1 | ms | ms | tokens/sec| GiB | 9.1 GiB |
51
+
52
+ **Video Process**
53
+ |Chips| input size | image num | image encoder |ttft(600 tokens) | w4a16 | CMM | Flash |
54
+ |--|--|--|--|--|--|--|--|
55
+ |AX650| 384*384 | 8 | ms | ms | tokens/sec| GiB | 9.1 GiB |
56
+
57
+ The DDR capacity refers to the CMM memory that needs to be consumed. Ensure that the CMM memory allocation on the development board is greater than this value.
58
+
59
+ ## How to use
60
+
61
+ Download all files from this repository to the device
62
+
63
+ **If you using AX650 Board**
64
+
65
+ ### Prepare tokenizer server
66
+
67
+ #### Install transformer
68
+
69
+ ```
70
+ pip install -r requirements.txt
71
+ ```
72
+
73
+ ### Demo Run
74
+
75
+ #### Image understand demo
76
+
77
+ ##### start tokenizer server for image understand demo
78
+
79
+ ```
80
+ python3 tokenizer_images.py --port 8080
81
+ ```
82
+
83
+ ##### run image understand demo
84
+
85
+ - input text
86
+
87
+ ```
88
+ 描述这张图片
89
+ ```
90
+
91
+ - input image
92
+
93
+ ![](./images/recoAll_attractions_1.jpg)
94
+
95
+ ```
96
+ root@ax650 ~/Qwen3-VL-2B-Instruct-GPTQ-Int4 # bash run_image_ax650.sh
97
+ [I][ Init][ 156]: LLM init start
98
+ [I][ Init][ 158]: Total CMM:4353 MB
99
+ [I][ Init][ 34]: connect http://127.0.0.1:8080 ok
100
+ bos_id: -1, eos_id: 151645
101
+ img_start_token: 151652
102
+ img_context_token: 151655
103
+ 3% | ██ | 1 / 31 [0.01s<0.46s, 66.67 count/s] tokenizer init ok[I][ Init][ 26]: LLaMaEmbedSelector use mmap
104
+ 6% | ███ | 2 / 31 [0.02s<0.34s, 90.91 count/s] embed_selector init ok[I][ Init][ 201]: attr.axmodel_num:28
105
+ 103% | ██████████████████████████████████ | 32 / 31 [34.03s<32.96s, 0.94 count/s] init vpm axmodel ok,remain_cmm(854 MB)[I][ Init][ 266]: IMAGE_CONTEXT_TOKEN: 151655, IMAGE_START_TOKEN: 151652
106
+ [I][ Init][ 309]: image encoder output float32
107
+
108
+ [I][ Init][ 339]: max_token_len : 2047
109
+ [I][ Init][ 344]: kv_cache_size : 1024, kv_cache_num: 2047
110
+ [I][ Init][ 352]: prefill_token_num : 128
111
+ [I][ Init][ 356]: grp: 1, prefill_max_token_num : 1
112
+ [I][ Init][ 356]: grp: 2, prefill_max_token_num : 128
113
+ [I][ Init][ 356]: grp: 3, prefill_max_token_num : 256
114
+ [I][ Init][ 356]: grp: 4, prefill_max_token_num : 384
115
+ [I][ Init][ 356]: grp: 5, prefill_max_token_num : 512
116
+ [I][ Init][ 356]: grp: 6, prefill_max_token_num : 640
117
+ [I][ Init][ 356]: grp: 7, prefill_max_token_num : 768
118
+ [I][ Init][ 356]: grp: 8, prefill_max_token_num : 896
119
+ [I][ Init][ 356]: grp: 9, prefill_max_token_num : 1024
120
+ [I][ Init][ 356]: grp: 10, prefill_max_token_num : 1152
121
+ [I][ Init][ 360]: prefill_max_token_num : 1152
122
+ [I][ Init][ 372]: LLM init ok
123
+ [I][ Init][ 374]: Left CMM:854 MB
124
+ Type "q" to exit, Ctrl+c to stop current running
125
+ prompt >> 描述这张图片
126
+ image >> images/recoAll_attractions_1.jpg
127
+ [I][ EncodeImage][ 440]: pixel_values size 1
128
+ [I][ EncodeImage][ 441]: grid_h 24 grid_w 24
129
+ [I][ EncodeImage][ 489]: image encode time : 237.778000 ms, size : 1
130
+ [I][ Encode][ 532]: input_ids size:168
131
+ [I][ Encode][ 540]: offset 15
132
+ [I][ Encode][ 569]: img_embed.size:1, 294912
133
+ [I][ Encode][ 583]: out_embed size:344064
134
+ [I][ Encode][ 584]: input_ids size 168
135
+ [I][ Encode][ 586]: position_ids size:168
136
+ [I][ Run][ 607]: input token num : 168, prefill_split_num : 2
137
+ [I][ Run][ 641]: input_num_token:128
138
+ [I][ Run][ 641]: input_num_token:40
139
+ [I][ Run][ 865]: ttft: 313.60 ms
140
+ 这是一张在埃及沙漠中拍摄的风景照片。画面中，三座巨大的金字塔在晴朗的天空下矗立，它们是古埃及文明的象征。这些金字塔由巨大的石块堆叠而成，表面因岁月侵蚀而显得斑驳。在金字塔的前方，有几个人影在沙地上行走，这为整个场景提供了比例感和尺度感。整个场景充满了历史的厚重感和神秘的氛围。
141
+
142
