添加chunk prefill模型和axcl算力卡脚本

.gitattributes

@@ -36,3 +36,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.axmodel filter=lfs diff=lfs merge=lfs -text
 main filter=lfs diff=lfs merge=lfs -text
 image/ssd_horse.jpg filter=lfs diff=lfs merge=lfs -text
+main_axcl_aarch64 filter=lfs diff=lfs merge=lfs -text
+main_axcl_x86 filter=lfs diff=lfs merge=lfs -text

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/Qwen2.5-VL-3B-Instruct_vision_nchw448.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:47c6a5c75e3941c49123018f352785dbcbd028dd7d1e741a16c6453f9c9209cf
+size 921254437

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/Qwen2.5-VL-3B-Instruct_vision_nhwc.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:035e9118aa3f106f35f6cf4caa6829a8c97695693b42aea009e43ae39e9b1a59
+size 777801587

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/model.embed_tokens.weight.bfloat16.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b84907567aa829c6f24cadbdeb68c3c44d25fc0a8be8e917fd603cb64f72810d
+size 622329856

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l0_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:51a79521d270e312062df608512706eaa3e259d97ac79514cba3461150a92496
+size 90748908

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l10_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cae58a2b7923870f93540409bd4a8f1c63398ebdaeb0d51905d1f59f2b5aa51f
+size 90755116

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l11_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:60cd4540069957200d622642c91015ff93744b2bc17a9e2494f8e1a3fb78a34d
+size 90756172

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l12_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8c4302238aeeb83ab4ef8f4317e2a4d39917532ba07c46d844a1540b0abac2fb
+size 90756268

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l13_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:34f1efba5e7bc27ddb8902e3c4347b881fdd3616e6ad7709f6234cdd81d0a931
+size 90756844

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l14_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6041968d265ecbe039719c0bb31b068d3c124fbda5501a231c0bc73b04806355
+size 90755628

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l15_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc59dd081c621b5f78a80665e5bb50708f665ce8df834b81bb959eaf68203aa6
+size 90754924

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l16_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3403d19c196f2a5a0f86c862b42e79034986edae8dce8cda526a9c58791bd0d0
+size 90756204

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l17_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:42638c3a6d26560aa078110bdd7b34221eaaf874d380d4afffcc5bf064137d9d
+size 90753068

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l18_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b3c87b2b8b35fa4a1c744a96f574c7ccd6672edbbcf28640bab8b015721b7c8c
+size 90755308

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l19_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9505a70f508794afc7b5c1a6524d227e8424b04cbaaec08d5676a637990da1f3
+size 90752876

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l1_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c2741d3af4b13c85f35e596b4c4af194aa489803007a1c4a92f93499a5fdec63
+size 90736556

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l20_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:847399a80afded39d556b153f53483b74b4e17308cb420df1b37b5e64a3c891a
+size 90751788

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l21_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c5d3e45a63ba393186a06184006aafcdea62129b414f5df8e86da40aa764cdf9
+size 90754380

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l22_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a8b09fc38f16380a3565fa430ae1b97a09c74e663cda85c40b1a49b737f4d1f
+size 90756364

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l23_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f736729cd081a16ccbd81d87fc420adf7612baa0a6b35e2f50557b8ccc44f966
+size 90752428

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l24_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d4c4571d43628a914526fcee5af0aa9ba37f8fb6bd1813b3257c49eb1243a2d
+size 90756204

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l25_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6eb156c97aab89cae01fe4e1e605c5fef025f9aa5e0f8cfd7c6e04c2ed06bbd3
+size 90754572

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l26_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a5755b87ecd176c77436079c32a19c8623ed015db200e7061c83327ce47db3d0
+size 90755116

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l27_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:35cc4ee732852c91b2fe9d63d158adbed81e780522459bc70ef3075a5b898de2
+size 90752364

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l28_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7a337e42d0b12ee92bbeecbe9b3e5d7e4580a6e5e9d56ee05147606e2dd4840
+size 90754476

