Datasets:

Hamerlate
/

tts

@@ -4,168 +4,174 @@ Try the 'nsys status --environment' command to learn more.
 WARNING: CPU context switch tracing not supported, disabling.
 Try the 'nsys status --environment' command to learn more.
-INFO 08-11 19:11:41 [__init__.py:244] Automatically detected platform cuda.
 --- vLLM 性能基准测试 (含 NVTX 标记) ---
 模型: Qwen/Qwen2-1.5B
 批量大小: [128, 64, 32, 16]
 输入/输出 Token (近似): 128 / 512
 ---------------------------------------------
 正在加载模型... 这可能需要一些时间。
-INFO 08-11 19:11:51 [config.py:841] This model supports multiple tasks: {'reward', 'generate', 'classify', 'embed'}. Defaulting to 'generate'.
-INFO 08-11 19:11:51 [config.py:1472] Using max model len 131072
-INFO 08-11 19:11:51 [config.py:2285] Chunked prefill is enabled with max_num_batched_tokens=8192.
-INFO 08-11 19:11:53 [core.py:526] Waiting for init message from front-end.
-INFO 08-11 19:11:53 [core.py:69] Initializing a V1 LLM engine (v0.9.2) with config: model='Qwen/Qwen2-1.5B', speculative_config=None, tokenizer='Qwen/Qwen2-1.5B', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, override_neuron_config={}, tokenizer_revision=None, trust_remote_code=True, dtype=torch.bfloat16, max_seq_len=131072, download_dir=None, load_format=LoadFormat.AUTO, tensor_parallel_size=1, pipeline_parallel_size=1, disable_custom_all_reduce=False, quantization=None, enforce_eager=False, kv_cache_dtype=auto,  device_config=cuda, decoding_config=DecodingConfig(backend='auto', disable_fallback=False, disable_any_whitespace=False, disable_additional_properties=False, reasoning_backend=''), observability_config=ObservabilityConfig(show_hidden_metrics_for_version=None, otlp_traces_endpoint=None, collect_detailed_traces=None), seed=0, served_model_name=Qwen/Qwen2-1.5B, num_scheduler_steps=1, multi_step_stream_outputs=True, enable_prefix_caching=True, chunked_prefill_enabled=True, use_async_output_proc=True, pooler_config=None, compilation_config={"level":3,"debug_dump_path":"","cache_dir":"","backend":"","custom_ops":[],"splitting_ops":["vllm.unified_attention","vllm.unified_attention_with_output"],"use_inductor":true,"compile_sizes":[],"inductor_compile_config":{"enable_auto_functionalized_v2":false},"inductor_passes":{},"use_cudagraph":true,"cudagraph_num_of_warmups":1,"cudagraph_capture_sizes":[512,504,496,488,480,472,464,456,448,440,432,424,416,408,400,392,384,376,368,360,352,344,336,328,320,312,304,296,288,280,272,264,256,248,240,232,224,216,208,200,192,184,176,168,160,152,144,136,128,120,112,104,96,88,80,72,64,56,48,40,32,24,16,8,4,2,1],"cudagraph_copy_inputs":false,"full_cuda_graph":false,"max_capture_size":512,"local_cache_dir":null}
-INFO 08-11 19:11:54 [parallel_state.py:1076] rank 0 in world size 1 is assigned as DP rank 0, PP rank 0, TP rank 0, EP rank 0
-WARNING 08-11 19:11:54 [topk_topp_sampler.py:59] FlashInfer is not available. Falling back to the PyTorch-native implementation of top-p & top-k sampling. For the best performance, please install FlashInfer.
-INFO 08-11 19:11:54 [gpu_model_runner.py:1770] Starting to load model Qwen/Qwen2-1.5B...
-INFO 08-11 19:11:54 [gpu_model_runner.py:1775] Loading model from scratch...
-INFO 08-11 19:11:54 [cuda.py:284] Using Flash Attention backend on V1 engine.
-INFO 08-11 19:11:55 [weight_utils.py:292] Using model weights format ['*.safetensors']
-INFO 08-11 19:11:55 [weight_utils.py:308] Time spent downloading weights for Qwen/Qwen2-1.5B: 0.522589 seconds
-INFO 08-11 19:11:56 [weight_utils.py:345] No model.safetensors.index.json found in remote.
