| ▲ | hrimfaxi 8 hours ago | |
Why does this G character appear to prefix most of the output? ("Ġlike") | ||
| ▲ | frwickst 8 hours ago | parent | next [-] | |
It is a tokenizer artifact most likely (https://github.com/huggingface/transformers/issues/4786). So the output is not properly decoded in this case, it should just be a space. | ||
| ▲ | kgeist 8 hours ago | parent | prev [-] | |
The original tokens have Ġ instead of space. I had this issue too when writing an inference engine for Qwen. You have to "normalize" those special characters. | ||