Hugging Face's logo

AXERA-TECH
/
Qwen3-4B

Text Generation
Transformers
English
Qwen
Qwen3
Int8
Model card Files
xet
 Community
wangli commited on
Commit
d3bf9ef
·
verified ·
1 Parent(s): 7d2783c

update axmodel for better ttft

Browse files
Files changed (38) hide show
  1. README.md +117 -79
  2. qwen3_p128_l0_together.axmodel +2 -2
  3. qwen3_p128_l10_together.axmodel +2 -2
  4. qwen3_p128_l11_together.axmodel +2 -2
  5. qwen3_p128_l12_together.axmodel +2 -2
  6. qwen3_p128_l13_together.axmodel +2 -2
  7. qwen3_p128_l14_together.axmodel +2 -2
  8. qwen3_p128_l15_together.axmodel +2 -2
  9. qwen3_p128_l16_together.axmodel +2 -2
  10. qwen3_p128_l17_together.axmodel +2 -2
  11. qwen3_p128_l18_together.axmodel +2 -2
  12. qwen3_p128_l19_together.axmodel +2 -2
  13. qwen3_p128_l1_together.axmodel +2 -2
  14. qwen3_p128_l20_together.axmodel +2 -2
  15. qwen3_p128_l21_together.axmodel +2 -2
  16. qwen3_p128_l22_together.axmodel +2 -2
  17. qwen3_p128_l23_together.axmodel +2 -2
  18. qwen3_p128_l24_together.axmodel +2 -2
  19. qwen3_p128_l25_together.axmodel +2 -2
  20. qwen3_p128_l26_together.axmodel +2 -2
  21. qwen3_p128_l27_together.axmodel +2 -2
  22. qwen3_p128_l28_together.axmodel +2 -2
  23. qwen3_p128_l29_together.axmodel +2 -2
  24. qwen3_p128_l2_together.axmodel +2 -2
  25. qwen3_p128_l30_together.axmodel +2 -2
  26. qwen3_p128_l31_together.axmodel +2 -2
  27. qwen3_p128_l32_together.axmodel +2 -2
  28. qwen3_p128_l33_together.axmodel +2 -2
  29. qwen3_p128_l34_together.axmodel +2 -2
  30. qwen3_p128_l35_together.axmodel +2 -2
  31. qwen3_p128_l3_together.axmodel +2 -2
  32. qwen3_p128_l4_together.axmodel +2 -2
  33. qwen3_p128_l5_together.axmodel +2 -2
  34. qwen3_p128_l6_together.axmodel +2 -2
  35. qwen3_p128_l7_together.axmodel +2 -2
  36. qwen3_p128_l8_together.axmodel +2 -2
  37. qwen3_p128_l9_together.axmodel +2 -2
  38. qwen3_post.axmodel +1 -1
README.md CHANGED
@@ -5,6 +5,7 @@ language:
5
  base_model:
6
  - Qwen/Qwen3-4B
7
  pipeline_tag: text-generation
 
8
  tags:
9
  - Qwen
10
  - Qwen3
@@ -14,11 +15,11 @@ tags:
14
 
15
  # Qwen3-4B
16
 
17
- This version of Qwen3-4B-Int8 has been converted to run on the Axera NPU using **w8a16** quantization.
18
 
19
  This model has been optimized with the following LoRA:
20
 
21
- Compatible with Pulsar2 version: 4.2(Not released yet)
22
 
23
  ## Convert tools links:
24
 
@@ -29,6 +30,17 @@ https://huggingface.co/Qwen/Qwen3-4B
29
 
30
  [AXera NPU LLM Runtime](https://github.com/AXERA-TECH/ax-llm)
31
 
 
 
 
 
 
 
 
 
 
 
 
32
 
33
  ## Support Platform
34
 
@@ -36,9 +48,9 @@ https://huggingface.co/Qwen/Qwen3-4B
36
  - [M4N-Dock(爱芯派Pro)](https://wiki.sipeed.com/hardware/zh/maixIV/m4ndock/m4ndock.html)
37
  - [M.2 Accelerator card](https://axcl-docs.readthedocs.io/zh-cn/latest/doc_guide_hardware.html)
38
 
