configs/infer.yaml.example

# Inference Configuration# Inference Configuration

# Predict gender and dialect from audio# Architecture: WavLM + Attentive Pooling + LayerNorm + Deeper Heads

# Copy this file to infer.yaml and update paths# Copy this file to infer.yaml and update paths



# Model# Model

model:model:

  checkpoint: "path/to/best_model"  checkpoint: "path/to/best_model"

  name: "microsoft/wavlm-base-plus"  name: "microsoft/wavlm-base-plus"

  head_hidden_dim: 256  head_hidden_dim: 256



# Audio Processing# Audio Processing

audio:audio:

  sampling_rate: 16000  sampling_rate: 16000

  max_duration: 5  max_duration: 5



# Inference# Inference

inference:inference:

  batch_size: 1  batch_size: 1

  device: "cuda"  device: "cuda"



# Input# Input

input:input:

  audio_path: null  audio_path: null

  audio_dir: null  audio_dir: null



# Output# Output

output:output:

  dir: "output/predictions"  dir: "output/predictions"

  save_results: true  save_results: true

  format: "json"  format: "json"



# Label Mappings# Label Mappings

labels:labels:

  gender:  gender:

    0: "Male"    0: "Male"

    1: "Female"    1: "Female"

  dialect:  dialect:

    0: "North"    0: "North"

    1: "Central"    1: "Central"

    2: "South"    2: "South"