claroche1 commited on
Commit
3c8c7ec
·
verified ·
1 Parent(s): 5ea73b5

upload LiSenNet checkpoint + ONNX (fp32, int8 dyn/static)

Browse files
Files changed (1) hide show
  1. README.md +2 -3
README.md CHANGED
@@ -51,8 +51,7 @@ reproduction lands within ~0.06 of the paper's reported ~3.07.
51
  | ------------------------- | --- |
52
  | `g_best` | PyTorch checkpoint (`{"generator": state_dict}`) |
53
  | `g_best_fp32.onnx` | FP32 ONNX of the mask sub-network: `feat (B,3,T,F)` -> `est_mag (B,T,F)`, dynamic batch/time |
54
- | `g_best_int8_dyn.onnx` | Dynamic weight-only int8 (robust, near loss-free) — **recommended** |
55
- | `g_best_int8_static.onnx` | Static full int8 (QDQ, per-tensor, percentile calibration; QAT-grade) |
56
  | `config.json` | Training + architecture config (STFT, channels, blocks) |
57
 
58
  The ONNX graphs cover only the mask sub-network (the pure tensor-op core). The
@@ -89,7 +88,7 @@ from huggingface_hub import hf_hub_download
89
 
90
  REPO = "claroche1/LiSenNet"
91
  sess = ort.InferenceSession(
92
- hf_hub_download(REPO, "g_best_int8_dyn.onnx"), # or _fp32 / _int8_static
93
  providers=["CPUExecutionProvider"],
94
  )
95
  # input feat: (B, 3, T, F=257) = [compressed_mag, group_delay/pi, ifd/pi]
 
51
  | ------------------------- | --- |
52
  | `g_best` | PyTorch checkpoint (`{"generator": state_dict}`) |
53
  | `g_best_fp32.onnx` | FP32 ONNX of the mask sub-network: `feat (B,3,T,F)` -> `est_mag (B,T,F)`, dynamic batch/time |
54
+ | `g_best_int8_static.onnx` | Static full int8 (QDQ, per-tensor, percentile calibration; embedded-deployable) |
 
55
  | `config.json` | Training + architecture config (STFT, channels, blocks) |
56
 
57
  The ONNX graphs cover only the mask sub-network (the pure tensor-op core). The
 
88
 
89
  REPO = "claroche1/LiSenNet"
90
  sess = ort.InferenceSession(
91
+ hf_hub_download(REPO, "g_best_int8_static.onnx"), # or g_best_fp32.onnx
92
  providers=["CPUExecutionProvider"],
93
  )
94
  # input feat: (B, 3, T, F=257) = [compressed_mag, group_delay/pi, ifd/pi]