From 78a30b75621537d13bbc048d2e3cd7d5b8d66cd8 Mon Sep 17 00:00:00 2001
From: tzhouam <tzhouam@connect.ust.hk>
Date: Wed, 9 Sep 2026 16:22:01 +0800
Subject: [PATCH] edge results: codec stream diagnosis run (chain4) + offline
 parity diag

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
---
 .../20260909-160935/client_0.json             | 60 ++++++++++++++++
 .../20260909-160935/client_1.json             | 60 ++++++++++++++++
 .../20260909-160935/client_2.json             | 60 ++++++++++++++++
 .../20260909-160935/client_3.json             | 60 ++++++++++++++++
 .../20260909-160935/parity_diag.json          | 58 +++++++++++++++
 .../codec_stream/20260909-160935/summary.json | 71 +++++++++++++++++++
 6 files changed, 369 insertions(+)
 create mode 100644 benchmarks/tts/edge_results/codec_stream/20260909-160935/client_0.json
 create mode 100644 benchmarks/tts/edge_results/codec_stream/20260909-160935/client_1.json
 create mode 100644 benchmarks/tts/edge_results/codec_stream/20260909-160935/client_2.json
 create mode 100644 benchmarks/tts/edge_results/codec_stream/20260909-160935/client_3.json
 create mode 100644 benchmarks/tts/edge_results/codec_stream/20260909-160935/parity_diag.json
 create mode 100644 benchmarks/tts/edge_results/codec_stream/20260909-160935/summary.json

diff --git a/benchmarks/tts/edge_results/codec_stream/20260909-160935/client_0.json b/benchmarks/tts/edge_results/codec_stream/20260909-160935/client_0.json
new file mode 100644
index 000000000..79e27565c
--- /dev/null
+++ b/benchmarks/tts/edge_results/codec_stream/20260909-160935/client_0.json
@@ -0,0 +1,60 @@
+{
+ "codec_start": {
+  "type": "codec.start",
+  "version": 1,
+  "utterance_index": 0,
+  "sentence_index": 0,
+  "sentence_text": "Hello, welcome to the voice synthesis benchmark test.",
+  "model": "Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice",
+  "codebooks": 16,
+  "codebook_size": 2048,
+  "frame_rate_hz": 12.5,
+  "sample_rate": 24000,
+  "decoder_id": "qwen3_tts_tokenizer_v2_12hz",
+  "header_format": "<IHB"
+ },
+ "schedule": [
+  2,
+  4,
+  8,
+  16,
+  25
+ ],
+ "total_frames": 52,
+ "ttfa_ms": 17410.63924692571,
+ "playback_start_ms": 17410.639968235046,
+ "stall_at_ttfa_ms": 0,
+ "audio_s": 4.16,
+ "total_s": 17.91295762034133,
+ "rtf": 4.305999427966666,
+ "decode_ms_total": 1106.4688973128796,
+ "chunks": [
+  {
+   "idx": 0,
+   "t_ms": 17410.639968235046,
+   "arrival_ms": 16784.192154183984,
+   "frames": 25,
+   "samples": 48000,
+   "decode_ms": 626.3310848735273
+  },
+  {
+   "idx": 1,
+   "t_ms": 17731.141499243677,
+   "arrival_ms": 16784.192154183984,
+   "frames": 25,
+   "samples": 48000,
+   "decode_ms": 320.48439234495163
+  },
+  {
+   "idx": 2,
+   "t_ms": 17890.80695528537,
+   "arrival_ms": 16784.192154183984,
+   "frames": 2,
+   "samples": 3840,
+   "decode_ms": 159.65342009440064
+  }
+ ],
+ "device": "cpu",
+ "wav": "/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/edge_results/codec_stream/20260909-160935/client_0.wav",
+ "codes": "/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/edge_results/codec_stream/20260909-160935/client_0.codes.npy"
+}
\ No newline at end of file