39
- |Chips|w8a16|w4a16|
40
- |--|--|--|
41
- |AX650| 4.5 tokens/sec|TBD|
42
 
43
  ## How to use
44
 
@@ -99,91 +111,118 @@ hf download AXERA-TECH/Qwen3-4B --local-dir .
99
 
100
  ```shell
101
  (base) root@ax650:~# axllm run AXERA-TECH/Qwen3-4B/
102
- [I][ Init][ 127]: LLM init start
103
  tokenizer_type = 1
104
- 97% | ████████████████████████████████ | 38 / 39 [13.13s<13.47s, 2.89 count/s] init post axmodel ok,remain_cmm(5175 MB)
105
- [I][ Init][ 188]: max_token_len : 1023
106
- [I][ Init][ 191]: kv_cache_size : 1024, kv_cache_num: 1023
107
- [I][ Init][ 194]: prefill_token_num : 128
108
- [I][ Init][ 198]: grp: 1, prefill_max_kv_cache_num : 1
109
- [I][ Init][ 198]: grp: 2, prefill_max_kv_cache_num : 128
110
- [I][ Init][ 198]: grp: 3, prefill_max_kv_cache_num : 256
111
- [I][ Init][ 198]: grp: 4, prefill_max_kv_cache_num : 384
112
- [I][ Init][ 198]: grp: 5, prefill_max_kv_cache_num : 512
113
- [I][ Init][ 203]: prefill_max_token_num : 512
114
- [I][ Init][ 27]: LLaMaEmbedSelector use mmap
115
- 100% | ████████████████████████████████ | 39 / 39 [13.13s<13.13s, 2.97 count/s] embed_selector init ok
116
- [I][ load_config][ 282]: load config:
117
- {
118
- "enable_repetition_penalty": false,
119
- "enable_temperature": false,
120
- "enable_top_k_sampling": true,
121
- "enable_top_p_sampling": false,
122
- "penalty_window": 20,
123
- "repetition_penalty": 1.2,
124
- "temperature": 0.9,
125
- "top_k": 1,
126
- "top_p": 0.8
127
- }
128
-
129
- [I][ Init][ 224]: LLM init ok
130
- Type "q" to exit
131
- Ctrl+c to stop current running
132
- "reset" to reset kvcache
133
- "dd" to remove last conversation.
134
- "pp" to print history.
 
 
 
 
 
135
  ----------------------------------------
136
  prompt >> who are you
137
- [I][ SetKVCache][ 357]: prefill_grpid:2 kv_cache_num:128 precompute_len:0 input_num_token:22
138
- [I][ SetKVCache][ 359]: current prefill_max_token_num:512
139
- [I][ SetKVCache][ 360]: first run
140
- [I][ Run][ 412]: input token num : 22, prefill_split_num : 1
141
- [I][ Run][ 474]: ttft: 910.64 ms
 
 
142
  <think>
143
- Okay, the user asked, "who are you?" I need to respond in a friendly and informative way. Let me start by introducing myself clearly. I should mention that I'm Qwen, developed by Alibaba Cloud. It's important to highlight my capabilities, like answering questions, creating content, and helping with various tasks. I should also note that I can communicate in multiple languages. Maybe add something about being a helpful assistant. Keep it concise but cover the key points. Let me check if there's anything else important to include. Oh, and make sure the tone is approachable and not too technical. Alright, that should cover it.
144
  </think>
145
 
146
- Hello! I'm Qwen, a large language model developed by Alibaba Cloud. I can help with answering questions, creating content, and assisting with various tasks. I can communicate in multiple languages and am designed to be helpful and friendly. How can I assist you today? 😊
147
-
148
- [N][ Run][ 554]: hit eos,avg 4.36 token/s
149
 
150
- [I][ GetKVCache][ 331]: precompute_len:211, remaining:301
151
- prompt >> q
 
152
  ```
153
 