-INFO 08-11 19:11:56 [default_loader.py:272] Loading weights took 0.89 seconds
-INFO 08-11 19:11:57 [gpu_model_runner.py:1801] Model loading took 2.9110 GiB and 2.424753 seconds
-INFO 08-11 19:12:04 [backends.py:508] Using cache directory: /home/cy/.cache/vllm/torch_compile_cache/19cf05a3aa/rank_0_0/backbone for vLLM's torch.compile
-INFO 08-11 19:12:04 [backends.py:519] Dynamo bytecode transform time: 6.61 s
-INFO 08-11 19:12:08 [backends.py:155] Directly load the compiled graph(s) for shape None from the cache, took 4.174 s
-INFO 08-11 19:12:09 [monitor.py:34] torch.compile takes 6.61 s in total
-INFO 08-11 19:12:10 [gpu_worker.py:232] Available KV cache memory: 16.88 GiB
-INFO 08-11 19:12:10 [kv_cache_utils.py:716] GPU KV cache size: 632,176 tokens
-INFO 08-11 19:12:10 [kv_cache_utils.py:720] Maximum concurrency for 131,072 tokens per request: 4.82x
-INFO 08-11 19:12:26 [gpu_model_runner.py:2326] Graph capturing finished in 16 secs, took 0.47 GiB
-INFO 08-11 19:12:26 [core.py:172] init engine (profile, create kv cache, warmup model) took 29.36 seconds
 模型加载完成。
 使用长度为 2760 的示例文本作为输入 Prompt。
 ===== 正在运行批量大小为 128 的基准测试 =====
 正在进行预热...
 开始计时和剖析...
 --- 结果 (批量大小: 128) ---
-执行时间: 5.142 秒
 实际平均输入 Token 数: 541.00
-总计生成 Token 数: 65534
-吞吐量 (Throughput): 12745.89 tokens/second
 ===== 正在运行批量大小为 64 的基准测试 =====
 正在进行预热...
 开始计时和剖析...
 --- 结果 (批量大小: 64) ---
-执行时间: 3.979 秒
 实际平均输入 Token 数: 541.00
 总计生成 Token 数: 32768
-吞吐量 (Throughput): 8235.59 tokens/second
 ===== 正在运行批量大小为 32 的基准测试 =====
 正在进行预热...
 开始计时和剖析...
 --- 结果 (批量大小: 32) ---
-执行时间: 3.408 秒
 实际平均输入 Token 数: 541.00
 总计生成 Token 数: 16384
-吞吐量 (Throughput): 4807.24 tokens/second
 ===== 正在运行批量大小为 16 的基准测试 =====
 正在进行预热...
 开始计时和剖析...
 --- 结果 (批量大小: 16) ---
-执行时间: 3.066 秒
 实际平均输入 Token 数: 541.00
 总计生成 Token 数: 8192
-吞吐量 (Throughput): 2671.95 tokens/second
 GPU 3: General Metrics for NVIDIA AD10x (any frequency)
-Generating '/tmp/nsys-report-b6a4.qdstrm'
-SKIPPED: No data available.
 [3/8] Executing 'nvtx_sum' stats report
 [4/8] Executing 'osrt_sum' stats report
  Time (%)   Total Time (ns)   Num Calls      Avg (ns)          Med (ns)      Min (ns)      Max (ns)       StdDev (ns)             Name
  --------  -----------------  ---------  ----------------  ----------------  ---------  --------------  ---------------  ----------------------
-     80.8  1,316,401,531,684         96  13,712,515,955.0  12,735,288,428.5    173,376  15,077,000,078  1,885,845,753.5  pthread_cond_wait
-      9.2    149,488,620,646      4,680      31,942,013.0       4,998,638.5      1,080  34,489,398,805    824,807,828.7  epoll_wait
-      5.9     95,329,273,419     10,814       8,815,357.3           1,450.0      1,000  10,010,050,788    179,315,085.6  poll
-      1.6     26,737,688,064         12   2,228,140,672.0         168,659.5     92,695  10,000,082,423  4,109,907,418.4  sem_timedwait
-      1.5     24,256,687,442      4,105       5,909,059.1       5,624,747.0    135,044      19,182,006      1,253,243.9  sem_wait
-      1.0     15,508,403,228         70     221,548,617.5         409,454.0      4,172     500,093,813    250,108,104.9  pthread_cond_timedwait
-      0.1        993,011,624     20,011          49,623.3           2,522.0      1,000     241,598,899      2,824,985.7  read
-      0.0        150,708,163     71,161           2,117.8           1,338.0      1,000         143,323          2,884.1  munmap
-      0.0        126,977,808        234         542,640.2           2,318.5      1,065      20,643,834      3,080,898.9  fopen
-      0.0         88,254,464      1,461          60,406.9           7,757.0      1,020      31,636,400        864,986.1  ioctl
-      0.0         81,004,509         16       5,062,781.8       5,062,915.0  5,053,743       5,074,146          4,767.2  nanosleep
-      0.0         75,692,289     18,126           4,175.9           2,564.0      1,009      23,499,766        174,532.7  open64
-      0.0         68,718,668        459         149,713.9           1,070.0      1,000      62,152,067      2,902,407.9  waitpid
-      0.0         57,403,976         66         869,757.2           3,393.0      1,042      22,370,923      4,028,517.2  open
-      0.0         50,385,956          2      25,192,978.0      25,192,978.0    971,165      49,414,791     34,254,816.4  fork
-      0.0         37,577,142         98         383,440.2          17,467.5     10,169       9,068,538      1,374,492.3  pthread_join
-      0.0         17,645,503      4,166           4,235.6           3,600.0      1,061          28,839          2,210.4  recv
-      0.0         14,081,767      3,210           4,386.8           2,402.5      1,000       1,851,335         33,686.7  mmap64
-      0.0         11,032,769      4,573           2,412.6           1,938.0      1,006         118,235          2,611.0  write
