Upload 3 files
Browse files- README.md +52 -0
- dataset_info.json +131 -0
- sesame_tts_synthetic_combined.parquet +3 -0
README.md
ADDED
|
@@ -0,0 +1,52 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Combined Fine-tuned Sesame CSM 1B Synthetic Speech Dataset
|
| 2 |
+
|
| 3 |
+
## Data Format
|
| 4 |
+
|
| 5 |
+
Each sample contains:
|
| 6 |
+
- **audio**: Audio array with sampling_rate (24kHz)
|
| 7 |
+
- **text**: Original transcript text
|
| 8 |
+
- **speaker_id**: Speaker identifier (F04, M02, FC02, MC01, F02, M04, 211, 4014)
|
| 9 |
+
- **corpus**: Source corpus (TORGO, UA-Speech, LibriSpeech)
|
| 10 |
+
- **condition**: Speaker condition (Dysarthric, Healthy)
|
| 11 |
+
- **model_name**: Fine-tuned model name
|
| 12 |
+
- **model_type**: "sesame_csm_1b_adapter"
|
| 13 |
+
- **base_model**: Base model used (unsloth/csm-1b)
|
| 14 |
+
- **adapter_path**: HuggingFace adapter path
|
| 15 |
+
- **duration**: Audio duration in seconds
|
| 16 |
+
- **sample_rate**: Audio sample rate (24000)
|
| 17 |
+
- **sample_index**: Sample index number
|
| 18 |
+
- **audio_file**: Original audio filename
|
| 19 |
+
- **generated_at**: Generation timestamp
|
| 20 |
+
|
| 21 |
+
## Usage Example
|
| 22 |
+
|
| 23 |
+
```python
|
| 24 |
+
from datasets import Dataset
|
| 25 |
+
import pandas as pd
|
| 26 |
+
|
| 27 |
+
# Load combined dataset
|
| 28 |
+
dataset = Dataset.from_parquet("sesame_tts_synthetic_combined.parquet")
|
| 29 |
+
|
| 30 |
+
print(f"Total samples: {len(dataset)}")
|
| 31 |
+
print(f"Speakers: {set(dataset['speaker_id'])}")
|
| 32 |
+
|
| 33 |
+
# Filter by model type
|
| 34 |
+
sesame_samples = dataset.filter(lambda x: x['model_type'] == 'sesame_csm_1b_adapter')
|
| 35 |
+
print(f"Sesame CSM samples: {len(sesame_samples)}")
|
| 36 |
+
|
| 37 |
+
# Filter by condition
|
| 38 |
+
dysarthric_samples = dataset.filter(lambda x: x['condition'] == 'Dysarthric')
|
| 39 |
+
healthy_samples = dataset.filter(lambda x: x['condition'] == 'Healthy')
|
| 40 |
+
|
| 41 |
+
print(f"Dysarthric: {len(dysarthric_samples)}, Healthy: {len(healthy_samples)}")
|
| 42 |
+
|
| 43 |
+
# Play sample
|
| 44 |
+
sample = dataset[0]
|
| 45 |
+
from IPython.display import Audio
|
| 46 |
+
Audio(sample['audio']['array'], rate=sample['audio']['sampling_rate'])
|
| 47 |
+
|
| 48 |
+
# Convert to DataFrame for analysis
|
| 49 |
+
df = dataset.to_pandas()
|
| 50 |
+
print(df.groupby(['corpus', 'condition'])['duration'].agg(['count', 'sum', 'mean']))
|
| 51 |
+
```
|
| 52 |
+
|
dataset_info.json
ADDED
|
@@ -0,0 +1,131 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"dataset_name": "combined_sesame_tts_synthetic",
|
| 3 |
+
"dataset_size": 759,
|
| 4 |
+
"description": "Combined fine-tuned Sesame CSM 1B synthetic speech dataset with LoRA adapters for pathological speech from 8 speakers",
|
| 5 |
+
"speakers": {
|
| 6 |
+
"MC01": {
|
| 7 |
+
"name": "TORGO Healthy Male",
|
| 8 |
+
"corpus": "TORGO",
|
| 9 |
+
"condition": "Healthy",
|
| 10 |
+
"gender": "Male",
|
| 11 |
+
"samples": 97,
|
| 12 |
+
"adapter_path": "resproj007/torgo_healthy_male_sesame_1b_MC01"
|
| 13 |
+
},
|
| 14 |
+
"FC02": {
|
| 15 |
+
"name": "TORGO Healthy Female",
|
| 16 |
+
"corpus": "TORGO",
|
| 17 |
+
"condition": "Healthy",
|
| 18 |
+
"gender": "Female",
|
| 19 |
+
"samples": 98,
|
| 20 |
+
"adapter_path": "resproj007/torgo_healthy_female_sesame_1b_FC02"
|
| 21 |
+
},
|
| 22 |
+
"F02": {
|
| 23 |
+
"name": "UA-Speech Female",
|
| 24 |
+
"corpus": "UA-Speech",
|
| 25 |
+
"condition": "Dysarthric",
|
| 26 |
+
"gender": "Female",
|
| 27 |
+
"samples": 182,
