Closed HaneulCheong closed 11 months ago
. . . RVC was properly using my GPU without an issue until I have pulled the latest Hay Say update yesterday . . .
Also, is there a way to downgrade my local Hay Say to confirm this? Sorry I'm not quite familiar with Docker.
Hello HaneulCheong, Would you please check the docker-compose.yaml file (which is likely located in your home directory, i.e. C:/Users/\<your-user-name>/) and verify that the following lines are not commented out with hashtags? If they are, remove the hashtag at the beginning of each line and then launch Hay Say again:
Yeah the lines were commented again; The pull must have reset my edits on the local yaml
file.
hay-say-ui-rvc_server-1 | 2023-10-16 01:29:46 | INFO | fairseq.tasks.hubert_pretraining | current directory is /root/hay_say/rvc
hay-say-ui-rvc_server-1 | 2023-10-16 01:29:46 | INFO | fairseq.tasks.hubert_pretraining | HubertPretrainingTask Config {'_name': 'hubert_pretraining', 'data': 'metadata', 'fine_tuning': False, 'labels': ['km'], 'label_dir': 'label', 'label_rate': 50.0, 'sample_rate': 16000, 'normalize': False, 'enable_padding': False, 'max_keep_size': None, 'max_sample_size': 250000, 'min_sample_size': 32000, 'single_target': False, 'random_crop': True, 'pad_audio': False}
hay-say-ui-rvc_server-1 | 2023-10-16 01:29:46 | INFO | fairseq.models.hubert.hubert | HubertModel Config: {'_name': 'hubert', 'label_rate': 50.0, 'extractor_mode': default, 'encoder_layers': 12, 'encoder_embed_dim': 768, 'encoder_ffn_embed_dim': 3072, 'encoder_attention_heads': 12, 'activation_fn': gelu, 'layer_type': transformer, 'dropout': 0.1, 'attention_dropout': 0.1, 'activation_dropout': 0.0, 'encoder_layerdrop': 0.05, 'dropout_input': 0.1, 'dropout_features': 0.1, 'final_dim': 256, 'untie_final_proj': True, 'layer_norm_first': False, 'conv_feature_layers': '[(512,10,5)] + [(512,3,2)] * 4 + [(512,2,2)] * 2', 'conv_bias': False, 'logit_temp': 0.1, 'target_glu': False, 'feature_grad_mult': 0.1, 'mask_length': 10, 'mask_prob': 0.8, 'mask_selection': static, 'mask_other': 0.0, 'no_mask_overlap': False, 'mask_min_space': 1, 'mask_channel_length': 10, 'mask_channel_prob': 0.0, 'mask_channel_selection': static, 'mask_channel_other': 0.0, 'no_mask_channel_overlap': False, 'mask_channel_min_space': 1, 'conv_pos': 128, 'conv_pos_groups': 16, 'latent_temp': [2.0, 0.5, 0.999995], 'skip_masked': False, 'skip_nomask': False, 'checkpoint_activations': False, 'required_seq_len_multiple': 2, 'depthwise_conv_kernel_size': 31, 'attn_type': '', 'pos_enc_type': 'abs', 'fp16': False}
hay-say-ui-rvc_server-1 | Found GPU NVIDIA GeForce RTX 3060 Laptop GPU
hay-say-ui-rvc_server-1 | Use Language: en_US
hay-say-ui-rvc_server-1 | loading weights/Twilight Sparkle.pth
hay-say-ui-rvc_server-1 | gin_channels: 256 self.spk_embed_dim: 109
hay-say-ui-rvc_server-1 | <All keys matched successfully>
hay-say-ui-rvc_server-1 | 172.18.0.8 - - [16/Oct/2023 01:31:19] "POST /generate HTTP/1.1" 200 -
hay-say-ui-hay_say_ui-1 | 172.18.0.1 - - [16/Oct/2023 01:31:35] "POST /_dash-update-component HTTP/1.1" 200 -
The issue itself is solved now, but do you think this comment process can be automated in the future? Or I suppose at least there could be a new paragraph added under the updating chapter of Readme.md
reminding people with a CUDA GPU to edit their yaml
file again.
Hello HaneulCheong,
I am glad to hear that the issue is resolved. I went ahead and updated the Readme file to specifically mention un-commenting lines after an update to re-enable GPU integration; thanks for the suggestion.
I do have plans to write a "launcher" application in the QT framework that will dynamically build a docker-compose.yaml file based on what the user selects in a UI, including one or more checkboxes to make Cuda-capable GPUs visible to Hay Say. These user settings will persist across updates to Hay Say, and the launcher will have the added benefit of being started like a normal everyday application, eliminating the need for users to open a command prompt or terminal.
Best, Hydrus
Thanks a lot! This issue is resolved.
Environment
Log
Additional Comments
I installed Hay Say on October 1st if I'm remembering this correctly. I am also pretty sure RVC was properly using my GPU without an issue until I have pulled the latest Hay Say update yesterday, and after the pull this started to happen. Do you have some clues on what's going on here?