-      0.0          6,630,899         98          67,662.2          69,180.0     49,884          77,867          4,854.4  sleep
-      0.0          5,571,075         58          96,053.0          73,668.0     33,776         415,696         73,221.2  pthread_create
-      0.0          4,604,598        499           9,227.7           1,912.0      1,000          87,359         12,344.0  fgets
-      0.0          1,394,084        255           5,467.0           5,088.0      1,136          26,156          2,139.6  send
-      0.0          1,310,419      1,107           1,183.8           1,054.0      1,000          12,416            660.0  fclose
-      0.0            708,201         48          14,754.2           2,643.5      1,811         288,844         51,216.4  futex
-      0.0            592,906        378           1,568.5           1,378.5      1,000           8,587          1,014.6  epoll_ctl
-      0.0            519,985        161           3,229.7           2,675.0      1,092          32,233          2,740.2  pthread_cond_signal
-      0.0            253,194         42           6,028.4           4,997.0      1,869          16,252          3,640.3  pipe2
-      0.0            197,022         18          10,945.7           5,368.5      1,408          74,228         16,532.4  mmap
-      0.0            162,812          6          27,135.3          22,212.0      2,546          71,892         26,887.7  bind
-      0.0            123,295         25           4,931.8           3,466.0      2,169          10,744          2,960.7  fopen64
-      0.0             98,953          9          10,994.8          10,105.0      4,623          17,643          4,086.1  socket
-      0.0             70,577          6          11,762.8          14,542.5      1,165          22,484          8,515.1  pthread_mutex_lock
-      0.0             45,925         15           3,061.7           2,004.0      1,695           9,510          1,986.5  stat
-      0.0             39,631         24           1,651.3           1,315.0      1,013           5,528            962.1  fcntl
-      0.0             38,556          2          19,278.0          19,278.0     15,634          22,922          5,153.4  connect
-      0.0             32,708          3          10,902.7           9,384.0      6,028          17,296          5,785.5  accept4
-      0.0             29,641          4           7,410.3           6,130.5      3,290          14,090          4,986.6  fread
-      0.0             26,997         13           2,076.7           2,340.0      1,145           2,990            651.7  dup2
-      0.0             25,520         14           1,822.9           1,896.0      1,058           2,488            348.0  sigaction
-      0.0             25,179          7           3,597.0           4,768.0      1,123           5,648          2,033.8  fflush
-      0.0             22,308          4           5,577.0           5,378.0      5,067           6,485            638.7  lstat
-      0.0             18,002          9           2,000.2           1,722.0      1,006           3,405            803.5  pread
-      0.0             17,322          3           5,774.0           4,877.0      4,674           7,771          1,732.4  pthread_cond_broadcast
-      0.0             12,954          2           6,477.0           6,477.0      5,866           7,088            864.1  fwrite
-      0.0             12,526          1          12,526.0          12,526.0     12,526          12,526              0.0  kill
-      0.0             12,478          3           4,159.3           4,097.0      4,061           4,320            140.3  fputs_unlocked
-      0.0             11,283          4           2,820.8           2,893.5      2,497           2,999            234.3  mprotect
-      0.0              7,550          4           1,887.5           1,868.0      1,615           2,199            239.5  listen
-      0.0              5,446          3           1,815.3           1,667.0      1,602           2,177            314.9  fstat
-      0.0              3,822          2           1,911.0           1,911.0      1,424           2,398            688.7  openat64
-      0.0              3,294          1           3,294.0           3,294.0      3,294           3,294              0.0  fputs
 [5/8] Executing 'cuda_api_sum' stats report
  Time (%)  Total Time (ns)  Num Calls    Avg (ns)    Med (ns)  Min (ns)   Max (ns)    StdDev (ns)            Name
  --------  ---------------  ---------  ------------  --------  --------  -----------  ------------  ----------------------
-    100.0      147,884,600         12  12,3SKIPPED: /data/cy/qwen_util_bs.sqlite does not contain CUDA kernel data.
