Spaces:
Running
Running
Questions?
#84
pinned
by
nouamanetazi
HF staff
- opened
If you have any questions about the content of the blog, feel free to ask here!
nouamanetazi
pinned discussion
Hi, I am just a nobbie, trying to learn about the training LLM models.
It might be dumb from me to ask this.
But can anyone tell me how the batch size effects the throughput (tokens generated /sec) also how does having a larger batch size will tend to make less use of each training token rendering convergence slower and potentially wasting compute.
Does the value in this graph really make sense?[40, 180, 320, 460]^(T) @ [20, 40]...