154
  ### 启动服务（OpenAI 兼容）
155
 
156
  ```shell
157
  (base) root@ax650:~# axllm serve AXERA-TECH/Qwen3-4B/
158
- [I][ Init][ 127]: LLM init start
159
  tokenizer_type = 1
160
- 97% | ████████████████████████████████ | 38 / 39 [9.17s<9.41s, 4.15 count/s] init post axmodel ok,remain_cmm(5175 MB)
161
- [I][ Init][ 188]: max_token_len : 1023
162
- [I][ Init][ 191]: kv_cache_size : 1024, kv_cache_num: 1023
163
- [I][ Init][ 194]: prefill_token_num : 128
164
- [I][ Init][ 198]: grp: 1, prefill_max_kv_cache_num : 1
165
- [I][ Init][ 198]: grp: 2, prefill_max_kv_cache_num : 128
166
- [I][ Init][ 198]: grp: 3, prefill_max_kv_cache_num : 256
167
- [I][ Init][ 198]: grp: 4, prefill_max_kv_cache_num : 384
168
- [I][ Init][ 198]: grp: 5, prefill_max_kv_cache_num : 512
169
- [I][ Init][ 203]: prefill_max_token_num : 512
170
- [I][ Init][ 27]: LLaMaEmbedSelector use mmap
171
- 100% | ████████████████████████████████ | 39 / 39 [9.17s<9.17s, 4.25 count/s] embed_selector init ok
172
- [I][ load_config][ 282]: load config:
173
- {
174
- "enable_repetition_penalty": false,
175
- "enable_temperature": false,
176
- "enable_top_k_sampling": true,
177
- "enable_top_p_sampling": false,
178
- "penalty_window": 20,
179
- "repetition_penalty": 1.2,
180
- "temperature": 0.9,
181
- "top_k": 1,
182
- "top_p": 0.8
183
- }
184
-
185
- [I][ Init][ 224]: LLM init ok
 
 
 
 
186
  Starting server on port 8000 with model 'AXERA-TECH/Qwen3-4B'...
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
187
  OpenAI API Server starting on http://0.0.0.0:8000
188
  Max concurrency: 1
189
  Models: AXERA-TECH/Qwen3-4B
@@ -195,7 +234,7 @@ Models: AXERA-TECH/Qwen3-4B
195
  from openai import OpenAI
196
 
197
  API_URL = "http://127.0.0.1:8000/v1"
198
- MODEL = "AXERA-TECH/Qwen3-1.7B"
199
 
200
  messages = [
201
  {"role": "system", "content": [{"type": "text", "text": "you are a helpful assistant."}]},
@@ -218,7 +257,7 @@ print(completion.choices[0].message.content)
218
  from openai import OpenAI
219
 
220
  API_URL = "http://127.0.0.1:8000/v1"
221
- MODEL = "AXERA-TECH/Qwen3-1.7B"
222
 
223
  messages = [
224
  {"role": "system", "content": [{"type": "text", "text": "you are a helpful assistant."}]},
@@ -237,7 +276,6 @@ for ev in stream:
237
  delta = getattr(ev.choices[0], "delta", None)
238
  if delta and getattr(delta, "content", None):
239
  print(delta.content, end="", flush=True)
240
- print("
241
- ")
242
  ```
243
 
 
5
  base_model:
6
  - Qwen/Qwen3-4B
7
  pipeline_tag: text-generation
8
+ library_name: transformers
9
  tags:
10
  - Qwen
11
  - Qwen3
 
15
 
16
  # Qwen3-4B
17
 
18
+ This version of Qwen3-4B has been converted to run on the Axera NPU using **w8a16** quantization.
19
 
20
  This model has been optimized with the following LoRA:
21
 
22
+ Compatible with Pulsar2 version: 5.2
23
 
24
  ## Convert tools links:
25
 
 
30
 
31
  [AXera NPU LLM Runtime](https://github.com/AXERA-TECH/ax-llm)
32
 
33
+ Convert the original Huggingface Qwen3-4B to axmodel, and then apply the w8a16 quantization to get the final axmodel for axllm runtime.
34
+
35
+ ```shell
36
+ export FLOAT_MATMUL_USE_CONV_EU=1 # only support AX650, for better performance, please set this env var before running the conversion command.
37
+
38
+ # context window size 2048, prefill length 1024
39
+ pulsar2 llm_build --input_path Qwen3-4B --output_path <your path> \
40
+ --hidden_state_type bf16 --kv_cache_len 2048 --prefill_len 128 --chip AX650 -c 1 --parallel 32 \
41
+ --last_kv_cache_len 128 --last_kv_cache_len 256 --last_kv_cache_len 384 --last_kv_cache_len 512 \
42
+ --last_kv_cache_len 640 --last_kv_cache_len 768 --last_kv_cache_len 896 --last_kv_cache_len 1024 -w s8
43
+ ```
44
 
45
  ## Support Platform
46
 
 
48
  - [M4N-Dock(爱芯派Pro)](https://wiki.sipeed.com/hardware/zh/maixIV/m4ndock/m4ndock.html)
49
  - [M.2 Accelerator card](https://axcl-docs.readthedocs.io/zh-cn/latest/doc_guide_hardware.html)
50
 
51
+ |Chips| w8a16 | CMM | Flash |
52
+ |--|--|--|--|
53
+ |AX650| 4.01 tokens/sec| 5.1 GiB| 5.3GiB |
54
 
