Upload folder using huggingface_hub

config.yaml

@@ -0,0 +1,74 @@
+acoustic_encoder_configs:
+  avg_prob: 0.2
+  emb_size: 256
+  ft_size: 512
+  is_24k: true
+  mix_prob: 0.0
+  output_ln: true
+  speech_upstream: utter-project/mHuBERT-147
+  transformer_configs:
+    attn_dropout: 0.05
+    depth: 4
+    dim: 512
+    dim_head: 64
+    ff_dropout: 0.05
+    heads: 8
+    input_dim: 800
+    num_register_tokens: 0
+    skip_posemb: false
+acoustic_quantizer_configs:
+  codebook_dim: 64
+  codebook_size: 4096
+  input_dim: 256
+  n_codebooks: 24
+  possibly_no_quantizer: false
+  quantizer_dropout: 1
+additive: true
+apply_ln: false
+avg_spk_prob: 0.0
+do_acoustic_vae: false
+do_auggan: false
+do_club: true
+drop_spk_prob: 0.0
+generator_configs:
+  cond_dim: 128
+  dim: 3072
+  hop_length: 480
+  input_channels: 288
+  intermediate_dim: 3072
+  n_fft: 1920
+  num_layers: 6
+  padding: same
+  use_film: false
+input_layer_configs:
+  dropout: 0.05
+  hidden_dims:
+  - 1024
+  - 1024
+  - 1024
+  input_dim: 768
+  output_dim: 256
+  output_ln: true
+  use_linear: false
+on_the_fly: false
+progress_encoding_configs:
+  emb_dim: 32
+  emb_n: 11
+quantizer_configs:
+  dim: 256
+  levels:
+  - 8
+  - 8
+  - 8
+  - 8
+  - 8
+  num_codebooks: 2
+  possibly_no_quantizer: false
+  quantizer_dropout: 0.5
+recon_content: true
+sample_rate: 24000
+segauc_ratio: 0.0
+use_perceptual_loss: false
+use_rvq_acoustic: true
+use_rvq_content: false
+use_segauc: false

model.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a25a8cb92096fcc3f1272e920094dbe47d779d05d515fc8199af2a1562befc2
+size 1224641535

segmenter.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c6d8106bba8fd556eb028d064bc8bbf5394d493dd6753d2f957b7ef7825a5b4
+size 405871563

segmenter_config.yaml

@@ -0,0 +1,29 @@
+ckpt_path: segmenter.ckpt
+normalize_target: true
+notch_layer: 8
+notch_model_configs:
+  attn_dropout: 0.05
+  depth: 4
+  dim: 768
+  dim_head: 64
+  ff_dropout: 0.05
+  heads: 12
+  num_register_tokens: 0
+  skip_posemb: false
+notch_threshold_range:
+- 0.2
+- 0.2
+prominence_range:
+- 0.1
+- 0.1
+speech_upstream: utter-project/mHuBERT-147
+target_pooling_layers:
+- 8
+upstream_configs:
+  activation_dropout: 0.0
+  attention_dropout: 0.0
+  feat_proj_dropout: 0.0
+  hidden_dropout: 0.0
+  hidden_dropout_prob: 0.0
+  layerdrop: 0.0
+  mask_time_prob: 0.0