Update model.tflite with dynamic-range quantization and fix architecture/parameter metadata

README.md

@@ -37,8 +37,8 @@ Multi-source attention sequence-to-sequence **Textual Echo Cancellation (TEC)**
 
 - **Condition**: Multiple interfering voices (LibriTTS + VCTK)
 - **Dataset Splits**: 400 training utterances (`train`), 100 validation utterances (`dev`), 100 `test-clean` utterances, and 100 `test-other` utterances at 24 kHz mixed at 0 dB SNR with synthetic image-method room impulse responses (RT60 = 0.25 s).
-- **Architecture**: Speech encoder (`SpeechEncoderV1`: 2× Conv2D + 4× Bi-LSTM + 1× Bi-CLSTM with stride `(2, 2)`), text encoder (`TtsEncoderV2`: 98-symbol character vocabulary `VOCAB_SIZE = 98`, 3× Conv1D + 1× Bi-LSTM, for TEC models), multi-source additive location-sensitive + content attention decoder (`SpeechEnhancementDecoderV1`: 1× autoregressive ZoneOut LSTM + 2× post-attention LSTM + 5×5 ConvTranspose2D + 2-layer Highway Postnet).
-- **Trainable Parameters**: `3,962,728`
+- **Architecture**: Speech encoder (`SpeechEncoderV1`: 2× strided 3×3 Conv2D with stride `(2, 2)`, 1× Bi-CLSTM with stride `(1, 1)`, and 3× Bi-LSTM layers), text encoder (`TtsEncoderV2`: 98-symbol ASCII character vocabulary `VOCAB_SIZE = 98`, 512-dim embedding, 3× Conv1D + 1× Bi-LSTM, for TEC models), and multi-source GMM monotonic attention decoder (`MultiSourceFbeDecoderV1` / `FbeDecoderV1`: 2×256 Pre-Net, 2-layer autoregressive LSTM, frame reduction factor `r = 4`, and 5-layer 1D convolutional `PostEditConvNet`).
+- **Trainable Parameters**: `18,206,463` (`18,214,084` total model variables; `218.51 MB` checkpoint shard; `19.02 MB` dynamic-range quantized `model.tflite`)
 - **Checkpoint Selection**: Validation-loss early stopping (`best.ckpt` selected on the 100-utterance `dev` split over 500 training steps with batch size 8 and Adam `lr = 1e-3`).
 
 ---
@@ -167,18 +167,18 @@ model_dir = snapshot_download(repo_id="wq2012/tec_multi_interfering")
 tflite_path = os.path.join(model_dir, "model.tflite")
 
 dummy_mel = np.zeros((100, 128), dtype=np.float32)
-enhanced_mel = inference.run_tflite_inference(
+tflite_out = inference.run_tflite_inference(
     tflite_path=tflite_path,
     mixed_mel=dummy_mel,
     interfering_text="currently in mountain view it is 72 degrees",
 )
-print("TFLite predicted log-Mel shape:", enhanced_mel.shape)
+print("TFLite predicted log-Mel shape:", tflite_out["predicted_mel"].shape)
 ```
 
 Expected console output:
 
 ```text
-TFLite predicted log-Mel shape: (1, 100, 128)
+TFLite predicted log-Mel shape: (100, 128)
 ```
 
 ---

evaluation_metrics.json

@@ -1,7 +1,7 @@
 {
   "model": "TecMultiInterfering",
   "condition": "Multiple interfering voices (LibriTTS + VCTK)",
-  "trainable_parameters": 3962728,
+  "trainable_parameters": 18206463,
   "test_clean": {
     "method": "TecMultiInterfering",
     "text_condition": "matched",
@@ -3995,5 +3995,7 @@
         "As Master thinks best, Charles Gordon, leader of the class, the council declined to comment."
       ]
     }
-  }
+  },
+  "total_model_variables": 18214084,
+  "tflite_quantized_bytes": 19016856
 }

model.tflite

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6d78a4a0473acb5141a61d1c5e889a172f5f3e8096592c5499c6c68bba36b2cd
-size 73675328
+oid sha256:db1ac4fa5d69d19ddff13373c48288107f110417141fa0228a0014d6444e72ab
+size 19016856