55
  ## How to use
56
 
 
111
 
112
  ```shell
113
  (base) root@ax650:~# axllm run AXERA-TECH/Qwen3-4B/
114
+ 15:25:51.178 INF Init:890 | LLM init start
115
  tokenizer_type = 1
116
+ 97% | ############################### | 38 / 39 [17.72s<18.19s, 2.14 count/s] init post axmodel ok,remain_cmm(4744 MB)
117
+ 15:26:08.897 INF Init:1045 | max_token_len : 2048
118
+ 15:26:08.897 INF Init:1048 | kv_cache_size : 1024, kv_cache_num: 2048
119
+ 15:26:08.897 INF init_groups_from_model:606 | prefill_token_num : 128
120
+ 15:26:08.897 INF init_groups_from_model:820 | decode grp: 0, gid: 0, max_token_len : 2048
121
+ 15:26:08.897 INF init_groups_from_model:824 | prefill grp: 0, gid: 1, history_cap: 0, total_cap: 128, symbolic_cap: 1
122
+ 15:26:08.897 INF init_groups_from_model:824 | prefill grp: 1, gid: 2, history_cap: 128, total_cap: 256, symbolic_cap: 128
123
+ 15:26:08.897 INF init_groups_from_model:824 | prefill grp: 2, gid: 3, history_cap: 256, total_cap: 384, symbolic_cap: 256
124
+ 15:26:08.897 INF init_groups_from_model:824 | prefill grp: 3, gid: 4, history_cap: 384, total_cap: 512, symbolic_cap: 384
125
+ 15:26:08.897 INF init_groups_from_model:824 | prefill grp: 4, gid: 5, history_cap: 512, total_cap: 640, symbolic_cap: 512
126
+ 15:26:08.897 INF init_groups_from_model:824 | prefill grp: 5, gid: 6, history_cap: 640, total_cap: 768, symbolic_cap: 640
127
+ 15:26:08.897 INF init_groups_from_model:824 | prefill grp: 6, gid: 7, history_cap: 768, total_cap: 896, symbolic_cap: 768
128
+ 15:26:08.897 INF init_groups_from_model:824 | prefill grp: 7, gid: 8, history_cap: 896, total_cap: 1024, symbolic_cap: 896
129
+ 15:26:08.897 INF init_groups_from_model:824 | prefill grp: 8, gid: 9, history_cap: 1024, total_cap: 1152, symbolic_cap: 1024
130
+ 15:26:08.897 INF init_groups_from_model:831 | prefill_max_token_num : 1152
131
+ 15:26:08.897 INF Init:27 | LLaMaEmbedSelector use mmap
132
+ 100% | ################################ | 39 / 39 [17.72s<17.72s, 2.20 count/s] embed_selector init ok
133
+ 15:26:08.898 INF load_config:282 | load config:
134
+ 15:26:08.898 INF load_config:282 | {
135
+ 15:26:08.898 INF load_config:282 | "enable_repetition_penalty": false,
136
+ 15:26:08.898 INF load_config:282 | "enable_temperature": false,
137
+ 15:26:08.898 INF load_config:282 | "enable_top_k_sampling": false,
138
+ 15:26:08.898 INF load_config:282 | "enable_top_p_sampling": false,
139
+ 15:26:08.898 INF load_config:282 | "penalty_window": 20,
140
+ 15:26:08.898 INF load_config:282 | "repetition_penalty": 1.2,
141
+ 15:26:08.898 INF load_config:282 | "temperature": 0.9,
142
+ 15:26:08.898 INF load_config:282 | "top_k": 10,
143
+ 15:26:08.898 INF load_config:282 | "top_p": 0.8
144
+ 15:26:08.898 INF load_config:282 | }
145
+ 15:26:08.898 INF Init:1139 | LLM init ok
146
+ Commands:
147
+ /q, /exit 退出
148
+ /reset 重置 kvcache
149
+ /dd 删除一轮对话
150
+ /pp 打印历史对话
151
+ Ctrl+C: 停止当前生成
152
  ----------------------------------------
153
  prompt >> who are you
154
+ 15:26:15.337 INF SetKVCache:1437 | decode_grpid:0 prefill_grpid:1 history_cap:0 total_cap:128 symbolic_cap:1 precompute_len:0 input_num_token:22 prefer_symbolic_group:0
155
+ 15:26:15.337 INF SetKVCache:1458 | current prefill_max_token_num:1152
156
+ 15:26:15.460 INF SetKVCache:1462 | first run
157
+ 15:26:15.469 INF Run:1553 | input token num : 22, prefill_split_num : 1
158
+ 15:26:15.470 INF Run:1640 | prefill chunk p=0 history_len=0 grpid=1 kv_cache_num=0 input_tokens=22
159
+ 15:26:15.470 INF Run:1665 | prefill indices shape: p=0 idx_elems=128 idx_rows=1 pos_rows=0
160
+ 15:26:15.908 INF Run:1837 | ttft: 438.28 ms
161
  <think>
162
+ Okay, the user asked, "who are you?" I need to respond appropriately. First, I should introduce myself clearly. I'm Qwen, a large-scale language model developed by Alibaba Cloud. I should mention my capabilities, like answering questions, creating content, and helping with tasks. Also, I should highlight my multilingual support and the fact that I'm designed to be helpful and friendly. I should keep the tone positive and open for further assistance. Let me make sure the response is concise and covers all the key points without being too technical. Alright, that should do it.
163
  </think>
164
 
165
+ Hello! I'm Qwen, a large-scale language model developed by Alibaba Cloud. I can help with answering questions, creating content, and assisting with various tasks. I support multiple languages and am designed to be helpful and friendly. How can I assist you today? 😊
 
 
166
 
167
+ 15:26:59.884 NTC Run:2102 | hit eos,decode avg 4.00 token/s
168
+ 15:26:59.884 INF GetKVCache:1408 | precompute_len:199, remaining:953
169
+ prompt >> /q
170
  ```
171
 
