first commit

checkpoint_85000.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2af1fb646f30a382b1a299af4336b283decefa7cb53aebc62140c5f435e6bb56
+size 1043207181

prepare_model.py

@@ -0,0 +1,67 @@
+import json
+import os
+import subprocess
+
+# Load the data from the provided dictionary
+data = {
+  "VCTK_p294": {"age": 33, "gender": "F", "accents": "American", "region": "San Francisco", "comments": ""},
+  "VCTK_p297": {"age": 20, "gender": "F", "accents": "American", "region": "New York", "comments": ""},
+  "VCTK_p299": {"age": 25, "gender": "F", "accents": "American", "region": "California", "comments": ""},
+  "VCTK_p300": {"age": 23, "gender": "F", "accents": "American", "region": "California", "comments": ""},
+  "VCTK_p301": {"age": 23, "gender": "F", "accents": "American", "region": "North Carolina", "comments": ""},
+  "VCTK_p305": {"age": 19, "gender": "F", "accents": "American", "region": "Philadelphia", "comments": ""},
+  "VCTK_p306": {"age": 21, "gender": "F", "accents": "American", "region": "New York", "comments": ""},
+  "VCTK_p308": {"age": 18, "gender": "F", "accents": "American", "region": "Alabama", "comments": ""},
+  "VCTK_p310": {"age": 21, "gender": "F", "accents": "American", "region": "Tennessee", "comments": ""},
+  "VCTK_p318": {"age": 32, "gender": "F", "accents": "American", "region": "Napa", "comments": ""},
+  "VCTK_p329": {"age": 23, "gender": "F", "accents": "American", "region": "America", "comments": ""},
+  "VCTK_p330": {"age": 26, "gender": "F", "accents": "American", "region": "America", "comments": ""},
+  "VCTK_p333": {"age": 19, "gender": "F", "accents": "American", "region": "Indiana", "comments": ""},
+  "VCTK_p361": {"age": 19, "gender": "F", "accents": "American", "region": "New Jersey", "comments": ""},
+  "VCTK_p362": {"age": 29, "gender": "F", "accents": "American", "region": "America", "comments": ""},
+  "VCTK_p339": {"age": 21, "gender": "F", "accents": "American", "region": "Pennsylvania", "comments": ""},
+  "VCTK_p341": {"age": 26, "gender": "F", "accents": "American", "region": "Ohio", "comments": ""}
+}
+
+
+# Convert the data to JSON format
+json_data = json.dumps(data, indent=2)
+
+# Save the JSON data to a file
+with open('speakers-log.json', 'w') as file:
+    file.write(json_data)
+
+# Run the TTS command to get the speaker indices
+command = "tts --model_path checkpoint_85000.pth --config_path config.json --list_speaker_idxs | grep -vE '^(\s*\||\s*>|\s*$)'"
+output = subprocess.check_output(command, shell=True, text=True)
+
+# Parse the JSON output into a Python dictionary
+speaker_indices = eval(output)
+
+# Load the speaker IDs from speakers.json
+with open('speakers-log.json', 'r') as file:
+    speaker_ids = json.load(file)
+
+
+for speaker_idx in speaker_indices:
+    # # Remove the 'VCTK_' prefix
+    speaker_id = speaker_idx
+    # speaker_id = speaker_idx.replace('VCTK_', '')
+
+    # Lookup the speaker ID in the loaded speaker IDs
+    if speaker_id in speaker_ids:
+        speaker_id_json = speaker_ids[speaker_id]
+    else:
+        continue
+
+    # # Generate the TTS command to create the audio file
+    text = f"Hello, I am from {speaker_id_json['region']}. I hope that you will select my voice for your project. Thank you."
+    # # make samples directory if it doesn't exist
+    if not os.path.exists("samples"):
+        os.makedirs("samples")
+
+    out_path = f"samples/{speaker_id}.wav"
+    tts_command = f"tts --text \"{text}\" --model_path checkpoint_85000.pth --language_idx en --config_path config.json --speaker_idx \"{speaker_id}\" --out_path {out_path}"
+
+    # Execute the TTS command
+    os.system(tts_command)