-SKIPPED: /data/cy/qwen_util_bs.sqlite does not contain GPU memory data.
-SKIPPED: /data/cy/qwen_util_bs.sqlite does not contain GPU memory data.
-23,716.7  21,398.0     3,612  147,706,958  42,634,665.1  cudaDeviceSynchronize
-      0.0            2,546          1       2,546.0   2,546.0     2,546        2,546           0.0  cuModuleGetLoadingMode
 [6/8] Executing 'cuda_gpu_kern_sum' stats report
 [7/8] Executing 'cuda_gpu_mem_time_sum' stats report

 WARNING: CPU context switch tracing not supported, disabling.
 Try the 'nsys status --environment' command to learn more.
+INFO 08-11 21:41:28 [__init__.py:244] Automatically detected platform cuda.
 --- vLLM 性能基准测试 (含 NVTX 标记) ---
 模型: Qwen/Qwen2-1.5B
 批量大小: [128, 64, 32, 16]
 输入/输出 Token (近似): 128 / 512
 ---------------------------------------------
 正在加载模型... 这可能需要一些时间。
+INFO 08-11 21:41:38 [config.py:841] This model supports multiple tasks: {'classify', 'embed', 'reward', 'generate'}. Defaulting to 'generate'.
+INFO 08-11 21:41:38 [config.py:1472] Using max model len 131072
+INFO 08-11 21:41:39 [config.py:2285] Chunked prefill is enabled with max_num_batched_tokens=8192.
+INFO 08-11 21:41:40 [core.py:526] Waiting for init message from front-end.
+INFO 08-11 21:41:40 [core.py:69] Initializing a V1 LLM engine (v0.9.2) with config: model='Qwen/Qwen2-1.5B', speculative_config=None, tokenizer='Qwen/Qwen2-1.5B', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, override_neuron_config={}, tokenizer_revision=None, trust_remote_code=True, dtype=torch.bfloat16, max_seq_len=131072, download_dir=None, load_format=LoadFormat.AUTO, tensor_parallel_size=1, pipeline_parallel_size=1, disable_custom_all_reduce=False, quantization=None, enforce_eager=False, kv_cache_dtype=auto,  device_config=cuda, decoding_config=DecodingConfig(backend='auto', disable_fallback=False, disable_any_whitespace=False, disable_additional_properties=False, reasoning_backend=''), observability_config=ObservabilityConfig(show_hidden_metrics_for_version=None, otlp_traces_endpoint=None, collect_detailed_traces=None), seed=0, served_model_name=Qwen/Qwen2-1.5B, num_scheduler_steps=1, multi_step_stream_outputs=True, enable_prefix_caching=True, chunked_prefill_enabled=True, use_async_output_proc=True, pooler_config=None, compilation_config={"level":3,"debug_dump_path":"","cache_dir":"","backend":"","custom_ops":[],"splitting_ops":["vllm.unified_attention","vllm.unified_attention_with_output"],"use_inductor":true,"compile_sizes":[],"inductor_compile_config":{"enable_auto_functionalized_v2":false},"inductor_passes":{},"use_cudagraph":true,"cudagraph_num_of_warmups":1,"cudagraph_capture_sizes":[512,504,496,488,480,472,464,456,448,440,432,424,416,408,400,392,384,376,368,360,352,344,336,328,320,312,304,296,288,280,272,264,256,248,240,232,224,216,208,200,192,184,176,168,160,152,144,136,128,120,112,104,96,88,80,72,64,56,48,40,32,24,16,8,4,2,1],"cudagraph_copy_inputs":false,"full_cuda_graph":false,"max_capture_size":512,"local_cache_dir":null}
+INFO 08-11 21:41:41 [parallel_state.py:1076] rank 0 in world size 1 is assigned as DP rank 0, PP rank 0, TP rank 0, EP rank 0
+WARNING 08-11 21:41:41 [topk_topp_sampler.py:59] FlashInfer is not available. Falling back to the PyTorch-native implementation of top-p & top-k sampling. For the best performance, please install FlashInfer.