172
  ### 启动服务（OpenAI 兼容）
173
 
174
  ```shell
175
  (base) root@ax650:~# axllm serve AXERA-TECH/Qwen3-4B/
176
+ 15:31:10.226 INF Init:890 | LLM init start
177
  tokenizer_type = 1
178
+ 97% | ############################### | 38 / 39 [13.45s<13.80s, 2.83 count/s] init post axmodel ok,remain_cmm(4744 MB)
179
+ 15:31:23.673 INF Init:1045 | max_token_len : 2048
180
+ 15:31:23.673 INF Init:1048 | kv_cache_size : 1024, kv_cache_num: 2048
181
+ 15:31:23.673 INF init_groups_from_model:606 | prefill_token_num : 128
182
+ 15:31:23.673 INF init_groups_from_model:820 | decode grp: 0, gid: 0, max_token_len : 2048
183
+ 15:31:23.673 INF init_groups_from_model:824 | prefill grp: 0, gid: 1, history_cap: 0, total_cap: 128, symbolic_cap: 1
184
+ 15:31:23.673 INF init_groups_from_model:824 | prefill grp: 1, gid: 2, history_cap: 128, total_cap: 256, symbolic_cap: 128
185
+ 15:31:23.673 INF init_groups_from_model:824 | prefill grp: 2, gid: 3, history_cap: 256, total_cap: 384, symbolic_cap: 256
186
+ 15:31:23.673 INF init_groups_from_model:824 | prefill grp: 3, gid: 4, history_cap: 384, total_cap: 512, symbolic_cap: 384
187
+ 15:31:23.673 INF init_groups_from_model:824 | prefill grp: 4, gid: 5, history_cap: 512, total_cap: 640, symbolic_cap: 512
188
+ 15:31:23.673 INF init_groups_from_model:824 | prefill grp: 5, gid: 6, history_cap: 640, total_cap: 768, symbolic_cap: 640
189
+ 15:31:23.673 INF init_groups_from_model:824 | prefill grp: 6, gid: 7, history_cap: 768, total_cap: 896, symbolic_cap: 768
190
+ 15:31:23.673 INF init_groups_from_model:824 | prefill grp: 7, gid: 8, history_cap: 896, total_cap: 1024, symbolic_cap: 896
191
+ 15:31:23.673 INF init_groups_from_model:824 | prefill grp: 8, gid: 9, history_cap: 1024, total_cap: 1152, symbolic_cap: 1024
192
+ 15:31:23.673 INF init_groups_from_model:831 | prefill_max_token_num : 1152
193
+ 15:31:23.674 INF Init:27 | LLaMaEmbedSelector use mmap
194
+ 100% | ################################ | 39 / 39 [13.45s<13.45s, 2.90 count/s] embed_selector init ok
195
+ 15:31:23.674 INF load_config:282 | load config:
196
+ 15:31:23.674 INF load_config:282 | {
197
+ 15:31:23.674 INF load_config:282 | "enable_repetition_penalty": false,
198
+ 15:31:23.674 INF load_config:282 | "enable_temperature": false,
199
+ 15:31:23.674 INF load_config:282 | "enable_top_k_sampling": false,
200
+ 15:31:23.674 INF load_config:282 | "enable_top_p_sampling": false,
201
+ 15:31:23.674 INF load_config:282 | "penalty_window": 20,
202
+ 15:31:23.674 INF load_config:282 | "repetition_penalty": 1.2,
203
+ 15:31:23.674 INF load_config:282 | "temperature": 0.9,
204
+ 15:31:23.674 INF load_config:282 | "top_k": 10,
205
+ 15:31:23.674 INF load_config:282 | "top_p": 0.8
206
+ 15:31:23.674 INF load_config:282 | }
207
+ 15:31:23.674 INF Init:1139 | LLM init ok
208
  Starting server on port 8000 with model 'AXERA-TECH/Qwen3-4B'...
209
+ API URLs:
210
+ GET http://127.0.0.1:8000/health
211
+ GET http://127.0.0.1:8000/v1/models
212
+ POST http://127.0.0.1:8000/v1/chat/completions
213
+ GET http://10.126.29.54:8000/health
214
+ GET http://10.126.29.54:8000/v1/models
215
+ POST http://10.126.29.54:8000/v1/chat/completions
216
+ GET http://172.17.0.1:8000/health
217
+ GET http://172.17.0.1:8000/v1/models
218
+ POST http://172.17.0.1:8000/v1/chat/completions
219
+ Aliases:
220
+ GET http://127.0.0.1:8000/models
221
+ POST http://127.0.0.1:8000/chat/completions
222
+ GET http://10.126.29.54:8000/models
223
+ POST http://10.126.29.54:8000/chat/completions
224
+ GET http://172.17.0.1:8000/models
225
+ POST http://172.17.0.1:8000/chat/completions
226
  OpenAI API Server starting on http://0.0.0.0:8000
227
  Max concurrency: 1
228
  Models: AXERA-TECH/Qwen3-4B
 
234
  from openai import OpenAI
235
 
236
  API_URL = "http://127.0.0.1:8000/v1"
237
+ MODEL = "AXERA-TECH/Qwen3-4B"
238
 
239
  messages = [
240
  {"role": "system", "content": [{"type": "text", "text": "you are a helpful assistant."}]},
 
257
  from openai import OpenAI
258
 
259
  API_URL = "http://127.0.0.1:8000/v1"
260
+ MODEL = "AXERA-TECH/Qwen3-4B"
261
 
262
  messages = [
263
  {"role": "system", "content": [{"type": "text", "text": "you are a helpful assistant."}]},
 
276
  delta = getattr(ev.choices[0], "delta", None)
277
  if delta and getattr(delta, "content", None):
278
  print(delta.content, end="", flush=True)
279
+ print(" ")
 
280
  ```
281
 