diff --git a/benchmarks/tts/edge_results/codec_stream/20260909-160935/client_1.json b/benchmarks/tts/edge_results/codec_stream/20260909-160935/client_1.json
new file mode 100644
index 000000000..a10dcde6f
--- /dev/null
+++ b/benchmarks/tts/edge_results/codec_stream/20260909-160935/client_1.json
@@ -0,0 +1,60 @@
+{
+ "codec_start": {
+  "type": "codec.start",
+  "version": 1,
+  "utterance_index": 0,
+  "sentence_index": 0,
+  "sentence_text": "She said she would be here by noon, but nobody showed up.",
+  "model": "Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice",
+  "codebooks": 16,
+  "codebook_size": 2048,
+  "frame_rate_hz": 12.5,
+  "sample_rate": 24000,
+  "decoder_id": "qwen3_tts_tokenizer_v2_12hz",
+  "header_format": "<IHB"
+ },
+ "schedule": [
+  2,
+  4,
+  8,
+  16,
+  25
+ ],
+ "total_frames": 56,
+ "ttfa_ms": 2220.2195092104375,
+ "playback_start_ms": 2220.219925045967,
+ "stall_at_ttfa_ms": 0,
+ "audio_s": 4.48,
+ "total_s": 2.6397858639247715,
+ "rtf": 0.5892379160546365,
+ "decode_ms_total": 573.353567160666,
+ "chunks": [
+  {
+   "idx": 0,
+   "t_ms": 2220.219925045967,
+   "arrival_ms": 2065.352803096175,
+   "frames": 25,
+   "samples": 48000,
+   "decode_ms": 154.79304362088442
+  },
+  {
+   "idx": 1,
+   "t_ms": 2424.035610165447,
+   "arrival_ms": 2065.352803096175,
+   "frames": 25,
+   "samples": 48000,
+   "decode_ms": 203.80566967651248
+  },
+  {
+   "idx": 2,
+   "t_ms": 2638.8036208227277,
+   "arrival_ms": 2065.352803096175,
+   "frames": 6,
+   "samples": 11520,
+   "decode_ms": 214.7548538632691
+  }
+ ],
+ "device": "cpu",
+ "wav": "/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/edge_results/codec_stream/20260909-160935/client_1.wav",
+ "codes": "/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/edge_results/codec_stream/20260909-160935/client_1.codes.npy"
+}
\ No newline at end of file
diff --git a/benchmarks/tts/edge_results/codec_stream/20260909-160935/client_2.json b/benchmarks/tts/edge_results/codec_stream/20260909-160935/client_2.json
new file mode 100644
index 000000000..f5e52b0d5
--- /dev/null
+++ b/benchmarks/tts/edge_results/codec_stream/20260909-160935/client_2.json
@@ -0,0 +1,60 @@
+{
+ "codec_start": {
+  "type": "codec.start",
+  "version": 1,
+  "utterance_index": 0,
+  "sentence_index": 0,
+  "sentence_text": "The quick brown fox jumps over the lazy dog near the riverbank.",
+  "model": "Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice",
+  "codebooks": 16,
+  "codebook_size": 2048,
+  "frame_rate_hz": 12.5,
+  "sample_rate": 24000,
+  "decoder_id": "qwen3_tts_tokenizer_v2_12hz",
+  "header_format": "<IHB"
+ },
+ "schedule": [
+  2,
+  4,
+  8,
+  16,
+  25
+ ],
+ "total_frames": 57,
+ "ttfa_ms": 2197.10724754259,
+ "playback_start_ms": 2197.1078077331185,
+ "stall_at_ttfa_ms": 0,
+ "audio_s": 4.56,
+ "total_s": 2.876847210805863,
+ "rtf": 0.630887546229356,
+ "decode_ms_total": 831.5234775654972,
+ "chunks": [
+  {
+   "idx": 0,
+   "t_ms": 2197.1078077331185,
+   "arrival_ms": 2044.1418839618564,
+   "frames": 25,
+   "samples": 48000,
+   "decode_ms": 152.90560433641076
+  },
+  {
+   "idx": 1,
+   "t_ms": 2421.8555367551744,
+   "arrival_ms": 2044.1418839618564,
+   "frames": 25,
+   "samples": 48000,
+   "decode_ms": 224.73750729113817
+  },
+  {