+INFO 08-11 21:41:41 [gpu_model_runner.py:1770] Starting to load model Qwen/Qwen2-1.5B...
+INFO 08-11 21:41:41 [gpu_model_runner.py:1775] Loading model from scratch...
+INFO 08-11 21:41:41 [cuda.py:284] Using Flash Attention backend on V1 engine.
+INFO 08-11 21:41:42 [weight_utils.py:292] Using model weights format ['*.safetensors']
+INFO 08-11 21:41:43 [weight_utils.py:345] No model.safetensors.index.json found in remote.
+INFO 08-11 21:41:44 [default_loader.py:272] Loading weights took 0.81 seconds
+INFO 08-11 21:41:44 [gpu_model_runner.py:1801] Model loading took 2.9110 GiB and 2.327024 seconds
+INFO 08-11 21:41:51 [backends.py:508] Using cache directory: /home/cy/.cache/vllm/torch_compile_cache/19cf05a3aa/rank_0_0/backbone for vLLM's torch.compile
+INFO 08-11 21:41:51 [backends.py:519] Dynamo bytecode transform time: 6.51 s
+INFO 08-11 21:41:55 [backends.py:155] Directly load the compiled graph(s) for shape None from the cache, took 4.107 s
+INFO 08-11 21:41:56 [monitor.py:34] torch.compile takes 6.51 s in total
+INFO 08-11 21:41:57 [gpu_worker.py:232] Available KV cache memory: 16.88 GiB
+INFO 08-11 21:41:57 [kv_cache_utils.py:716] GPU KV cache size: 632,176 tokens
+INFO 08-11 21:41:57 [kv_cache_utils.py:720] Maximum concurrency for 131,072 tokens per request: 4.82x
+INFO 08-11 21:42:14 [gpu_model_runner.py:2326] Graph capturing finished in 17 secs, took 0.47 GiB
+INFO 08-11 21:42:14 [core.py:172] init engine (profile, create kv cache, warmup model) took 29.59 seconds
 模型加载完成。
 使用长度为 2760 的示例文本作为输入 Prompt。
 ===== 正在运行批量大小为 128 的基准测试 =====
 正在进行预热...
 开始计时和剖析...
 --- 结果 (批量大小: 128) ---
+执行时间: 5.116 秒
 实际平均输入 Token 数: 541.00
+总计生成 Token 数: 65536
+吞吐量 (Throughput): 12811.03 tokens/second
 ===== 正在运行批量大小为 64 的基准测试 =====
 正在进行预热...
 开始计时和剖析...
 --- 结果 (批量大小: 64) ---
+执行时间: 3.964 秒
 实际平均输入 Token 数: 541.00
 总计生成 Token 数: 32768
+吞吐量 (Throughput): 8265.40 tokens/second
 ===== 正在运行批量大小为 32 的基准测试 =====
 正在进行预热...
 开始计时和剖析...
 --- 结果 (批量大小: 32) ---
+执行时间: 3.412 秒
 实际平均输入 Token 数: 541.00
 总计生成 Token 数: 16384
+吞吐量 (Throughput): 4801.29 tokens/second
 ===== 正在运行批量大小为 16 的基准测试 =====
 正在进行预热...
 开始计时和剖析...
