Skip to content

Commit

Permalink
add comments
Browse files Browse the repository at this point in the history
Signed-off-by: jiqing-feng <jiqing.feng@intel.com>
  • Loading branch information
jiqing-feng committed Nov 27, 2024
1 parent 1975027 commit 5b71206
Showing 1 changed file with 2 additions and 1 deletion.
3 changes: 2 additions & 1 deletion optimum/intel/ipex/modeling_base.py
Original file line number Diff line number Diff line change
Expand Up @@ -59,7 +59,7 @@

_IPEX_SUPPORT_MODEL_TYPES = ("llama", "bert", "vit", "falcon", "gpt2")
_IPEX_EXPORTED_GENERATION_METHODS = ("sample", "greedy_search", "beam_sample", "beam_search", "assisted_generation")
# TODO: Try to fix these models compile issue in torch 2.6
# TODO: Fix these models compile issue in torch 2.6
_COMPILE_NOT_READY_MODEL_TYPES = ("electra", "roformer", "beit")


Expand Down Expand Up @@ -112,6 +112,7 @@ def __init__(
):
from torch._inductor import config

# System level optimization
torch._inductor.config.cpp_wrapper = True
os.environ["TORCHINDUCTOR_FREEZING"] = "1"
logger.info("Enable torch.compile optimization, start warm up")
Expand Down

0 comments on commit 5b71206

Please sign in to comment.