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

Merge pull request #604 from aliencaocao/main-1

Browse source 
Add doc string for n_gpu_layers argument and make -1 offload all layers
This commit is contained in:
Andrei 2023-08-14 22:40:10 -04:00 committed by GitHub
commit b99e758045
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
1 changed files with 2 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

3
llama_cpp/llama.py
View file

@ -239,6 +239,7 @@ class Llama:
n_ctx: Maximum context size.
n_parts: Number of parts to split the model into. If -1, the number of parts is automatically determined.
seed: Random seed. -1 for random.
n_gpu_layers: Number of layers to offload to GPU (-ngl). If -1, all layers are offloaded.
f16_kv: Use half-precision for key/value cache.
logits_all: Return logits for all tokens, not just the last token.
vocab_only: Only load the vocabulary no weights.
@ -267,7 +268,7 @@ class Llama:
self.params = llama_cpp.llama_context_default_params()
self.params.n_ctx = n_ctx
self.params.n_gpu_layers = n_gpu_layers
self.params.n_gpu_layers = 0x7FFFFFFF if n_gpu_layers == -1 else n_gpu_layers # 0x7FFFFFFF is INT32 max, will be auto set to all layers
self.params.seed = seed
self.params.f16_kv = f16_kv
self.params.logits_all = logits_all