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l29_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0357c4ab39ee73fc410ee703df3fb14f53a75dae04555c4747879af343be7d16
+size 90756588

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l2_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:637456ba1f3a34cfecb227b2c6f1497fcf2e06819ef17d21968fb90972862486
+size 90740012

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l30_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4a9126537341353c715539cab61dc5104cae6fccf3c2e1f63879e1f5b534b3e7
+size 90755180

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l31_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:edbb9a2f10db51dc636ba1690455e7cd36c38cfee0acdea6906705db675a21d9
+size 90753260

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l32_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:823d29dfe1d3ab990d00cd34979faa5b7b02232576f679785bfe9e63e5db3d8c
+size 90754508

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l33_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dc35da5cf2958dfab6ceab4f9125055a8f2f3f4bc06f4eb19f4c682f3a1ead49
+size 90756492

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l34_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:29a39b7f7595a0ab9366a8ea7c7f02fd23689cca03906e86ade07de7a71648a3
+size 90757612

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l35_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f578c09e0e5570af1807c95076f2f4da430ce094c01a90409a1801c43e4c6c97
+size 90756812

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l3_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c19d25d84398b51fcb06b6450e36b8272bd71a02e7cf98368fcec9386608c3cd
+size 90748940

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l4_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc453a3db86348b7ff30029dadb00f7b050aa19b7c705b836d616521806375d9
+size 90747148

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l5_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3fed937c6a11c7a561a5b53f357d2e21c8693b04b5799f10e78d7ce65ab036db
+size 90753676

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l6_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:901709229028a2537434fb6b9649466b0a667118e8536b72bd9751496ef1d486
+size 90754828

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l7_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d01a13ede4f8af89726c63087c598f9e225139a2ddba5e016347385ddf1ce2ee
+size 90756524

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l8_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90d9eb43359555d0fe4813e29aa8a17380b1cbba99f686ce5b7c5fbdc6635933
+size 90757132

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_p128_l9_together.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:67f8801cef3bd233b806acdb373598efc493cdf5f1e364944dce9b35fcd46a13
+size 90756908

Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512/qwen2_5_vl_post.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:87c2abcce5f1701f751c57848fa50cdce35e06a84d6df12816291e1216fcd1af
+size 339965928

README.md

@@ -62,6 +62,8 @@ root@ax650:/mnt/qtang/llm-test/qwen2.5-vl-3b# tree -L 2
 ├── image
 │   └── ssd_car.jpg
 ├── main
+├── main_axcl_x86
+├── main_axcl_aarch64
 ├── python
 │   ├── cv_resize.py
 │   ├── infer_image.py
@@ -97,6 +99,10 @@ root@ax650:/mnt/qtang/llm-test/qwen2.5-vl-3b# tree -L 2
 ├── qwen2_tokenizer_video_308.py
 ├── run_qwen2_5_vl_image.sh
 ├── run_qwen2_5_vl_video.sh
+├── run_qwen2_5_vl_image_axcl_x86.sh
+├── run_qwen2_5_vl_image_axcl_aarch64.sh
+├── run_qwen2_5_vl_video_axcl_x86.sh
+├── run_qwen2_5_vl_video_axcl_aarch64.sh
 └── video
     ├── frame_0075.jpg
 ......
@@ -210,4 +216,119 @@ video/frame_0056.jpg
 #### Inference with M.2 Accelerator card
 What is M.2 Accelerator card?, Show this DEMO based on Raspberry PI 5.
 