qwen3_p128_l0_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:fd61cf7ac7196904510a51be616f3aefe63a59879b861ba4de0050c2a8343d55
3
- size 125435994
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:02b0899a5556ef22d998483684daf73cb9c2d17dc27c6d1a9574c874d2b4670a
3
+ size 123979452
qwen3_p128_l10_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:42664547adf097be89a8bafbcf4f3a8f20f7c86dfcfe06e5b722db36289507c6
3
- size 125435866
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9596abc7306d71629783add37a7a06dc37742bc28ec9cb9fef99aa832da2914d
3
+ size 123979388
qwen3_p128_l11_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:2b1bfd35954826cae66327a92f60c8d7b97892adaf6c966a06945664212da538
3
- size 125435770
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5b0b3c840707dea93a76aaa2415f6402d93830ec60fecc0d1531d3281ccf0eba
3
+ size 123979676
qwen3_p128_l12_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:9057624167eea41908d6ae0fda33bdef1fc61eb0b6e6e6fd4f7c0a45a83b52a0
3
- size 125435930
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c20759dc73598b909ede9f0e31bc47fc83314371981bf6adfc3d461c84f2cb79
3
+ size 123979612
qwen3_p128_l13_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:f282cb43221711d9b145de5001fb0fbb531939139f90514362991038215a973e
3
- size 125435994
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6bdfa75a3ac502ad6bb3d5868f1ed25749e7d7ea0c85c3f571aeaeeeffbd060a
3
+ size 123979644
qwen3_p128_l14_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:8b2422806e1ca9e1fcb36927b2a0861f5cd9b5b2531a0c639b0295d089423144
3
- size 125435866
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2f85bd6df6fd895dec5d5847939588a113ba5205c57e20da378858dcf70b891e
3
+ size 123979388
qwen3_p128_l15_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:0dc68fa4d4a4118d3bed91246423ebd49bedb05b6ed0a1f421b3740198f6318d
3
- size 125435866
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c75363e4f5ebd2dafcab0bd2bfd3356b70e983648dbe74ba38e14b43c01353f9
3
+ size 123979388
qwen3_p128_l16_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:fd5ed3869713ec8159f4f36f9e64f230910550ec5b7b74c856ee8b066a87b43d
3
- size 125435866
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c3cb11566ead82242eec59fa99b30465779a42e3a16c24837b363673e1641732
3
+ size 123979388
qwen3_p128_l17_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:dfe4b4b90a08c34e3a5300aede9c4f13aa8046b64a0cfc7174a4d374d6d11049
3
- size 125436026
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:327017a908862a7f19d4348279d0997fb7a68139c66275c33d4dd7ff7f305575
3
+ size 123979516
qwen3_p128_l18_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:3a3a16f84a0249f18e713f4efe132a1f553cabf994baf06dbe460ea9fc28f6a7
3
- size 125435866
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2258b4dcbab0b16cd6c89db7afe63ea7e57793548590f2b8376ec0ffbf8b27ac
3
+ size 123979388
qwen3_p128_l19_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:85f6dcfc2b5d6c4c6a79f1e24fa07ec0dd5e51a164b8fc0a2c5bd503515a5f86
3
- size 125435866
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5b80ba1e6f48c554eee2b26b671b62f3be842eae8814d3797dfae34106f3dbf3
3
+ size 123979388
qwen3_p128_l1_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:9b398b242b539a43e7216c41849e8e1d4eda19f9ad940f88eba4c5b0035e02c0
3
- size 125440026
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:10d28d462863fd0b3fb8d2e7deb190e402e11e36bbddd2b4c4283cf8b2c9c535
3
+ size 123988604
qwen3_p128_l20_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:a66d39ef6172cd4bc52bcd6caca988d709d0525cfae0d7e4bd5f750afad5b6d9