 --- 结果 (批量大小: 16) ---
+执行时间: 3.084 秒
 实际平均输入 Token 数: 541.00
 总计生成 Token 数: 8192
+吞吐量 (Throughput): 2656.39 tokens/second
 GPU 3: General Metrics for NVIDIA AD10x (any frequency)
+Generating '/tmp/nsys-report-d8d0.qdstrm'
 [3/8] Executing 'nvtx_sum' stats report
+ Time (%)  Total Time (ns)  Instances     Avg (ns)         Med (ns)        Min (ns)       Max (ns)     StdDev (ns)   Style      Range
+ --------  ---------------  ---------  ---------------  ---------------  -------------  -------------  -----------  -------  ------------
+     32.8    5,115,252,310          1  5,115,252,310.0  5,115,252,310.0  5,115,252,310  5,115,252,310          0.0  PushPop  :bs:128 qwen
+     25.5    3,964,311,307          1  3,964,311,307.0  3,964,311,307.0  3,964,311,307  3,964,311,307          0.0  PushPop  :bs:64 qwen
+     21.9    3,412,263,536          1  3,412,263,536.0  3,412,263,536.0  3,412,263,536  3,412,263,536          0.0  PushPop  :bs:32 qwen
+     19.8    3,083,737,486          1  3,083,737,486.0  3,083,737,486.0  3,083,737,486  3,083,737,486          0.0  PushPop  :bs:16 qwen
 [4/8] Executing 'osrt_sum' stats report
  Time (%)   Total Time (ns)   Num Calls      Avg (ns)          Med (ns)      Min (ns)      Max (ns)       StdDev (ns)             Name
  --------  -----------------  ---------  ----------------  ----------------  ---------  --------------  ---------------  ----------------------
+     81.5  1,370,973,183,949         96  14,280,970,666.1  13,375,999,749.0    133,132  15,604,144,341  1,919,025,045.4  pthread_cond_wait
+      8.8    148,715,278,362      4,676      31,803,951.7       5,002,643.5      1,009  34,336,697,739    821,232,877.1  epoll_wait
+      5.6     94,912,909,337      9,798       9,686,967.7           1,375.5      1,000  10,010,047,625    187,778,851.4  poll
+      1.6     26,570,890,385         12   2,214,240,865.4         165,453.5     99,894  10,000,097,576  4,093,524,018.4  sem_timedwait
+      1.4     24,098,966,884      4,104       5,872,068.0       5,385,563.5    146,730      18,879,015      1,264,152.9  sem_wait
+      0.9     15,505,809,401         65     238,550,913.9         456,999.0     10,476     500,098,586    251,665,440.9  pthread_cond_timedwait
+      0.1      1,245,043,955     17,459          71,312.4           2,384.0      1,000     517,177,937      4,597,714.0  read
+      0.0        215,661,289        234         921,629.4           2,287.0      1,103      34,390,374      4,274,980.9  fopen
+      0.0        154,104,361     73,084           2,108.6           1,267.0      1,006         167,363          3,324.7  munmap
+      0.0         81,045,901         16       5,065,368.8       5,064,039.5  5,053,918       5,080,925          7,510.2  nanosleep
+      0.0         80,555,283      1,475          54,613.8           7,151.0      1,026      28,044,030        753,940.6  ioctl
+      0.0         54,719,358         67         816,706.8           3,552.0      1,029      18,800,673      3,782,575.7  open
+      0.0         51,541,123          2      25,770,561.5      25,770,561.5  1,044,901      50,496,222     34,967,364.4  fork
+      0.0         50,514,209     18,121           2,787.6           2,350.0      1,000         716,512          5,542.5  open64
+      0.0         48,865,146        569          85,879.0           1,055.0      1,000      42,675,079      1,790,964.1  waitpid
+      0.0         39,788,388         97         410,189.6          12,283.0      6,446       6,066,262      1,109,283.3  pthread_join
+      0.0         18,545,605         98         189,240.9          70,024.5     46,406      11,983,727      1,203,717.1  sleep
+      0.0         16,159,369      4,167           3,877.9           3,636.0      1,160          17,449          1,038.1  recv
+      0.0         13,480,816      3,260           4,135.2           2,216.5      1,000       1,471,516         26,612.3  mmap64
+      0.0         10,271,775      4,569           2,248.1           1,946.0      1,027          50,716          1,978.1  write
+      0.0          7,098,308         58         122,384.6          80,090.5     27,747         670,356        135,549.3  pthread_create
+      0.0          4,057,179        538           7,541.2           1,660.0      1,004          93,747         10,727.2  fgets
+      0.0          1,464,194        255           5,741.9           4,939.0      1,487          20,660          2,741.1  send
+      0.0          1,269,204      1,084           1,170.9           1,034.0      1,000           9,944            658.1  fclose
+      0.0            489,450        307           1,594.3           1,400.0      1,003           9,045            749.7  epoll_ctl