-TODO
+#### Image understand demo
+
+##### start tokenizer server for image understand demo
+
+```
+python3 qwen2_tokenizer_image_448.py --port 12345
+```
+
+##### run image understand demo
+
+- input text
+
+```
+描述这张图片
+```
+
+- input image
+
+![](./image/ssd_car.jpg)
+
+```
+(base) axera@raspberrypi:~/lhj/Qwen2.5-VL-3B-Instruct $ bash run_qwen2_5_vl_image_axcl_aarch64.sh 
+[I][                            Init][ 162]: LLM init start
+[I][                            Init][  34]: connect http://127.0.0.1:12345 ok
+[I][                            Init][ 267]: IMAGE_CONTEXT_TOKEN: 151655, IMAGE_START_TOKEN: 151652
+[I][                            Init][ 328]: image encoder output float32
+
+[I][                            Init][ 340]: max_token_len : 1023
+[I][                            Init][ 343]: kv_cache_size : 256, kv_cache_num: 1023
+[I][                            Init][ 351]: prefill_token_num : 128
+[I][                            Init][ 355]: grp: 1, prefill_max_token_num : 1
+[I][                            Init][ 355]: grp: 2, prefill_max_token_num : 128
+[I][                            Init][ 355]: grp: 3, prefill_max_token_num : 256
+[I][                            Init][ 355]: grp: 4, prefill_max_token_num : 384
+[I][                            Init][ 355]: grp: 5, prefill_max_token_num : 512
+[I][                            Init][ 359]: prefill_max_token_num : 512
+________________________
+|    ID| remain cmm(MB)|
+========================
+|     0|           2286|
+¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯
+[E][                     load_config][ 278]: config file(post_config.json) open failed
+[W][                            Init][ 452]: load postprocess config(post_config.json) failed
+[I][                            Init][ 456]: LLM init ok
+Type "q" to exit, Ctrl+c to stop current running
+prompt >> 描述这张图片
+image >> image/ssd_car.jpg
+[I][                          Encode][ 539]: image encode time : 772.851990 ms, size : 524288
+[I][                             Run][ 625]: input token num : 280, prefill_split_num : 3
+[I][                             Run][ 659]: input_num_token:128
+[I][                             Run][ 659]: input_num_token:128
+[I][                             Run][ 659]: input_num_token:24
+[I][                             Run][ 796]: ttft: 2067.18 ms
+这张图片展示了一条繁忙的城市街道。前景中，一名女子站在人行道上，穿着黑色外套，面带微笑。她旁边是一辆红色的双层巴士，巴士上有一个广告，上面写着“THINGS GET MORE EXITING WHEN YOU SAY ‘YES’ VirginMoney.co.uk”。巴士的车牌号是“L15”。巴士旁边停着一辆黑色的面包车。背景中可以看到一些商店和行人，街道两旁有路灯和商店的招牌。整体环境显得非常繁忙和现代。
+
+[N][                             Run][ 949]: hit eos,avg 4.12 token/s
+```
+
+#### Video understand demo
+
+Please pre-process the image of the video file into a 308x308 size picture
+
+##### start tokenizer server for image understand demo
+
+```
+python qwen2_tokenizer_video_308.py --port 12345
+```
+
+##### run image understand demo
+
+```
+(base) axera@raspberrypi:~/lhj/Qwen2.5-VL-3B-Instruct $ bash run_qwen2_5_vl_video_axcl_aarch64.sh 
+[I][                            Init][ 162]: LLM init start
+[I][                            Init][  34]: connect http://127.0.0.1:12345 ok
+[I][                            Init][ 267]: IMAGE_CONTEXT_TOKEN: 151656, IMAGE_START_TOKEN: 151652
+[I][                            Init][ 328]: image encoder output float32
+
+[I][                            Init][ 340]: max_token_len : 1023
+[I][                            Init][ 343]: kv_cache_size : 256, kv_cache_num: 1023
+[I][                            Init][ 351]: prefill_token_num : 128
+[I][                            Init][ 355]: grp: 1, prefill_max_token_num : 1
+[I][                            Init][ 355]: grp: 2, prefill_max_token_num : 128
+[I][                            Init][ 355]: grp: 3, prefill_max_token_num : 256
+[I][                            Init][ 355]: grp: 4, prefill_max_token_num : 384
+[I][                            Init][ 355]: grp: 5, prefill_max_token_num : 512
+[I][                            Init][ 359]: prefill_max_token_num : 512
+________________________
+|    ID| remain cmm(MB)|
+========================
+|     0|           2464|
+¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯
+[E][                     load_config][ 278]: config file(post_config.json) open failed
+[W][                            Init][ 452]: load postprocess config(post_config.json) failed
+[I][                            Init][ 456]: LLM init ok
+Type "q" to exit, Ctrl+c to stop current running
+prompt >> 描述这个视频的内容
+image >> video
+video/frame_0000.jpg
+video/frame_0008.jpg
+video/frame_0016.jpg
+video/frame_0024.jpg
+video/frame_0032.jpg
+video/frame_0040.jpg
+video/frame_0048.jpg
+video/frame_0056.jpg
+[I][                          Encode][ 539]: image encode time : 1481.107056 ms, size : 991232
+[I][                             Run][ 625]: input token num : 509, prefill_split_num : 4
+[I][                             Run][ 659]: input_num_token:128
+[I][                             Run][ 659]: input_num_token:128
+[I][                             Run][ 659]: input_num_token:128
+[I][                             Run][ 659]: input_num_token:125
+[I][                             Run][ 796]: ttft: 3049.59 ms
+视频展示了两只松鼠在户外的场景。背景是模糊的山脉和蓝天，前景中有松鼠在互动。松鼠的毛色是棕色和灰色的混合，它们的爪子是橙色的。松鼠似乎在互相玩耍或争抢，它们的爪子和嘴巴都伸向对方。整个场景显得非常自然和生动。
+
+[N][                             Run][ 949]: hit eos,avg 4.15 token/s
+```

