lynn

> Then the second part of my reply should be correct (?), i.e. max len should be 2048 or longer to be good on speed. Yes.

> Hi ! Some of these datasets are script-based datasets. Maybe the issue comes from here ? I see some of them have hardcoded HF URLs in their scripts: >...

> I'm trying to convert deepseek v3 using A100 > > ``` > python fp8_cast_bf16.py --input-fp8-hf-path /models/DeepSeek-V3/ --output-bf16-hf-path /models/DeepSeek-V3-BF16 > ``` > > but got assert builder.options.allow_fp8e4nv, "fp8e4nv data type...