speakers-log.json

@@ -0,0 +1,121 @@
+{
+  "VCTK_p294": {
+    "age": 33,
+    "gender": "F",
+    "accents": "American",
+    "region": "San Francisco",
+    "comments": ""
+  },
+  "VCTK_p297": {
+    "age": 20,
+    "gender": "F",
+    "accents": "American",
+    "region": "New York",
+    "comments": ""
+  },
+  "VCTK_p299": {
+    "age": 25,
+    "gender": "F",
+    "accents": "American",
+    "region": "California",
+    "comments": ""
+  },
+  "VCTK_p300": {
+    "age": 23,
+    "gender": "F",
+    "accents": "American",
+    "region": "California",
+    "comments": ""
+  },
+  "VCTK_p301": {
+    "age": 23,
+    "gender": "F",
+    "accents": "American",
+    "region": "North Carolina",
+    "comments": ""
+  },
+  "VCTK_p305": {
+    "age": 19,
+    "gender": "F",
+    "accents": "American",
+    "region": "Philadelphia",
+    "comments": ""
+  },
+  "VCTK_p306": {
+    "age": 21,
+    "gender": "F",
+    "accents": "American",
+    "region": "New York",
+    "comments": ""
+  },
+  "VCTK_p308": {
+    "age": 18,
+    "gender": "F",
+    "accents": "American",
+    "region": "Alabama",
+    "comments": ""
+  },
+  "VCTK_p310": {
+    "age": 21,
+    "gender": "F",
+    "accents": "American",
+    "region": "Tennessee",
+    "comments": ""
+  },
+  "VCTK_p318": {
+    "age": 32,
+    "gender": "F",
+    "accents": "American",
+    "region": "Napa",
+    "comments": ""
+  },
+  "VCTK_p329": {
+    "age": 23,
+    "gender": "F",
+    "accents": "American",
+    "region": "America",
+    "comments": ""
+  },
+  "VCTK_p330": {
+    "age": 26,
+    "gender": "F",
+    "accents": "American",
+    "region": "America",
+    "comments": ""
+  },
+  "VCTK_p333": {
+    "age": 19,
+    "gender": "F",
+    "accents": "American",
+    "region": "Indiana",
+    "comments": ""
+  },
+  "VCTK_p361": {
+    "age": 19,
+    "gender": "F",
+    "accents": "American",
+    "region": "New Jersey",
+    "comments": ""
+  },
+  "VCTK_p362": {
+    "age": 29,
+    "gender": "F",
+    "accents": "American",
+    "region": "America",
+    "comments": ""
+  },
+  "VCTK_p339": {
+    "age": 21,
+    "gender": "F",
+    "accents": "American",
+    "region": "Pennsylvania",
+    "comments": ""
+  },
+  "VCTK_p341": {
+    "age": 26,
+    "gender": "F",
+    "accents": "American",
+    "region": "Ohio",
+    "comments": ""
+  }
+}

