Enable prefill for running CausalLM using ET runtime#73

Merged

guangy10 merged 15 commits intohuggingface:mainfrom

guangy10:prefill

Jun 26, 2025

Commits on Jun 25, 2025

Enable prefill for running CausalLM using ET runtime

guangy10
authored and
Guang Yang
committed
export cache_position dynamically
Guang Yang
committed
bug fix in generate loop

guangy10
authored and
Guang Yang
committed
fix for older version of transformers
Guang Yang
committed
move pinned transformers for prefill

guangy10
authored and
Guang Yang
committed
temporarily disble running etLLM generated phi4 in optimum-et due to missing cache_position support

guangy10
authored and
Guang Yang
committed
gated prefill for executorch<=0.6 for backwards compatibility

guangy10
authored and
Guang Yang
committed
more fixes for executorch 0.6.0

guangy10
authored and
Guang Yang
committed
set upper bound of seq_len dynamically based on users config

guangy10
authored and
Guang Yang
committed
fix phi4 tests

guangy10
authored and
Guang Yang
committed
fix seq_len dim for models using hybrid cache

guangy10
authored and
Guang Yang
committed
Update install_dev.py

guangy10
authored and
Guang Yang
committed
gated export with dynamic shapes for models using hf's hybrid cache impl
Guang Yang
committed
fix bc for 0.6
Guang Yang
committed

Commits on Jun 26, 2025

fix generate by extracting seq_len from the method meta
Guang Yang
committed