+ [N][ Run][ 992]: hit eos,avg 14.14 token/s
143
+ ```
144
+
145
+ #### Video understand demo
146
+
147
+ ##### start tokenizer server for image understand demo
148
+
149
+ ```
150
+ python tokenizer_video.py --port 8080
151
+ ```
152
+
153
+ ##### run video understand demo
154
+ - input text
155
+
156
+ ```
157
+ 描述这个视频
158
+ ```
159
+
160
+ - input video
161
+
162
+ ./video
163
+
164
+ ```
165
+ root@ax650 ~/Qwen3-VL-2B-Instruct-GPTQ-Int4 # bash run_video_ax650.sh
166
+ [I][ Init][ 156]: LLM init start
167
+ [I][ Init][ 158]: Total CMM:7884 MB
168
+ [I][ Init][ 34]: connect http://127.0.0.1:8080 ok
169
+ bos_id: -1, eos_id: 151645
170
+ img_start_token: 151652
171
+ img_context_token: 151656
172
+ 3% | ██ | 1 / 31 [0.01s<0.34s, 90.91 count/s] tokenizer init ok[I][ Init][ 26]: LLaMaEmbedSelector use mmap
173
+ 6% | ███ | 2 / 31 [0.01s<0.23s, 133.33 count/s] embed_selector init ok[I][ Init][ 201]: attr.axmodel_num:28
174
+ 103% | ██████████████████████████████████ | 32 / 31 [32.37s<31.36s, 0.99 count/s] init vpm axmodel ok,remain_cmm(4385 MB)[I][ Init][ 266]: IMAGE_CONTEXT_TOKEN: 151656, IMAGE_START_TOKEN: 151652
175
+ [I][ Init][ 309]: image encoder output float32
176
+
177
+ [I][ Init][ 339]: max_token_len : 2047
178
+ [I][ Init][ 344]: kv_cache_size : 1024, kv_cache_num: 2047
179
+ [I][ Init][ 352]: prefill_token_num : 128
180
+ [I][ Init][ 356]: grp: 1, prefill_max_token_num : 1
181
+ [I][ Init][ 356]: grp: 2, prefill_max_token_num : 128
182
+ [I][ Init][ 356]: grp: 3, prefill_max_token_num : 256
183
+ [I][ Init][ 356]: grp: 4, prefill_max_token_num : 384
184
+ [I][ Init][ 356]: grp: 5, prefill_max_token_num : 512
185
+ [I][ Init][ 356]: grp: 6, prefill_max_token_num : 640
186
+ [I][ Init][ 356]: grp: 7, prefill_max_token_num : 768
187
+ [I][ Init][ 356]: grp: 8, prefill_max_token_num : 896
188
+ [I][ Init][ 356]: grp: 9, prefill_max_token_num : 1024
189
+ [I][ Init][ 356]: grp: 10, prefill_max_token_num : 1152
190
+ [I][ Init][ 360]: prefill_max_token_num : 1152
191
+ [I][ Init][ 372]: LLM init ok
192
+ [I][ Init][ 374]: Left CMM:4385 MB
193
+ Type "q" to exit, Ctrl+c to stop current running
194
+ prompt >> 描述这个视频
195
+ video >> video
196
+ video/frame_0000.jpg
197
+ video/frame_0008.jpg
198
+ video/frame_0016.jpg
199
+ video/frame_0024.jpg
200
+ video/frame_0032.jpg
201
+ video/frame_0040.jpg
202
+ video/frame_0048.jpg
203
+ video/frame_0056.jpg
204
+ [I][ EncodeImage][ 440]: pixel_values size 4
205
+ [I][ EncodeImage][ 441]: grid_h 24 grid_w 24
206
+ [I][ EncodeImage][ 489]: image encode time : 751.481018 ms, size : 4
207
+ [I][ Encode][ 532]: input_ids size:600
208
+ [I][ Encode][ 540]: offset 15
209
+ [I][ Encode][ 569]: img_embed.size:4, 294912
210
+ [I][ Encode][ 574]: offset:159
211
+ [I][ Encode][ 574]: offset:303
212
+ [I][ Encode][ 574]: offset:447
213
+ [I][ Encode][ 583]: out_embed size:1228800
214
+ [I][ Encode][ 584]: input_ids size 600
215
+ [I][ Encode][ 586]: position_ids size:600
216
+ [I][ Run][ 607]: input token num : 600, prefill_split_num : 5
217
+ [I][ Run][ 641]: input_num_token:128
218
+ [I][ Run][ 641]: input_num_token:128
219
+ [I][ Run][ 641]: input_num_token:128
220
+ [I][ Run][ 641]: input_num_token:128
221
+ [I][ Run][ 641]: input_num_token:88
222
+ [I][ Run][ 865]: ttft: 843.36 ms
223
+ 这是一段关于两只山地旱獭（也称“山地土拨鼠”）在山地环境中互动的视频。
224
+
225
+ 在画面中，两只山地旱獭正站在布满碎石的山坡上，背景是连绵起伏的山脉和蓝天。它们的毛色以灰、棕、黑相间，脸部和耳朵周围有明显的黑白条纹，显得非常可爱。
226
+
227
+ 这两只旱獭正在进行一场激烈的“拳击”或“格斗”游戏。它们的前爪高高举起，像在互相击打，但它们的姿势和动作表明它们可能是在进行一场激烈的“拳击”或“格斗”游戏。它们的嘴巴和前爪在空中挥舞，似乎在互相攻击或展示力量。
228
+
229
+ 整个场景充满了动感和活力，展现了这些小动物在自然环境中充满活力和趣味的一面。
230
+
231
+ [N][ Run][ 992]: hit eos,avg 14.16 token/s
232
+
233
+ ```
config.json ADDED
File without changes
images/demo.jpg ADDED