|
| 28 |
+
"adapter_path": "resproj007/uaspeech_female_sesame_1b_F02"
|
| 29 |
+
},
|
| 30 |
+
"M04": {
|
| 31 |
+
"name": "UA-Speech Male",
|
| 32 |
+
"corpus": "UA-Speech",
|
| 33 |
+
"condition": "Dysarthric",
|
| 34 |
+
"gender": "Male",
|
| 35 |
+
"samples": 184,
|
| 36 |
+
"adapter_path": "resproj007/uaspeech_male_sesame_1b_M04"
|
| 37 |
+
},
|
| 38 |
+
"4014": {
|
| 39 |
+
"name": "LibriSpeech Male",
|
| 40 |
+
"corpus": "LibriSpeech",
|
| 41 |
+
"condition": "Healthy",
|
| 42 |
+
"gender": "Male",
|
| 43 |
+
"samples": 30,
|
| 44 |
+
"adapter_path": "resproj007/librispeech_male_sesame_1b_4014"
|
| 45 |
+
},
|
| 46 |
+
"M02": {
|
| 47 |
+
"name": "TORGO Dysarthric Male",
|
| 48 |
+
"corpus": "TORGO",
|
| 49 |
+
"condition": "Dysarthric",
|
| 50 |
+
"gender": "Male",
|
| 51 |
+
"samples": 69,
|
| 52 |
+
"adapter_path": "resproj007/torgo_dysarthric_male_sesame_1b_M02"
|
| 53 |
+
},
|
| 54 |
+
"F04": {
|
| 55 |
+
"name": "TORGO Dysarthric Female",
|
| 56 |
+
"corpus": "TORGO",
|
| 57 |
+
"condition": "Dysarthric",
|
| 58 |
+
"gender": "Female",
|
| 59 |
+
"samples": 69,
|
| 60 |
+
"adapter_path": "resproj007/torgo_dysarthric_female_sesame_1b_F04"
|
| 61 |
+
},
|
| 62 |
+
"211": {
|
| 63 |
+
"name": "LibriSpeech Female",
|
| 64 |
+
"corpus": "LibriSpeech",
|
| 65 |
+
"condition": "Healthy",
|
| 66 |
+
"gender": "Female",
|
| 67 |
+
"samples": 30,
|
| 68 |
+
"adapter_path": "resproj007/librispeech_female_sesame_1b_211"
|
| 69 |
+
}
|
| 70 |
+
},
|
| 71 |
+
"corpora": [
|
| 72 |
+
"TORGO",
|
| 73 |
+
"UA-Speech",
|
| 74 |
+
"LibriSpeech"
|
| 75 |
+
],
|
| 76 |
+
"conditions": [
|
| 77 |
+
"Dysarthric",
|
| 78 |
+
"Healthy"
|
| 79 |
+
],
|
| 80 |
+
"model_info": {
|
| 81 |
+
"base_model": "unsloth/csm-1b",
|
| 82 |
+
"model_type": "sesame_csm_1b_adapter",
|
| 83 |
+
"fine_tuning_method": "lora_adapter",
|
| 84 |
+
"architecture": "sesame_csm_1b",
|
| 85 |
+
"training_corpora": [
|
| 86 |
+
"TORGO",
|
| 87 |
+
"UA-Speech",
|
| 88 |
+
"LibriSpeech"
|
| 89 |
+
]
|
| 90 |
+
},
|
| 91 |
+
"audio_info": {
|
| 92 |
+
"sampling_rate": 24000,
|
| 93 |
+
"format": "wav",
|
| 94 |
+
"dtype": "float32",
|
| 95 |
+
"channels": 1
|
| 96 |
+
},
|
| 97 |
+
"features": {
|
| 98 |
+
"audio": {
|
| 99 |
+
"feature_type": "Audio",
|
| 100 |
+
"sampling_rate": 24000
|
| 101 |
+
},
|
| 102 |
+
"text": "string",
|
| 103 |
+
"speaker_id": "string",
|
| 104 |
+
"corpus": "string",
|
| 105 |
+
"condition": "string",
|
| 106 |
+
"model_name": "string",
|
| 107 |
+
"model_type": "string",
|
| 108 |
+
"base_model": "string",
|
| 109 |
+
"adapter_path": "string",
|
| 110 |
+
"duration": "float64",
|
| 111 |
+
"sample_rate": "int64",
|
| 112 |
+
"sample_index": "int64",
|
| 113 |
+
"audio_file": "string",
|
| 114 |
+
"generated_at": "string"
|
| 115 |
+
},
|
| 116 |
+
"statistics": {
|
| 117 |
+
"total_test_samples": 800,
|
| 118 |
+
"successfully_generated": 759,
|
| 119 |
+
"skipped_samples": 41,
|
| 120 |
+
"overall_success_rate": "94.9%",
|
| 121 |
+
"total_speakers": 8,
|
| 122 |
+
"total_duration_hours": 0.8077777777777778,
|
| 123 |
+
"avg_duration_per_sample": 3.831357048748353
|
| 124 |
+
},
|
| 125 |
+
"generation_info": {
|
| 126 |
+
"generated_date": "2025-09-14T15:08:08.091968",
|
| 127 |
+
"converter": "SesameTTSParquetConverter",
|
| 128 |
+
"version": "1.0",
|
| 129 |
+
"source_directory": "sesame_synthetic_data"
|
| 130 |
+
}
|
| 131 |
+
}
|
sesame_tts_synthetic_combined.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8bce9d40f31336d8fc762343e8e4ff24018927cff384798f669833d955fc09ac
|
| 3 |
+
size 133589427
|