jhansss commited on
Commit
9d28841
·
1 Parent(s): 09fa5bf

Add new configuration file for test and update model options in interface configuration

Browse files
config/cli/yaoyin_test.yaml ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ asr_model: openai/whisper-small
2
+ llm_model: google/gemma-2-2b
3
+ svs_model: espnet/aceopencpop_svs_visinger2_40singer_pretrain
4
+ melody_source: sample-lyric-kising
5
+ language: mandarin
6
+ prompt_template_character: Yaoyin
7
+ speaker: 9
8
+ cache_dir: .cache
9
+
10
+ track_latency: True
config/interface/options.yaml CHANGED
@@ -5,6 +5,8 @@ asr_models:
5
  name: Whisper large-v3
6
  - id: openai/whisper-medium
7
  name: Whisper medium
 
 
8
  - id: sanchit-gandhi/whisper-small-dv
9
  name: Whisper small-dv
10
  - id: facebook/wav2vec2-base-960h
@@ -15,6 +17,8 @@ llm_models:
15
  name: Gemma 2 2B
16
  - id: MiniMaxAI/MiniMax-M1-80k
17
  name: MiniMax M1 80k
 
 
18
 
19
  svs_models:
20
  - id: mandarin-espnet/mixdata_svs_visinger2_spkemb_lang_pretrained
 
5
  name: Whisper large-v3
6
  - id: openai/whisper-medium
7
  name: Whisper medium
8
+ - id: openai/whisper-small
9
+ name: Whisper small
10
  - id: sanchit-gandhi/whisper-small-dv
11
  name: Whisper small-dv
12
  - id: facebook/wav2vec2-base-960h
 
17
  name: Gemma 2 2B
18
  - id: MiniMaxAI/MiniMax-M1-80k
19
  name: MiniMax M1 80k
20
+ - id: meta-llama/Llama-3.2-3B-Instruct
21
+ name: Llama 3.2 3B Instruct
22
 
23
  svs_models:
24
  - id: mandarin-espnet/mixdata_svs_visinger2_spkemb_lang_pretrained