Liang

Please refer to [#14561](https://github.com/huggingface/transformers/issues/14561) . Also if you would like to set the same value (e.g. `model_max_len`) as you train tokenizer before, you can modify `PreTrainedTokenizerFast` as following ``` PreTrainedTokenizerFast(tokenizer_object=tokenizer,...

@Narsil You're right, `tokenizers` has no special treatment of special tokens. As describled in https://huggingface.co/course/chapter6/8?fw=pt: > To wrap the tokenizer in a PreTrainedTokenizerFast, we can either pass the tokenizer we...

@karankapur999 Not yet

Ref here: https://github.com/microsoft/autogen/issues/217#issuecomment-1783895515

I recently experimented with using monkey patching to implement streaming functionality. This method enables the dynamic and flexible addition of new Agents without the need to alter the original autogen...

Hi @Alvaromah , thank you for your contribution, which has enabled autogen to stream in the terminal. However, I would like to ask if there's a way to support streaming...

I encountered the same issue as @selectorseb while deploying a finetuned Llama-3.2 model using vLLM with Docker. Initially, I faced the same problem mentioned in the original post @djannot, but...

> It is likely because the model is too small, it is not fully utilizing the GPU, which causes the effect of liger kernel to be non significant. > >...

Hi @ChunChenLin , may I ask which model you are using and whether you are utilizing any distributed training?

@ChunChenLin That's quite strange. At the beginning of this thread, we thought it was because my model has fewer parameters, so even with Liger, it couldn't achieve acceleration. But now,...