3
- size 125435866
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eb0346fd6da1aa4ddbb269d9cbd2c05b7ddde45deb0b172cc12e3b0d52ac994a
3
+ size 123979388
qwen3_p128_l21_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:76135864b098a5c4d034942eeb4e9c1f53faa9ad3fae331ed0ae23b7104eabfb
3
- size 125435866
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f29e9dedf09a284beae4494f32c972c5f2e478dcea6c38f670524554ea70e57b
3
+ size 123979676
qwen3_p128_l22_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:978a4369fbf912f246bbe5385ec56aa104c6b9c3abedb6f0ec6e06be9776b487
3
- size 125436314
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ab32b903e822a6af08bd5e589a55d533bc7e76014412121ce65b8554a3585ccb
3
+ size 123980252
qwen3_p128_l23_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:6ca5b13b94a9014c41c5dd368d6dd4bd321b3ee92ebcd07e5ac91806af1c8a5d
3
- size 125435866
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1be749ff3aa7e9ea666bc62ebc99fc4f2e0179ac7c1058359e7c282dcbef5cd5
3
+ size 123979548
qwen3_p128_l24_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b40143f306365b23742950fc4cb3e41561c89f1939396555a5b4f2bbad9579d5
3
- size 125436346
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:52d8bdc2f7367da317c9b48768e8aa9c49f015ff9b457b01615cdead8b24df68
3
+ size 123980828
qwen3_p128_l25_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:636e640a567fdf71a8056b47b722ce17e304d325089bcf43829c1b66391eda0e
3
- size 125436698
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5bfcc4edbb964297010f1a140c489e8d1091f935eec0add2777d5ca4816c0c2a
3
+ size 123980092
qwen3_p128_l26_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:2c198a64b51bc0fd8f45f8a30bb1fc531a31710937f0ed8b9d13818ef9370842
3
- size 125436218
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:00d6a1f4bff5f0abc38ef0e3a362ae49aed79360804424145fe76e7a528d5320
3
+ size 123980188
qwen3_p128_l27_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:a4ffb990aa89ff89aa4eaa06c03194c912c7fd8e467d12949b146d36362dd9cc
3
- size 125436890
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:70bfb2bcec42a38fecfa51fe09bdbda12945f2aa64467918f847e2f3863d7cf8
3
+ size 123980700
qwen3_p128_l28_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:efe327aafdeb9cbd7c2d866355bc3bbb80605b8d518a2cc337e08f8bf1a10e72
3
- size 125436954
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:80bd65220499684d5ef14044ffbed0cb140c8428866254d3f8aa1ecfc50d891a
3
+ size 123981468
qwen3_p128_l29_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:968f868eb64ff7128d9e093f1c1ab3a624efd5d4655e354c5f432ed4fa625c09
3
- size 125436474
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:34c928efd1abe007b8022dcac6d9eaf1c0ba4fbb5c238703f6e7b2b234268b7e
3
+ size 123980092
qwen3_p128_l2_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:8418d37509e45570604d55944bf20d5315c49fe062aa299c08fb0479f8a57133
3
- size 125439642
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e602dd9a672e5b01604f306c388587a2373e55e9d889812ee3678bef79161c23
3
+ size 123986620
qwen3_p128_l30_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:e7a6612bd0d607929e9e6608bc408bf11d94bdf214e37982c142f91094821ff9
3
- size 125436314
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b8f3355fd20986edee0f2fa0f9066fe090da51bfa6ddb58909421092b365fb58
3
+ size 123979740
qwen3_p128_l31_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:9574520e448c65cd95d27f3fd28638124fc1c5363044e917eb4975aef0f095a9
3
- size 125436794
