Yang

CrossEntropyLoss的weights设置后报错

2


When I was inputting long text into a large model, that is, when the len of the text was 1024*1024, a StackOverflow error occurred. ``` thread '' panicked at src/lib.rs:227:33:...

Why hasn't the deepseek-v3 MTP layer of the 61st layer been quantized?

2

Why hasn't the deepseek-v3 MTP layer of the 61st layer been quantized? Can you tell me how to quantify it or do you plan to add quantification of the mtp...