Re-enable cache
This commit is contained in:
parent
6639371407
commit
21acd7901f
1 changed files with 3 additions and 5 deletions
|
@ -831,9 +831,7 @@ class Llama:
|
|||
"logprobs is not supported for models created with logits_all=False"
|
||||
)
|
||||
|
||||
# Temporarily disable usage of the cache
|
||||
# See: https://github.com/abetlen/llama-cpp-python/issues/348#issuecomment-1583072408
|
||||
if self.cache and False:
|
||||
if self.cache:
|
||||
try:
|
||||
cache_item = self.cache[prompt_tokens]
|
||||
cache_prefix_len = Llama.longest_token_prefix(
|
||||
|
@ -1071,14 +1069,14 @@ class Llama:
|
|||
}
|
||||
],
|
||||
}
|
||||
if self.cache and False:
|
||||
if self.cache:
|
||||
if self.verbose:
|
||||
print("Llama._create_completion: cache save", file=sys.stderr)
|
||||
self.cache[prompt_tokens + completion_tokens] = self.save_state()
|
||||
print("Llama._create_completion: cache saved", file=sys.stderr)
|
||||
return
|
||||
|
||||
if self.cache and False:
|
||||
if self.cache:
|
||||
if self.verbose:
|
||||
print("Llama._create_completion: cache save", file=sys.stderr)
|
||||
self.cache[prompt_tokens + completion_tokens] = self.save_state()
|
||||
|
|
Loading…
Reference in a new issue