+   "idx": 2,
+   "t_ms": 2875.748209655285,
+   "arrival_ms": 2044.1418839618564,
+   "frames": 7,
+   "samples": 13440,
+   "decode_ms": 453.8803659379482
+  }
+ ],
+ "device": "cpu",
+ "wav": "/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/edge_results/codec_stream/20260909-160935/client_2.wav",
+ "codes": "/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/edge_results/codec_stream/20260909-160935/client_2.codes.npy"
+}
\ No newline at end of file
diff --git a/benchmarks/tts/edge_results/codec_stream/20260909-160935/client_3.json b/benchmarks/tts/edge_results/codec_stream/20260909-160935/client_3.json
new file mode 100644
index 000000000..13700b9ed
--- /dev/null
+++ b/benchmarks/tts/edge_results/codec_stream/20260909-160935/client_3.json
@@ -0,0 +1,60 @@
+{
+ "codec_start": {
+  "type": "codec.start",
+  "version": 1,
+  "utterance_index": 0,
+  "sentence_index": 0,
+  "sentence_text": "I can't believe how beautiful the sunset looks from up here on the mountain.",
+  "model": "Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice",
+  "codebooks": 16,
+  "codebook_size": 2048,
+  "frame_rate_hz": 12.5,
+  "sample_rate": 24000,
+  "decoder_id": "qwen3_tts_tokenizer_v2_12hz",
+  "header_format": "<IHB"
+ },
+ "schedule": [
+  2,
+  4,
+  8,
+  16,
+  25
+ ],
+ "total_frames": 67,
+ "ttfa_ms": 2587.735196109861,
+ "playback_start_ms": 2587.735620792955,
+ "stall_at_ttfa_ms": 0,
+ "audio_s": 5.36,
+ "total_s": 3.123869556002319,
+ "rtf": 0.5828114843287908,
+ "decode_ms_total": 704.384888522327,
+ "chunks": [
+  {
+   "idx": 0,
+   "t_ms": 2587.735620792955,
+   "arrival_ms": 2418.170887976885,
+   "frames": 25,
+   "samples": 48000,
+   "decode_ms": 169.50282221660018
+  },
+  {
+   "idx": 1,
+   "t_ms": 2822.7644758298993,
+   "arrival_ms": 2418.170887976885,
+   "frames": 25,
+   "samples": 48000,
+   "decode_ms": 235.01680931076407
+  },
+  {
+   "idx": 2,
+   "t_ms": 3122.6435392163694,
+   "arrival_ms": 2418.170887976885,
+   "frames": 17,
+   "samples": 32640,
+   "decode_ms": 299.8652569949627
+  }
+ ],
+ "device": "cpu",
+ "wav": "/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/edge_results/codec_stream/20260909-160935/client_3.wav",
+ "codes": "/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/edge_results/codec_stream/20260909-160935/client_3.codes.npy"
+}
\ No newline at end of file
diff --git a/benchmarks/tts/edge_results/codec_stream/20260909-160935/parity_diag.json b/benchmarks/tts/edge_results/codec_stream/20260909-160935/parity_diag.json
new file mode 100644
index 000000000..e95e64eab
--- /dev/null
+++ b/benchmarks/tts/edge_results/codec_stream/20260909-160935/parity_diag.json
@@ -0,0 +1,58 @@
+[
+ {
+  "i": 0,
+  "frames": 52,
+  "exact_vs_client_chunked": {
+   "snr_db": 75.30462286668491,
+   "corr": 0.9999999963199535,
+   "lag": 0
+  },
+  "full_server_vs_exact": {
+   "snr_db": -3.0102342111930867,
+   "corr": -0.004136945264628016,
+   "lag": -1920
+  }
+ },
+ {
+  "i": 1,
+  "frames": 56,
+  "exact_vs_client_chunked": {
+   "snr_db": 75.94970682010397,
+   "corr": 0.9999999968181664,
+   "lag": 0
+  },
+  "full_server_vs_exact": {
+   "snr_db": -2.944756779166787,
+   "corr": 0.012639352627620306,
+   "lag": -1920
+  }