qwen2_tokenizer_image_448.py

@@ -87,10 +87,6 @@ class Tokenizer_Http():
 
         # official implementation 
         text = f'<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n<|im_start|>user\n<|vision_start|>' + '<|image_pad|>' * 256 + f'<|vision_end|>{content}<|im_end|>\n<|im_start|>assistant\n'
-
-        # better for quantation model
-        # text = f'<|im_start|>user\n{content}<|im_end|>\n<|im_start|>user\n<|vision_start|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|vision_end|><|im_end|>\n<|im_start|>assistant\n'
-
         
         output_kwargs = {'text_kwargs': {'padding': True, 'return_tensors': 'pt'}, 'images_kwargs': {'return_tensors': 'pt'}, 'audio_kwargs': {'padding': True, 'return_tensors': 'pt'}, 'videos_kwargs': {'fps': 2.0, 'return_tensors': 'pt'}, 'common_kwargs': {'return_tensors': 'pt'}}
         
@@ -117,6 +113,13 @@ class Tokenizer_Http():
     def eos_token(self):
         return self.tokenizer.eos_token
 
+    @property
+    def img_start_token(self):
+        return self.tokenizer.encode("<|vision_start|>")[0]
+
+    @property
+    def img_context_token(self):
+        return self.tokenizer.encode("<|image_pad|>")[0]
 
 tokenizer = Tokenizer_Http()
 
@@ -164,6 +167,18 @@ class Request(BaseHTTPRequestHandler):
                 msg = json.dumps({'eos_id': -1})
             else:
                 msg = json.dumps({'eos_id': eos_id})
+        elif self.path == '/img_start_token':
+            img_start_token = tokenizer.img_start_token
+            if img_start_token is None:
+                msg = json.dumps({'img_start_token': -1})
+            else:
+                msg = json.dumps({'img_start_token': img_start_token})
+        elif self.path == '/img_context_token':
+            img_context_token = tokenizer.img_context_token
+            if img_context_token is None:
+                msg = json.dumps({'img_context_token': -1})
+            else:
+                msg = json.dumps({'img_context_token': img_context_token})
         else:
             msg = 'error'
 

qwen2_tokenizer_video_308.py

@@ -4,6 +4,69 @@ from http.server import HTTPServer, BaseHTTPRequestHandler
 import json
 import argparse
 
