🍒 Upload config, processor and pretrained for FastSpeech2 Korean.

config.yml

@@ -0,0 +1,79 @@
+ This is the hyperparameter configuration file for FastSpeech2 v1.
+# Please make sure this is adjusted for the KSS dataset. If you want to
+# apply to the other dataset, you might need to carefully change some parameters.
+# This configuration performs 200k iters but a best checkpoint is around 150k iters.
+
+###########################################################
+#                FEATURE EXTRACTION SETTING               #
+###########################################################
+hop_size: 256            # Hop size.
+format: "npy"
+
+
+###########################################################
+#              NETWORK ARCHITECTURE SETTING               #
+###########################################################
+model_type: "fastspeech2"
+
+fastspeech2_params:
+    dataset: "kss"
+    n_speakers: 1
+    encoder_hidden_size: 384
+    encoder_num_hidden_layers: 4
+    encoder_num_attention_heads: 2
+    encoder_attention_head_size: 192  # hidden_size // num_attention_heads
+    encoder_intermediate_size: 1024
+    encoder_intermediate_kernel_size: 3
+    encoder_hidden_act: "mish"
+    decoder_hidden_size: 384
+    decoder_num_hidden_layers: 4
+    decoder_num_attention_heads: 2
+    decoder_attention_head_size: 192  # hidden_size // num_attention_heads
+    decoder_intermediate_size: 1024
+    decoder_intermediate_kernel_size: 3
+    decoder_hidden_act: "mish"
+    variant_prediction_num_conv_layers: 2
+    variant_predictor_filter: 256
+    variant_predictor_kernel_size: 3
+    variant_predictor_dropout_rate: 0.5
+    num_mels: 80
+    hidden_dropout_prob: 0.2
+    attention_probs_dropout_prob: 0.1
+    max_position_embeddings: 2048
+    initializer_range: 0.02
+    output_attentions: False
+    output_hidden_states: False
+
+###########################################################
+#                  DATA LOADER SETTING                    #
+###########################################################
+batch_size: 16              # Batch size for each GPU with assuming that gradient_accumulation_steps == 1.
+remove_short_samples: true  # Whether to remove samples the length of which are less than batch_max_steps.
+allow_cache: true           # Whether to allow cache in dataset. If true, it requires cpu memory.
+mel_length_threshold: 32    # remove all targets has mel_length <= 32 
+is_shuffle: true            # shuffle dataset after each epoch.
+###########################################################
+#             OPTIMIZER & SCHEDULER SETTING               #
+###########################################################
+optimizer_params:
+    initial_learning_rate: 0.001
+    end_learning_rate: 0.00005
+    decay_steps: 150000          # < train_max_steps is recommend.
+    warmup_proportion: 0.02
+    weight_decay: 0.001
+
+gradient_accumulation_steps: 1
+var_train_expr: null  # trainable variable expr (eg. 'embeddings|encoder|decoder' )
+                      # must separate by |. if var_train_expr is null then we 
+                      # training all variable 
+###########################################################
+#                    INTERVAL SETTING                     #
+###########################################################
+train_max_steps: 200000               # Number of training steps.
+save_interval_steps: 5000             # Interval steps to save checkpoint.
+eval_interval_steps: 500              # Interval steps to evaluate the network.
+log_interval_steps: 200               # Interval steps to record the training log.
+###########################################################
+#                     OTHER SETTING                       #
+###########################################################
+num_save_intermediate_results: 1  # Number of batch to be saved as intermediate results.

model.h5

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1e64a252d3f151086f3df5ef27a9d7fd21d15bd1f2ab4ad57b83f70dc4fd10b
+size 124691376

processor.json

@@ -0,0 +1 @@
+{"symbol_to_id": {"pad": 0, "-": 7, "!": 2, "'": 3, "(": 4, ")": 5, ",": 6, ".": 8, ":": 9, ";": 10, "?": 11, " ": 12, "\u1100": 13, "\u1101": 14, "\u1102": 15, "\u1103": 16, "\u1104": 17, "\u1105": 18, "\u1106": 19, "\u1107": 20, "\u1108": 21, "\u1109": 22, "\u110a": 23, "\u110b": 24, "\u110c": 25, "\u110d": 26, "\u110e": 27, "\u110f": 28, "\u1110": 29, "\u1111": 30, "\u1112": 31, "\u1161": 32, "\u1162": 33, "\u1163": 34, "\u1164": 35, "\u1165": 36, "\u1166": 37, "\u1167": 38, "\u1168": 39, "\u1169": 40, "\u116a": 41, "\u116b": 42, "\u116c": 43, "\u116d": 44, "\u116e": 45, "\u116f": 46, "\u1170": 47, "\u1171": 48, "\u1172": 49, "\u1173": 50, "\u1174": 51, "\u1175": 52, "\u11a8": 53, "\u11a9": 54, "\u11aa": 55, "\u11ab": 56, "\u11ac": 57, "\u11ad": 58, "\u11ae": 59, "\u11af": 60, "\u11b0": 61, "\u11b1": 62, "\u11b2": 63, "\u11b3": 64, "\u11b4": 65, "\u11b5": 66, "\u11b6": 67, "\u11b7": 68, "\u11b8": 69, "\u11b9": 70, "\u11ba": 71, "\u11bb": 72, "\u11bc": 73, "\u11bd": 74, "\u11be": 75, "\u11bf": 76, "\u11c0": 77, "\u11c1": 78, "\u11c2": 79, "eos": 80}, "id_to_symbol": {"0": "pad", "1": "-", "2": "!", "3": "'", "4": "(", "5": ")", "6": ",", "7": "-", "8": ".", "9": ":", "10": ";", "11": "?", "12": " ", "13": "\u1100", "14": "\u1101", "15": "\u1102", "16": "\u1103", "17": "\u1104", "18": "\u1105", "19": "\u1106", "20": "\u1107", "21": "\u1108", "22": "\u1109", "23": "\u110a", "24": "\u110b", "25": "\u110c", "26": "\u110d", "27": "\u110e", "28": "\u110f", "29": "\u1110", "30": "\u1111", "31": "\u1112", "32": "\u1161", "33": "\u1162", "34": "\u1163", "35": "\u1164", "36": "\u1165", "37": "\u1166", "38": "\u1167", "39": "\u1168", "40": "\u1169", "41": "\u116a", "42": "\u116b", "43": "\u116c", "44": "\u116d", "45": "\u116e", "46": "\u116f", "47": "\u1170", "48": "\u1171", "49": "\u1172", "50": "\u1173", "51": "\u1174", "52": "\u1175", "53": "\u11a8", "54": "\u11a9", "55": "\u11aa", "56": "\u11ab", "57": "\u11ac", "58": "\u11ad", "59": "\u11ae", "60": "\u11af", "61": "\u11b0", "62": "\u11b1", "63": "\u11b2", "64": "\u11b3", "65": "\u11b4", "66": "\u11b5", "67": "\u11b6", "68": "\u11b7", "69": "\u11b8", "70": "\u11b9", "71": "\u11ba", "72": "\u11bb", "73": "\u11bc", "74": "\u11bd", "75": "\u11be", "76": "\u11bf", "77": "\u11c0", "78": "\u11c1", "79": "\u11c2", "80": "eos"}, "speakers_map": {"kss": 0}, "processor_name": "KSSProcessor"}