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:11947240679adc33ba71d02c947cd8880f2baa0552150e9f47169cc322ff838b
3
+ size 123981308
qwen3_p128_l32_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b86f01a281913b7279f3c49ce470193df3e2d85c98cce97ef3fb4b7a1f037c28
3
- size 125436090
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3b8caa94e42126844bc58dd5b7240c05f5a8e9a7f4cbc264992ce0f540688c26
3
+ size 123980316
qwen3_p128_l33_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:0078a10187bfef039e39b57d982481a4be54fc8e81bd5a900d4cf4efb3471031
3
- size 125436186
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b7ac5778753734f5d20d8a05d972151827d6d096cd86d2985a778a9347b0d5a1
3
+ size 123980220
qwen3_p128_l34_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:c951c6dc142fb7cb2f8e72bd925373af1de69eec282c0a3cb596e5e05ba75d3a
3
- size 125435866
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8b489866fb74727196cfa87a6ea6b363c94144a9f361d02f92a4635f183c3651
3
+ size 123980156
qwen3_p128_l35_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:ed942d4188232181386acb934b1f6905ae8c8bfedec3755eac73277706c9814d
3
- size 125436186
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:66a334e9f55c0b99ddfe579e1300eb33f2b28c598828aeaebcc30333c177c35c
3
+ size 123979676
qwen3_p128_l3_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:32a7e2ecce7c86084be1d79e25339156c2ea9ac7e832032a7069436ee88e79e8
3
- size 125444634
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:44ed5fd5472184ad39581afbdfe91fe9fda7cd113ba6d5bb30d4ca47d8a875dc
3
+ size 123998076
qwen3_p128_l4_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:d2ffaee8a58efb08b4fd62fcf7c9245e22780b1c144f555b236e4b3ebeb6355c
3
- size 125435866
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a48452d1b23cad0987c36ebca404ca7aec6b567f46aa42f637be31fc258c554c
3
+ size 123979388
qwen3_p128_l5_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:9a4570bb23c558885d9ea3d68aa6a92d58e4f78fe539f5219d31de4e611b02da
3
- size 125435962
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:af5a5a5dbaf0260eb1bbc6c13ff0bbc199d9cc7f1ec573b2952403ef3b0fc17c
3
+ size 123979580
qwen3_p128_l6_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:ac517690334efa7884c031d5dde283eb42a45a5945ac85f7db8f0bb2542a411b
3
- size 125438682
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:495a48a40115045977de974403345ddc27afec699ddee36ae37aa609e0a3848a
3
+ size 123985692
qwen3_p128_l7_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:8cb167635c3749887ca90e0fc1f816ed67d28bce21fe39c729e7c97ccd178803
3
- size 125438458
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:77c3418850c1028ba1b913aee271de1676050eb3c41edc4aecaed15043a4a10a
3
+ size 123985116
qwen3_p128_l8_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:0d3d0223bc14f04f5d30461be1423d2121cefd1b4b72102c4129a9c4c51f6e52
3
- size 125435866
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8c26544765be766c2e160cdd895ec0ba143b75f434deb3efa34b37d4e1287379
3
+ size 123979388
qwen3_p128_l9_together.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:609721ac72bf8f0c49dc2774fd230d66d140ba787b684a98237de3c6c4c77bd7
3
- size 125435866
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e0f9927075c249ef4c5ceb0f95b7bd638e1ed8e74e1960b74e443c19684a5120
3
+ size 123979516
qwen3_post.axmodel CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:0f4641b04d989f444a37dd984993ee95b6f18cfbf39e79714a202a1b0633561e
3
  size 424609056
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2b1f8706a646d9d9ddf8d85bf4e0477c75eed9346d19da37de2bb5ea80abe70f
3
  size 424609056