+def _prompt_split_image(
+    image_seq_len,
+    image_rows,
+    image_cols,
+    fake_token_around_image,
+    image_token,
+    global_img_token,
+):
+    """Prompt with expanded image tokens for when the image is split into patches."""
+    text_split_images = ""
+    for n_h in range(image_rows):
+        for n_w in range(image_cols):
+            text_split_images += (
+                f"{fake_token_around_image}"
+                + f"<row_{n_h + 1}_col_{n_w + 1}>"
+                + f"{image_token}" * image_seq_len
+            )
+        text_split_images += "\n"
+
+    text_split_images += (
+        f"\n{fake_token_around_image}"
+        + f"{global_img_token}"
+        + f"{image_token}" * image_seq_len
+        + f"{fake_token_around_image}"
+    )
+    return text_split_images
+
+
+def _prompt_single_image(
+    image_seq_len, fake_token_around_image, image_token, global_img_token
+):
+    """Prompt with expanded image tokens for a single image."""
+    return (
+        f"{fake_token_around_image}"
+        + f"{global_img_token}"
+        + f"{image_token}" * image_seq_len
+        + f"{fake_token_around_image}"
+    )
+
+
+def get_image_prompt_string(
+    image_rows,
+    image_cols,
+    image_seq_len,
+    fake_token_around_image,
+    image_token,
+    global_img_token,
+):
+    if image_rows == 0 and image_cols == 0:
+        return _prompt_single_image(
+            image_seq_len,
+            fake_token_around_image=fake_token_around_image,
+            image_token=image_token,
+            global_img_token=global_img_token,
+        )
+    return _prompt_split_image(
+        image_seq_len,
+        image_rows,
+        image_cols,
+        fake_token_around_image,
+        image_token,
+        global_img_token,
+    )
 
 class Tokenizer_Http():
 
@@ -24,9 +87,7 @@ class Tokenizer_Http():
         # official implementation 
         text = f"<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n<|im_start|>user\n<|vision_start|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|vision_end|>{content}<|im_end|>\n<|im_start|>assistant\n"
 
-        # better for quantation model
-        # text = f"<|im_start|>user\n{content}<|im_end|>\n<|im_start|>user\n<|vision_start|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|video_pad|><|vision_end|><|im_end|>\n<|im_start|>assistant\n"
-       
+           
         output_kwargs = {'text_kwargs': {'padding': True, 'return_tensors': 'pt'}, 'images_kwargs': {'return_tensors': 'pt'}, 'audio_kwargs': {'padding': True, 'return_tensors': 'pt'}, 'videos_kwargs': {'return_tensors': 'pt'}, 'common_kwargs': {'return_tensors': 'pt'}}
         
         text_inputs = self.tokenizer(text, **output_kwargs["text_kwargs"])
@@ -52,6 +113,13 @@ class Tokenizer_Http():
     def eos_token(self):
         return self.tokenizer.eos_token
 
+    @property
+    def img_start_token(self):
+        return self.tokenizer.encode("<|vision_start|>")[0]
+
+    @property
+    def img_context_token(self):
+        return self.tokenizer.encode("<|video_pad|>")[0]
 
 tokenizer = Tokenizer_Http()
 
@@ -99,6 +167,18 @@ class Request(BaseHTTPRequestHandler):
                 msg = json.dumps({'eos_id': -1})
             else:
                 msg = json.dumps({'eos_id': eos_id})
+        elif self.path == '/img_start_token':
+            img_start_token = tokenizer.img_start_token
+            if img_start_token is None:
+                msg = json.dumps({'img_start_token': -1})
+            else:
+                msg = json.dumps({'img_start_token': img_start_token})
+        elif self.path == '/img_context_token':
+            img_context_token = tokenizer.img_context_token
+            if img_context_token is None:
+                msg = json.dumps({'img_context_token': -1})
+            else:
+                msg = json.dumps({'img_context_token': img_context_token})
         else:
             msg = 'error'
 

