fix: use relative checkpoint paths (gpt.pth, s2mel.pth) in config.yaml
Browse files- config.yaml +2 -31
config.yaml
CHANGED
|
@@ -10,7 +10,6 @@ dataset:
|
|
| 10 |
n_mels: 100
|
| 11 |
mel_fmin: 0
|
| 12 |
normalize: false
|
| 13 |
-
|
| 14 |
gpt:
|
| 15 |
model_dim: 1280
|
| 16 |
max_mel_tokens: 1815
|
|
@@ -41,7 +40,6 @@ gpt:
|
|
| 41 |
num_blocks: 4
|
| 42 |
input_layer: "conv2d2"
|
| 43 |
perceiver_mult: 2
|
| 44 |
-
|
| 45 |
semantic_codec:
|
| 46 |
codebook_size: 8192
|
| 47 |
hidden_size: 1024
|
|
@@ -49,7 +47,6 @@ semantic_codec:
|
|
| 49 |
vocos_dim: 384
|
| 50 |
vocos_intermediate_dim: 2048
|
| 51 |
vocos_num_layers: 12
|
| 52 |
-
|
| 53 |
s2mel:
|
| 54 |
preprocess_params:
|
| 55 |
sr: 22050
|
|
@@ -60,7 +57,6 @@ s2mel:
|
|
| 60 |
n_mels: 80
|
| 61 |
fmin: 0
|
| 62 |
fmax: "None"
|
| 63 |
-
|
| 64 |
dit_type: "DiT"
|
| 65 |
reg_loss_type: "l1"
|
| 66 |
style_encoder:
|
|
@@ -106,37 +102,12 @@ s2mel:
