An error occurred in merging lora weights with initial weights

[Alex-DeepL]

Self Checks

• This template is only for bug reports. For questions, please visit Discussions.
• I have thoroughly reviewed the project documentation (installation, training, inference) but couldn't find information to solve my problem. English 中文 日本語 Portuguese (Brazil)
• I have searched for existing issues, including closed ones. Search issues
• I confirm that I am using English to submit this report (我已阅读并同意 Language Policy).
• [FOR CHINESE USERS] 请务必使用英文提交 Issue，否则会被关闭。谢谢！:）
• Please do not modify this template and fill in all required fields.

Cloud or Self Hosted

Cloud

Environment Details

python=3.10

Steps to Reproduce

✔️ Expected Behavior

No response

❌ Actual Behavior

No response

[Whale-Dolphin]

Can you provide your merge command? I think this may because of the error of your lora model.

[Alex-DeepL]

python tools/llama/merge_lora.py --lora-config r_8_alpha_16 --base-weight /home/root123/workspace/lwl/fish-speech/tools/checkpoints/fish-speech-1.5/ --lora-weight /home/root123/workspace/lwl/fish-speech/results/checkpoints/step_000000600.ckpt --output /home/root123/workspace/lwl/fish-speech/tools/checkpoints/result/fish-speech-1.5-yth-lora/

[Alex-DeepL]

this is my command

[Alex-DeepL]

I just used the latest code to fine-tune it, and no errors occurred, but an error like "Loaded weights with error: _IncompatibleKeys(missing_keys="'embeddings. ora_A','embeddings. ora_B','codebook_embeddings. ora_A','codebook_embeddings. ora_B'," occurred when loading the model at the beginning of training, but the model was trained normally.
When I successfully merged Lora weights with the original model to reason, an error occurred if I used the--compile parameter when generating semantic tokens from text in the second part ImportError: /home/root123/.triton/cache/41ce1f58e0a8aa9865e66b90d58b3307bb64c5a006830e49543444 faf56202 fc/cuda_utils.so: undefined symbol: cuModuleGetFunction Set TORCH_LOGS="+dynamo" and TORCHDYNAMO_VERBOSE=1 for more information

[github-actions]

This issue is stale because it has been open for 30 days with no activity.