supplemental/config_se.json

@@ -0,0 +1,121 @@
+{
+    "model": "speaker_encoder",
+    "run_name": "speaker_encoder",
+    "run_description": "resnet speaker encoder trained with commonvoice all languages dev and train, Voxceleb 1 dev and Voxceleb 2 dev",
+    "epochs": 100000,
+    "batch_size": null,
+    "eval_batch_size": null,
+    "mixed_precision": false,
+    "run_eval": true,
+    "test_delay_epochs": 0,
+    "print_eval": false,
+    "print_step": 50,
+    "tb_plot_step": 100,
+    "tb_model_param_stats": false,
+    "save_step": 1000,
+    "checkpoint": true,
+    "keep_all_best": false,
+    "keep_after": 10000,
+    "num_loader_workers": 8,
+    "num_val_loader_workers": 0,
+    "use_noise_augment": false,
+    "output_path": "../checkpoints/speaker_encoder/language_balanced/normalized/angleproto-4-samples-by-speakers/",
+    "distributed_backend": "nccl",
+    "distributed_url": "tcp://localhost:54321",
+    "audio": {
+        "fft_size": 512,
+        "win_length": 400,
+        "hop_length": 160,
+        "frame_shift_ms": null,
+        "frame_length_ms": null,
+        "stft_pad_mode": "reflect",
+        "sample_rate": 16000,
+        "resample": false,
+        "preemphasis": 0.97,
+        "ref_level_db": 20,
+        "do_sound_norm": false,
+        "do_trim_silence": false,
+        "trim_db": 60,
+        "power": 1.5,
+        "griffin_lim_iters": 60,
+        "num_mels": 64,
+        "mel_fmin": 0.0,
+        "mel_fmax": 8000.0,
+        "spec_gain": 20,
+        "signal_norm": false,
+        "min_level_db": -100,
+        "symmetric_norm": false,
+        "max_norm": 4.0,
+        "clip_norm": false,
+        "stats_path": null,
+        "do_rms_norm": true,
+        "db_level": -27.0
+    },
+    "datasets": [
+        {
+            "name": "voxceleb2",
+            "path": "/workspace/scratch/ecasanova/datasets/VoxCeleb/vox2_dev_aac/",
+            "meta_file_train": null,
+            "ununsed_speakers": null,
+            "meta_file_val": null,
+            "meta_file_attn_mask": "",
+            "language": "voxceleb"
+        }
+    ],
+    "model_params": {
+        "model_name": "resnet",
+        "input_dim": 64,
+        "use_torch_spec": true,
+        "log_input": true,
+        "proj_dim": 512
+    },
+    "audio_augmentation": {
+        "p": 0.5,
+        "rir": {
+            "rir_path": "/workspace/store/ecasanova/ComParE/RIRS_NOISES/simulated_rirs/",
+            "conv_mode": "full"
+        },
+        "additive": {
+            "sounds_path": "/workspace/store/ecasanova/ComParE/musan/",
+            "speech": {
+                "min_snr_in_db": 13,
+                "max_snr_in_db": 20,
+                "min_num_noises": 1,
+                "max_num_noises": 1
+            },
+            "noise": {
+                "min_snr_in_db": 0,
+                "max_snr_in_db": 15,
+                "min_num_noises": 1,
+                "max_num_noises": 1
+            },
+            "music": {
+                "min_snr_in_db": 5,
+                "max_snr_in_db": 15,
+                "min_num_noises": 1,
+                "max_num_noises": 1
+            }
+        },
+        "gaussian": {
+            "p": 0.0,
+            "min_amplitude": 0.0,
+            "max_amplitude": 1e-05
+        }
+    },
+    "storage": {
+        "sample_from_storage_p": 0.5,
+        "storage_size": 40
+    },
+    "max_train_step": 1000000,
+    "loss": "angleproto",
+    "grad_clip": 3.0,
+    "lr": 0.0001,
+    "lr_decay": false,
+    "warmup_steps": 4000,
+    "wd": 1e-06,
+    "steps_plot_stats": 100,
+    "num_speakers_in_batch": 100,
+    "num_utters_per_speaker": 4,
+    "skip_speakers": true,
+    "voice_len": 2.0
+}

supplemental/language_ids.json

@@ -0,0 +1,5 @@
+{
+    "en": 0,
+    "fr-fr": 1,
+    "pt-br": 2
+}

supplemental/model_se.pth.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8f96efb20cbeeefd81fd8336d7f0155bf8902f82f9474e58ccb19d9e12345172
+size 44610930

supplemental/speaker_ids.json

@@ -0,0 +1,25 @@
+{
+    "female-en-5": 0,
+    "female-en-5\n": 1,
+    "female-pt-4\n": 2,
+    "male-en-2": 3,
+    "male-en-2\n": 4,
+    "male-pt-3\n": 5,
+    "VCTK_p294": 6,
+    "VCTK_p297": 7,
+    "VCTK_p299": 8,
+    "VCTK_p300": 9,
+    "VCTK_p301": 10,
+    "VCTK_p305": 11,
+    "VCTK_p306": 12,
+    "VCTK_p308": 13,
+    "VCTK_p310": 14,
+    "VCTK_p318": 15,
+    "VCTK_p329": 16,
+    "VCTK_p330": 17,
+    "VCTK_p333": 18,
+    "VCTK_p339": 19,
+    "VCTK_p341": 20,
+    "VCTK_p361": 21,
+    "VCTK_p362": 22
+}

supplemental/speakers-combined.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c1cc64ed8d2b769cafcfef9bbc89ea6579994a3311ebb3ab585d30b4fe2fb21
+size 31821551

supplemental/speakers-dataset.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:34564e100952d66a78676f41a2f4b8a236e3f3e0222fdbc2b97ae07d98c1e33d
+size 31640687