+      0.0            425,006        171           2,485.4           2,020.0      1,018          14,308          1,637.7  pthread_cond_signal
+      0.0            306,299         48           6,381.2           2,926.0      2,128          70,259         11,393.8  futex
+      0.0            254,267         42           6,054.0           5,096.0      2,110          16,888          3,347.8  pipe2
+      0.0            201,080         18          11,171.1           5,789.5      2,323          71,972         16,039.3  mmap
+      0.0            144,323          6          24,053.8          18,950.5      5,438          59,553         22,282.8  bind
+      0.0            129,607          5          25,921.4          21,642.0     16,639          48,236         12,693.9  pthread_mutex_lock
+      0.0            124,038         25           4,961.5           3,170.0      2,078          11,898          3,257.1  fopen64
+      0.0             85,284          9           9,476.0           9,089.0      4,110          16,407          3,605.8  socket
+      0.0             48,050          3          16,016.7          17,656.0      9,950          20,444          5,435.7  accept4
+      0.0             41,473          2          20,736.5          20,736.5     14,059          27,414          9,443.4  connect
+      0.0             38,324         15           2,554.9           2,128.0      1,231           5,855          1,264.6  stat
+      0.0             37,561          4           9,390.3           7,437.0      3,187          19,500          7,496.2  fread
+      0.0             29,294         14           2,092.4           2,214.5      1,057           3,087            692.1  dup2
+      0.0             28,412         18           1,578.4           1,291.5      1,031           4,781            850.6  fcntl
+      0.0             27,322         16           1,707.6           1,694.5      1,040           2,489            363.9  sigaction
+      0.0             24,199          4           6,049.8           5,448.0      5,073           8,230          1,480.2  lstat
+      0.0             23,534          7           3,362.0           4,194.0      1,117           5,334          1,850.2  fflush
+      0.0             19,487          3           6,495.7           5,131.0      4,816           9,540          2,641.2  pthread_cond_broadcast
+      0.0             16,536          9           1,837.3           1,786.0      1,096           2,456            385.0  pread
+      0.0             12,881          3           4,293.7           4,369.0      4,063           4,449            203.7  fputs_unlocked
+      0.0             11,145          2           5,572.5           5,572.5      5,504           5,641             96.9  fwrite
+      0.0             10,851          4           2,712.8           2,681.0      2,427           3,062            310.7  mprotect
+      0.0              9,864          1           9,864.0           9,864.0      9,864           9,864              0.0  kill
+      0.0              6,254          4           1,563.5           1,699.5      SKIPPED: /data/cy/qwen_util_bs.sqlite does not contain CUDA kernel data.
+SKIPPED: /data/cy/qwen_util_bs.sqlite does not contain GPU memory data.
+SKIPPED: /data/cy/qwen_util_bs.sqlite does not contain GPU memory data.
+1,040           1,815            353.3  listen
+      0.0              4,609          3           1,536.3           1,475.0      1,238           1,896            333.3  fstat
+      0.0              3,625          1           3,625.0           3,625.0      3,625           3,625              0.0  fputs
+      0.0              2,994          2           1,497.0           1,497.0      1,125           1,869            526.1  openat64
 [5/8] Executing 'cuda_api_sum' stats report
  Time (%)  Total Time (ns)  Num Calls    Avg (ns)    Med (ns)  Min (ns)   Max (ns)    StdDev (ns)            Name
  --------  ---------------  ---------  ------------  --------  --------  -----------  ------------  ----------------------
+    100.0      151,119,781         12  12,593,315.1  20,533.0     3,744  150,954,978  43,572,624.4  cudaDeviceSynchronize
+      0.0            1,897          1       1,897.0   1,897.0     1,897        1,897           0.0  cuModuleGetLoadingMode
 [6/8] Executing 'cuda_gpu_kern_sum' stats report
 [7/8] Executing 'cuda_gpu_mem_time_sum' stats report

qwen_util_bs.nsys-rep CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:23694f022f2361cf5122a4269de279cfd98207e8e557650f4cc51a00e369e7bf
-size 18193309

 version https://git-lfs.github.com/spec/v1
+oid sha256:5d22febf26a5798bf0e477815e2ddcdc40166ff25ed61a7b76260cd5533922c8
+size 18219333