gcw_6a3w2yXz/multimolecule-maxentscan-score5-NPU
Model Introduction
Files & versions
Pull Requests
Discussion
Analysis

multimolecule/maxentscan-score5 on Ascend NPU

模型与任务

  • 模型:multimolecule/maxentscan-score5
  • Revision:bca49854adf4c53362a82a6cfa58bea4de0d4e92
  • 适配路线:splicing
  • 设备:npu:0

环境依赖

{
  "torch": "2.9.0+cpu",
  "torch_npu": "2.9.0.post1+gitee7ba04",
  "npu_available": true,
  "device_count": 2,
  "device_name": "Ascend910_9362",
  "model_dir": "/tmp/maxentscan-score5_weights",
  "python": "3.11.14",
  "window": 9,
  "mode": "score5"
}

安装:pip install -r requirements.txt

NPU 推理

默认执行:python inference.py

真实输入

{
  "sequences": [
    "strong_CAGGTAAGT",
    "strong_AAGGTAAGT",
    "moderate_CAGGTGAGT",
    "weak_polyT",
    "example_GTAAGTTTT",
    "brca_like_CAGGTATGG"
  ],
  "raw_sequences": [
    "CAGGTAAGT",
    "AAGGTAAGT",
    "CAGGTGAGT",
    "TTTTTTTTT",
    "GTAAGTTTT",
    "CAGGTATGG"
  ],
  "sequence_length": 9,
  "batch": 6,
  "alphabet": "A/C/G/U/N (T->U)",
  "window": 9,
  "mode": "score5",
  "sha256": {
    "strong_CAGGTAAGT": "46595d64b6f66ade4e93e207cbbc5318eb555d012595534cad90c6cd7e8b594b",
    "strong_AAGGTAAGT": "042498bd8b613e16b4639fef734e85f6936689cb02b1d1ccbfc80b724a61c8ca",
    "moderate_CAGGTGAGT": "75f8b3a2b2a17e1b9a9467b27617f83d5da3d188179568d275ea8c5a90ee8b39",
    "weak_polyT": "eecf22c3c7391e0583aaf74dd120ac92fab5d99da5dbc72e58122923c500f2ba",
    "example_GTAAGTTTT": "9e3a4c018d67fb961f58d58281e699bc4696127ebdd26b8f0d8cc89c6b97ff19",
    "brca_like_CAGGTATGG": "9af781e08e9d2ed361022615a13870d18c3cac1c0ea14b5bb3488a3c2b69e1df"
  }
}

真实推理结果

{
  "logits_shape": [
    6,
    1
  ],
  "logits_dtype": "torch.float32",
  "logits_device": "npu:0",
  "scores_cpu": [
    10.858312606811523,
    10.998353958129883,
    10.673585891723633,
    -31.079252243041992,
    -30.759328842163086,
    9.9851655960083
  ],
  "scores_npu": [
    10.858312606811523,
    10.998353958129883,
    10.673585891723633,
    -31.079252243041992,
    -30.759326934814453,
    9.9851655960083
  ],
  "ranking_cpu": [
    "strong_AAGGTAAGT",
    "strong_CAGGTAAGT",
    "moderate_CAGGTGAGT",
    "brca_like_CAGGTATGG",
    "example_GTAAGTTTT",
    "weak_polyT"
  ],
  "ranking_npu": [
    "strong_AAGGTAAGT",
    "strong_CAGGTAAGT",
    "moderate_CAGGTGAGT",
    "brca_like_CAGGTATGG",
    "example_GTAAGTTTT",
    "weak_polyT"
  ]
}

CPU-NPU 一致性

{
  "max_abs_diff": 1.9073486328125e-06,
  "mean_abs_diff": 3.178914482759865e-07,
  "atol": 1e-05,
  "rtol": 1e-05,
  "shape_ok": true,
  "ranking_identical": true,
  "passed": true
}

性能

{
  "first_latency_ms": 149.8370347544551,
  "avg_latency_ms": 0.548607250675559,
  "min_ms": 0.5256100557744503,
  "max_ms": 0.57629169896245,
  "p50_ms": 0.5591907538473606,
  "sequences_per_s": 10936.78582011367,
  "bases_per_s": 98431.07238102304,
  "batch": 6,
  "seq_len": 9,
  "dtype": "float32",
  "stable_latencies_ms": [
    0.5615018308162689,
    0.5620219744741917,
    0.5591907538473606,
    0.57629169896245,
    0.5630119703710079,
    0.5290708504617214,
    0.5502812564373016,
    0.5256100557744503,
    0.5310410633683205,
    0.5280510522425175
  ]
}

自验证截图

Agent workflow

NPU device call

Model result

三张图片均由 xterm.js 根据本次真实日志生成。

已知限制

  • 仅对本次记录的模型 revision、输入契约、dtype 与 Ascend 环境完成验证。

标签

#NPU #Ascend #Ascend910