Git LFS Details

  • SHA256: 344d5f0e43bfd6a4a6ed655dc9c3dc76a2a6ecb1de3afeee998e61722378e8b3
  • Pointer size: 130 Bytes
  • Size of remote file: 64.5 kB
images/demo1.jpg ADDED

Git LFS Details

  • SHA256: 8d6156cbaa86cf0f4a9ebbfc8c2bf307dcf6fc64e32176d69333f6f7c7b294c5
  • Pointer size: 131 Bytes
  • Size of remote file: 101 kB
images/demo_720p.jpg ADDED

Git LFS Details

  • SHA256: 48215026377aaf674f2e0f771f02136962a8e914012a3fa49fd1edf25e022352
  • Pointer size: 130 Bytes
  • Size of remote file: 73.3 kB
images/recoAll_attractions_1.jpg ADDED

Git LFS Details

  • SHA256: 0072430513e76580c4134b78e452a1fb729112fe5725d1f8481e697c7b5cd4a1
  • Pointer size: 130 Bytes
  • Size of remote file: 73.3 kB
images/recoAll_attractions_2.jpg ADDED

Git LFS Details

  • SHA256: 28acba6c284d06039ca57c9ce182094baafa8a48b49068dc44482a6d643689cc
  • Pointer size: 131 Bytes
  • Size of remote file: 105 kB
images/recoAll_attractions_3.jpg ADDED

Git LFS Details

  • SHA256: f308b3ebb1855b74768d9a24a256616f683ba59e52c238326501f4008a3ab9a4
  • Pointer size: 130 Bytes
  • Size of remote file: 59.4 kB
images/recoAll_attractions_4.jpg ADDED

Git LFS Details

  • SHA256: e47e13a69c16073e221b143a43263fcbe956bcfeb8ad47fddbdbc03a2b6f7261
  • Pointer size: 131 Bytes
  • Size of remote file: 115 kB
images/ssd_car.jpg ADDED

Git LFS Details

  • SHA256: 92d459a39a9eef03956257cf9fec84114d9e5df8fb9c0662fb257488cdd4f365
  • Pointer size: 130 Bytes
  • Size of remote file: 50.5 kB