|
|
| 106 |
dilation_rate: 1
|
| 107 |
p_dropout: 0.2
|
| 108 |
style_condition: true
|
| 109 |
-
|
| 110 |
-
# gpt_checkpoint: /cubefs/users/liyunpei/fsdp/model_dir/6112986_gpt2_step1_diff_wav/checkpoint_153000.pth
|
| 111 |
-
# gpt_checkpoint: /cubefs/users/liyunpei/fsdp/model_dir/6120860_xjc_emo_with_fixed_emo/checkpoint_12000.yaml
|
| 112 |
-
# gpt_checkpoint: /cubefs/users/liyunpei/fsdp/model_dir/6120905_gpt2_step1_diff_wav/checkpoint_171000.pth
|
| 113 |
-
# gpt_checkpoint: /cubefs/users/liyunpei/fsdp/model_dir/6127529_xjc_emo_with_fixed_emo/checkpoint_2000.pth
|
| 114 |
-
# gpt_checkpoint: /cubefs/users/liyunpei/fsdp/model_dir/6127529_xjc_emo_with_fixed_emo/checkpoint_13000.pth
|
| 115 |
-
# gpt_checkpoint: /cubefs/users/liyunpei/fsdp/model_dir/6130259_gpt2_step1_diff_wav/checkpoint_176000.pth
|
| 116 |
-
# gpt_checkpoint: /cubefs/users/liyunpei/fsdp/model_dir/6130297_gpt_step4/checkpoint_7000.pth
|
| 117 |
-
# gpt_checkpoint: /cubefs/users/liyunpei/fsdp/model_dir/6132994_gpt2_step1_diff_wav/checkpoint_180000.pth
|
| 118 |
-
# gpt_checkpoint: /cubefs/users/liyunpei/fsdp/model_dir/6127529_xjc_emo_with_fixed_emo/checkpoint_18000.pth
|
| 119 |
-
# gpt_checkpoint: /cubefs/users/liyunpei/fsdp/model_dir/6136618_xjc_emo_with_fixed_emo/checkpoint_9000.pth
|
| 120 |
-
# gpt_checkpoint: /cubefs/users/liyunpei/fsdp/model_dir/6135639_gpt2_step1_diff_wav/checkpoint_195000.pth
|
| 121 |
-
# gpt_checkpoint: /cubefs/users/liyunpei/fsdp/model_dir/6140984_xjc_emo_with_fixed_emo/checkpoint_6000.pth good?
|
| 122 |
-
# gpt_checkpoint: /cubefs/users/liyunpei/fsdp/model_dir/6142099_gpt2_step1_diff_wav/checkpoint_198000.pth
|
| 123 |
-
# gpt_checkpoint: /cubefs/users/liyunpei/fsdp/model_dir/6140984_xjc_emo_with_fixed_emo/checkpoint_8000.pth
|
| 124 |
-
# gpt_checkpoint: /cubefs/users/liyunpei/fsdp/model_dir/6140984_xjc_emo_with_fixed_emo/checkpoint_5000.pth good
|
| 125 |
-
# gpt_checkpoint: /cubefs/users/liyunpei/fsdp/model_dir/6140984_xjc_emo_with_fixed_emo/checkpoint_9000.pth haixing
|
| 126 |
-
# gpt_checkpoint: /cubefs/users/liyunpei/fsdp/model_dir/6135457_gpt_step4/checkpoint_10000.pth
|
| 127 |
-
# gpt_checkpoint: /cubefs/users/liyunpei/fsdp/model_dir/6142099_gpt2_step1_diff_wav/checkpoint_206000.pth
|
| 128 |
-
# gpt_checkpoint: /cubefs/users/xujiacheng/Code/checkpoint-emo/6142099_checkpoint_206000_with_emo_tokenizer.pth
|
| 129 |
-
# gpt_checkpoint: /cubefs/users/liyunpei/fsdp/model_dir/6146344_gpt_step4/checkpoint_1000.pth
|
| 130 |
-
gpt_checkpoint: /cubefs/users/liyunpei/fsdp/model_dir/6140984_xjc_emo_with_fixed_emo/checkpoint_5000.pth
|
| 131 |
-
# gpt_checkpoint: /cubefs/users/liyunpei/fsdp/model_dir/6087548_xjc_emo_with_fixed_emo/checkpoint_12000.pth
|
| 132 |
w2v_stat: wav2vec2bert_stats.pt
|
| 133 |
-
s2mel_checkpoint:
|
| 134 |
emo_matrix: feat2.pt
|
| 135 |
spk_matrix: feat1.pt
|
| 136 |
emo_num: [3, 17, 2, 8, 4, 5, 10, 24]
|
| 137 |
-
# emo_matrix: emo_index2.5.pt
|
| 138 |
-
# spk_matrix: spk_index2.5.pt
|
| 139 |
-
# emo_num: [7, 13, 6, 8, 6, 8, 11, 8]
|
| 140 |
qwen_emo_path: qwen0.6bemo4-merge/
|
| 141 |
vocoder:
|
| 142 |
type: "bigvgan"
|
|
|
|
| 10 |
n_mels: 100
|
| 11 |
mel_fmin: 0
|
| 12 |
normalize: false
|
|
|
|
| 13 |
gpt:
|
| 14 |
model_dim: 1280
|
| 15 |
max_mel_tokens: 1815
|
|
|
|
| 40 |
num_blocks: 4
|
| 41 |
input_layer: "conv2d2"
|
| 42 |
perceiver_mult: 2
|
|
|
|
| 43 |
semantic_codec:
|
| 44 |
codebook_size: 8192
|
| 45 |
hidden_size: 1024
|
|
|
|
| 47 |
vocos_dim: 384
|
| 48 |
vocos_intermediate_dim: 2048
|
| 49 |
vocos_num_layers: 12
|
|
|
|
| 50 |
s2mel:
|
| 51 |
preprocess_params:
|
| 52 |
sr: 22050
|
|
|
|
| 57 |
n_mels: 80
|
| 58 |
fmin: 0
|
| 59 |
fmax: "None"
|
|
|
|
| 60 |
dit_type: "DiT"
|
| 61 |
reg_loss_type: "l1"
|
| 62 |
style_encoder:
|
|
|
|
| 102 |
dilation_rate: 1
|
| 103 |
p_dropout: 0.2
|
| 104 |
style_condition: true
|
| 105 |
+
gpt_checkpoint: gpt.pth
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 106 |
w2v_stat: wav2vec2bert_stats.pt
|
| 107 |
+
s2mel_checkpoint: s2mel.pth
|
| 108 |
emo_matrix: feat2.pt
|
| 109 |
spk_matrix: feat1.pt
|
| 110 |
emo_num: [3, 17, 2, 8, 4, 5, 10, 24]
|
|
|
|
|
|
|
|
|
|
| 111 |
qwen_emo_path: qwen0.6bemo4-merge/
|
| 112 |
vocoder:
|
| 113 |
type: "bigvgan"
|