run_qwen2_5_vl_image_axcl_aarch64.sh

@@ -0,0 +1,18 @@
+AXMODEL_DIR=./Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512
+
+./main_axcl_aarch64 \
+--template_filename_axmodel "${AXMODEL_DIR}/qwen2_5_vl_p128_l%d_together.axmodel" \
+--axmodel_num 36 \
+--filename_image_encoder_axmodedl "${AXMODEL_DIR}/Qwen2.5-VL-3B-Instruct_vision_nchw448.axmodel" \
+--use_mmap_load_embed 1 \
+--filename_tokenizer_model "http://127.0.0.1:12345" \
+--filename_post_axmodel "${AXMODEL_DIR}/qwen2_5_vl_post.axmodel" \
+--filename_tokens_embed "${AXMODEL_DIR}/model.embed_tokens.weight.bfloat16.bin" \
+--tokens_embed_num 151936 \
+--tokens_embed_size 2048 \
+--live_print 1 \
+--img_width 448 \
+--img_height 448 \
+--vision_start_token_id 151652 \
+--post_config_path post_config.json \
+--devices 0,

run_qwen2_5_vl_image_axcl_x86.sh

@@ -0,0 +1,18 @@
+AXMODEL_DIR=./Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512
+
+./main_axcl_x86 \
+--template_filename_axmodel "${AXMODEL_DIR}/qwen2_5_vl_p128_l%d_together.axmodel" \
+--axmodel_num 36 \
+--filename_image_encoder_axmodedl "${AXMODEL_DIR}/Qwen2.5-VL-3B-Instruct_vision_nchw448.axmodel" \
+--use_mmap_load_embed 1 \
+--filename_tokenizer_model "http://127.0.0.1:12345" \
+--filename_post_axmodel "${AXMODEL_DIR}/qwen2_5_vl_post.axmodel" \
+--filename_tokens_embed "${AXMODEL_DIR}/model.embed_tokens.weight.bfloat16.bin" \
+--tokens_embed_num 151936 \
+--tokens_embed_size 2048 \
+--live_print 1 \
+--img_width 448 \
+--img_height 448 \
+--vision_start_token_id 151652 \
+--post_config_path post_config.json \
+--devices 0,

run_qwen2_5_vl_video_axcl_aarch64.sh

@@ -0,0 +1,18 @@
+AXMODEL_DIR=./Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512
+
+./main_axcl_aarch64 \
+--template_filename_axmodel "${AXMODEL_DIR}/qwen2_5_vl_p128_l%d_together.axmodel" \
+--axmodel_num 36 \
+--filename_image_encoder_axmodedl "${AXMODEL_DIR}/Qwen2.5-VL-3B-Instruct_vision_nhwc.axmodel" \
+--use_mmap_load_embed 1 \
+--filename_tokenizer_model "http://127.0.0.1:12345" \
+--filename_post_axmodel "${AXMODEL_DIR}/qwen2_5_vl_post.axmodel" \
+--filename_tokens_embed "${AXMODEL_DIR}/model.embed_tokens.weight.bfloat16.bin" \
+--tokens_embed_num 151936 \
+--tokens_embed_size 2048 \
+--live_print 1 \
+--img_width 308 \
+--img_height 308 \
+--vision_start_token_id 151652 \
+--post_config_path post_config.json \
+--devices 0,

run_qwen2_5_vl_video_axcl_x86.sh

@@ -0,0 +1,18 @@
+AXMODEL_DIR=./Qwen2.5-VL-3B-Instruct-AX650-chunk_prefill_512
+
+./main_axcl_x86 \
+--template_filename_axmodel "${AXMODEL_DIR}/qwen2_5_vl_p128_l%d_together.axmodel" \
+--axmodel_num 36 \
+--filename_image_encoder_axmodedl "${AXMODEL_DIR}/Qwen2.5-VL-3B-Instruct_vision_nhwc.axmodel" \
+--use_mmap_load_embed 1 \
+--filename_tokenizer_model "http://127.0.0.1:12345" \
+--filename_post_axmodel "${AXMODEL_DIR}/qwen2_5_vl_post.axmodel" \
+--filename_tokens_embed "${AXMODEL_DIR}/model.embed_tokens.weight.bfloat16.bin" \
+--tokens_embed_num 151936 \
+--tokens_embed_size 2048 \
+--live_print 1 \
+--img_width 308 \
+--img_height 308 \
+--vision_start_token_id 151652 \
+--post_config_path post_config.json \
+--devices 0,