Skip to content

Commit

Permalink
Model: Formatting
Browse files Browse the repository at this point in the history
turboderp committed Aug 23, 2024

Verified

This commit was created on GitHub.com and signed with GitHub’s verified signature.
1 parent a3733ca commit a676c4b
Showing 1 changed file with 7 additions and 3 deletions.
10 changes: 7 additions & 3 deletions backends/exllamav2/model.py
Original file line number Diff line number Diff line change
@@ -585,7 +585,7 @@ def progress(loaded_modules: int, total_modules: int)
cache_class=draft_cache_class,
autosplit=True,
use_tp=False,
model=self.draft_model
model=self.draft_model,
)

for value in self.draft_model.load_autosplit_gen(
@@ -636,7 +636,7 @@ def progress(loaded_modules: int, total_modules: int)
cache_class=cache_class,
autosplit=self.gpu_split_auto,
use_tp=self.use_tp,
model=self.model
model=self.model,
)

# Load model with autosplit (without TP)
@@ -671,7 +671,11 @@ def get_cache_class(self, cache_mode: str):
return ExLlamaV2Cache

def create_cache(
self, cache_class: ExLlamaV2CacheBase, autosplit: bool, use_tp: bool, model: ExLlamaV2
self,
cache_class: ExLlamaV2CacheBase,
autosplit: bool,
use_tp: bool,
model: ExLlamaV2,
):
"""Utility function to create a model cache."""

0 comments on commit a676c4b

Please sign in to comment.