+ },
+ {
+  "i": 2,
+  "frames": 57,
+  "exact_vs_client_chunked": {
+   "snr_db": 74.71863619940385,
+   "corr": 0.9999999957637626,
+   "lag": 0
+  },
+  "full_server_vs_exact": {
+   "snr_db": -3.039177960515268,
+   "corr": -0.0073217557348321,
+   "lag": -1920
+  }
+ },
+ {
+  "i": 3,
+  "frames": 67,
+  "exact_vs_client_chunked": {
+   "snr_db": 72.65030455154775,
+   "corr": 0.9999999932262861,
+   "lag": 0
+  },
+  "full_server_vs_exact": {
+   "snr_db": -2.9956981500244613,
+   "corr": 0.0008606761784026995,
+   "lag": -1920
+  }
+ }
+]
\ No newline at end of file
diff --git a/benchmarks/tts/edge_results/codec_stream/20260909-160935/summary.json b/benchmarks/tts/edge_results/codec_stream/20260909-160935/summary.json
new file mode 100644
index 000000000..0963b22b7
--- /dev/null
+++ b/benchmarks/tts/edge_results/codec_stream/20260909-160935/summary.json
@@ -0,0 +1,71 @@
+{
+ "model": "Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice",
+ "timestamp": "20260909-160935",
+ "prompts": [
+  "Hello, welcome to the voice synthesis benchmark test.",
+  "She said she would be here by noon, but nobody showed up.",
+  "The quick brown fox jumps over the lazy dog near the riverbank.",
+  "I can't believe how beautiful the sunset looks from up here on the mountain."
+ ],
+ "client_device": "cpu",
+ "rows": [
+  {
+   "i": 0,
+   "text": "Hello, welcome to the voice synthesis benchmark test.",
+   "frames": 52,
+   "client_ttfa_ms": 17410.63924692571,
+   "client_playback_start_ms": 17410.639968235046,
+   "client_rtf": 4.305999427966666,
+   "client_decode_ms": 1106.4688973128796,
+   "client_audio_s": 4.16,
+   "full_server_s": 16.10692495200783,
+   "full_audio_s": 4.16,
+   "snr_db": -3.03693684919166,
+   "length_ratio": 1.0
+  },
+  {
+   "i": 1,
+   "text": "She said she would be here by noon, but nobody showed up.",
+   "frames": 56,
+   "client_ttfa_ms": 2220.2195092104375,
+   "client_playback_start_ms": 2220.219925045967,
+   "client_rtf": 0.5892379160546365,
+   "client_decode_ms": 573.353567160666,
+   "client_audio_s": 4.48,
+   "full_server_s": 2.1515910821035504,
+   "full_audio_s": 4.48,
+   "snr_db": -2.953329833875654,
+   "length_ratio": 1.0
+  },
+  {
+   "i": 2,
+   "text": "The quick brown fox jumps over the lazy dog near the riverbank.",
+   "frames": 57,
+   "client_ttfa_ms": 2197.10724754259,
+   "client_playback_start_ms": 2197.1078077331185,
+   "client_rtf": 0.630887546229356,
+   "client_decode_ms": 831.5234775654972,
+   "client_audio_s": 4.56,
+   "full_server_s": 2.14743230631575,
+   "full_audio_s": 4.56,
+   "snr_db": -3.011027358981732,
+   "length_ratio": 1.0
+  },
+  {
+   "i": 3,
+   "text": "I can't believe how beautiful the sunset looks from up here on the mountain.",
+   "frames": 67,
+   "client_ttfa_ms": 2587.735196109861,
+   "client_playback_start_ms": 2587.735620792955,
+   "client_rtf": 0.5828114843287908,
+   "client_decode_ms": 704.384888522327,
+   "client_audio_s": 5.36,
+   "full_server_s": 2.551622844301164,
+   "full_audio_s": 5.36,
+   "snr_db": -2.9195661162834483,
+   "length_ratio": 1.0
+  }
+ ],
+ "talker_server_start_s": 97.05230398196727,
+ "full_server_start_s": 166.07538488693535
+}
\ No newline at end of file