baalajimaestro/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

Update llama.cpp

Browse source
This commit is contained in:
Andrei Betlen 2023-04-28 15:32:43 -04:00
parent 9339929f56
commit ea0faabae1
2 changed files with 42 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

41
llama_cpp/llama_cpp.py
View file

@ -212,13 +212,16 @@ def llama_get_kv_cache_token_count(ctx: llama_context_p) -> c_int:
_lib.llama_get_kv_cache_token_count.argtypes = [llama_context_p] _lib.llama_get_kv_cache_token_count.argtypes = [llama_context_p]
_lib.llama_get_kv_cache_token_count.restype = c_int _lib.llama_get_kv_cache_token_count.restype = c_int
# Sets the current rng seed. # Sets the current rng seed.
def llama_set_rng_seed(ctx: llama_context_p, seed: c_int): def llama_set_rng_seed(ctx: llama_context_p, seed: c_int):
return _lib.llama_set_rng_seed(ctx, seed) return _lib.llama_set_rng_seed(ctx, seed)
_lib.llama_set_rng_seed.argtypes = [llama_context_p, c_int] _lib.llama_set_rng_seed.argtypes = [llama_context_p, c_int]
_lib.llama_set_rng_seed.restype = None _lib.llama_set_rng_seed.restype = None
# Returns the size in bytes of the state (rng, logits, embedding and kv_cache) # Returns the size in bytes of the state (rng, logits, embedding and kv_cache)
def llama_get_state_size(ctx: llama_context_p) -> c_size_t: def llama_get_state_size(ctx: llama_context_p) -> c_size_t:
return _lib.llama_get_state_size(ctx) return _lib.llama_get_state_size(ctx)
@ -249,6 +252,44 @@ _lib.llama_set_state_data.argtypes = [llama_context_p, POINTER(c_uint8)]
_lib.llama_set_state_data.restype = c_size_t _lib.llama_set_state_data.restype = c_size_t
# Save/load session file
def llama_load_session_file(
ctx: llama_context_p,
path_session: bytes,
tokens_out,
n_token_capacity: c_size_t,
n_token_count_out,
) -> c_size_t:
return _lib.llama_load_session_file(
ctx, path_session, tokens_out, n_token_capacity, n_token_count_out
)
_lib.llama_load_session_file.argtypes = [
llama_context_p,
c_char_p,
llama_token_p,
c_size_t,
POINTER(c_size_t),
]
_lib.llama_load_session_file.restype = c_size_t
def llama_save_session_file(
ctx: llama_context_p, path_session: bytes, tokens, n_token_count: c_size_t
) -> c_size_t:
return _lib.llama_save_session_file(ctx, path_session, tokens, n_token_count)
_lib.llama_save_session_file.argtypes = [
llama_context_p,
c_char_p,
llama_token_p,
c_size_t,
]
_lib.llama_save_session_file.restype = c_size_t
# Run the llama inference to obtain the logits and probabilities for the next token. # Run the llama inference to obtain the logits and probabilities for the next token.
# tokens + n_tokens is the provided batch of new tokens to process # tokens + n_tokens is the provided batch of new tokens to process
# n_past is the number of tokens to use from previous eval calls # n_past is the number of tokens to use from previous eval calls

2
vendor/llama.cpp vendored

@ -1 +1 @@
Subproject commit 0b2da20538d01926b77ea237dd1c930c4d20b686 Subproject commit 7f15c5c477d9933689a9d1c40794483e350c2f19