Minor fixes

egs/zipvoice/README.md

@@ -105,7 +105,7 @@ pip install -r ../../requirements.txt
 #### 1.1. Prepare the Emilia dataset
 
 ```bash
-bash scripts/prepare_emilia.sh --stage 0 --stop-stage 4
+bash scripts/prepare_emilia.sh
 ```
 
 See [scripts/prepare_emilia.sh](scripts/prepare_emilia.sh) for step by step instructions.
@@ -113,7 +113,7 @@ See [scripts/prepare_emilia.sh](scripts/prepare_emilia.sh) for step by step inst
 #### 1.2 Prepare the LibriTTS dataset
 
 ```bash
-bash scripts/prepare_libritts.sh --stage 0 --stop-stage 3
+bash scripts/prepare_libritts.sh
 ```
 
 See [scripts/prepare_libritts.sh](scripts/prepare_libritts.sh) for step by step instructions.
@@ -139,7 +139,7 @@ python3 zipvoice/train_flow.py \
         --lr-hours 30000 \
         --lr-batches 7500 \
         --token-file "data/tokens_emilia.txt" \
-        --manifest-dir "data/fbank_emilia" \
+        --manifest-dir "data/fbank" \
         --num-epochs 11 \
         --exp-dir zipvoice/exp_zipvoice
 ```
@@ -172,7 +172,7 @@ python3 zipvoice/train_distill.py \
         --base-lr 0.0005 \
         --max-duration 500 \
         --token-file "data/tokens_emilia.txt" \
-        --manifest-dir "data/fbank_emilia" \
+        --manifest-dir "data/fbank" \
         --teacher-model zipvoice/exp_zipvoice/epoch-11-avg-4.pt \
         --num-updates 60000 \
         --distill-stage "first" \
@@ -205,7 +205,7 @@ python3 zipvoice/train_distill.py \
         --base-lr 0.0001 \
         --max-duration 200 \
         --token-file "data/tokens_emilia.txt" \
-        --manifest-dir "data/fbank_emilia" \
+        --manifest-dir "data/fbank" \
         --teacher-model zipvoice/exp_zipvoice_distill_1stage/iter-60000-avg-7.pt \
         --num-updates 2000 \
         --distill-stage "second" \
@@ -233,7 +233,7 @@ python3 zipvoice/train_flow.py \
         --lr-epochs 10 \
         --lr-batches 7500 \
         --token-file "data/tokens_libritts.txt" \
-        --manifest-dir "data/fbank_libritts" \
+        --manifest-dir "data/fbank" \
         --num-epochs 60 \
         --exp-dir zipvoice/exp_zipvoice_libritts
 ```
@@ -266,7 +266,7 @@ python3 zipvoice/train_distill.py \
         --base-lr 0.001 \
         --max-duration 250 \
         --token-file "data/tokens_libritts.txt" \
-        --manifest-dir "data/fbank_libritts" \
+        --manifest-dir "data/fbank" \
         --teacher-model zipvoice/exp_zipvoice_libritts/epoch-60-avg-10.pt \
         --num-epochs 6 \
         --distill-stage "first" \
@@ -299,7 +299,7 @@ python3 zipvoice/train_distill.py \
         --base-lr 0.001 \
         --max-duration 250 \
         --token-file "data/tokens_libritts.txt" \
-        --manifest-dir "data/fbank_libritts" \
+        --manifest-dir "data/fbank" \
         --teacher-model zipvoice/exp_zipvoice_distill_1stage_libritts/epoch-6-avg-3.pt \
         --num-epochs 6 \
         --distill-stage "second" \

egs/zipvoice/scripts/run_eval.sh

@@ -1,75 +0,0 @@
-#!/usr/bin/env bash
-
-export PYTHONPATH=../../../:$PYTHONPATH
-
-stage=1
-stop_stage=10
-generated_wav_path="flow-matching/exp/generated_wavs"
-
-. shared/parse_options.sh || exit 1
-
-
-log() {
-  # This function is from espnet
-  local fname=${BASH_SOURCE[1]##*/}
-  echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
-}
-
-if [ $stage -le -2 ] && [ $stop_stage -ge -2 ]; then
-  log "Stage -2: Install dependencies and download models"
-
-  pip install -r requirements-eval.txt
-  pip install git+https://github.com/sarulab-speech/UTMOSv2.git
-  modelscope download --model k2-fsa/TTS_eval_models --local_dir TTS_eval_models
-fi
-
-
-if [ $stage -le -1 ] && [ $stop_stage -ge -1 ]; then
-  log "Stage -1: Prepare evaluation data."
-
-  mkdir -p data/reference/librispeech-test-clean
-
-  gunzip -c data/fbank/librispeech_cuts_with_prompts_test-clean.jsonl.gz | \
-    jq -r '"\(.recording.sources[0].source)"' | \
-    awk '{split($1, a, "/"); cmd="cp "$1" data/reference/librispeech-test-clean/"a[length(a)]; print cmd; system(cmd)}'
-
-
-  mkdir -p data/reference/librispeech-test-clean-prompt
-  gunzip -c data/fbank/librispeech_cuts_with_prompts_test-clean.jsonl.gz | \
-    jq -r '"\(.custom.prompt.recording.sources[0].source) \(.recording.sources[0].source)"' | \
-    awk '{split($2, a, "/"); cmd="cp "$1" data/reference/librispeech-test-clean-prompt/"a[length(a)]; print cmd; system(cmd)}'
-fi
-
-
-if [ $stage -le 1 ] && [ $stop_stage -ge 1 ]; then
-  log "Stage 1: Evaluate the model with FSD."
-
-  python local/evaluate_fsd.py --real-path data/reference/librispeech-test-clean \
-    --eval-path $generated_wav_path
-fi
-
-if [ $stage -le 2 ] && [ $stop_stage -ge 2 ]; then
-  log "Stage 2: Evaluate the model with SIM."
-
-  python local/evaluate_sim.py --real-path data/reference/librispeech-test-clean \
-    --eval-path $generated_wav_path
-fi
-
-if [ $stage -le 3 ] && [ $stop_stage -ge 3 ]; then
-  log "Stage 3: Evaluate the model with UTMOS."
-
-  python local/evaluate_utmos.py --wav-path $generated_wav_path
-fi
-
-if [ $stage -le 4 ] && [ $stop_stage -ge 4 ]; then
-  log "Stage 4: Evaluate the model with UTMOSv2."
-
-  python local/evaluate_utmosv2.py --wav-path $generated_wav_path
-fi
-
-if [ $stage -le 5 ] && [ $stop_stage -ge 5 ]; then
-  log "Stage 5: Evaluate the model with WER."
-
-  python local/evaluate_wer_hubert.py --wav-path $generated_wav_path \
-    --decode-path $generated_wav_path/decode 
-fi