Mirror/LocalAI
1
0
Fork 0
mirror of https://github.com/mudler/LocalAI.git synced 2024-06-07 19:40:48 +00:00
Code Issues Packages Projects Releases Wiki Activity

feat(llama.cpp): Bump llama.cpp, adapt grpc server (#1211)

Browse Source 
* feat(llama.cpp): Bump llama.cpp, adapt grpc server

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* ci: fixups

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

---------

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
This commit is contained in:
Ettore Di Giacinto 2023-10-25 20:56:25 +02:00 committed by GitHub
parent c132dbadce
commit f227e918f9
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
3 changed files with 13 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
.github/workflows/test-gpu.yml vendored
View File

@ -40,6 +40,8 @@ jobs:
if [ ! -e /run/systemd/system ]; then
sudo mkdir /run/systemd/system
fi
sudo mkdir -p /host/tests/${{ github.head_ref || github.ref }}
sudo chmod -R 777 /host/tests/${{ github.head_ref || github.ref }}
make \
TEST_DIR="/host/tests/${{ github.head_ref || github.ref }}" \
BUILD_TYPE=cublas \
@ -57,4 +59,5 @@ jobs:
make \
TEST_DIR="/host/tests/${{ github.head_ref || github.ref }}" \
teardown-e2e || true
docker system prune -f -a --volumes || true
sudo rm -rf /host/tests/${{ github.head_ref || github.ref }}
docker system prune -f -a --volumes || true

2
Makefile
View File

@ -8,7 +8,7 @@ GOLLAMA_VERSION?=aeba71ee842819da681ea537e78846dc75949ac0
GOLLAMA_STABLE_VERSION?=50cee7712066d9e38306eccadcfbb44ea87df4b7
CPPLLAMA_VERSION?=96981f37b1e3f450d9e63e571514217bf60f0a7f
CPPLLAMA_VERSION?=9d02956443e5c1ded29b7b5ed8a21bc01ba6f563
# gpt4all version
GPT4ALL_REPO?=https://github.com/nomic-ai/gpt4all

16
backend/cpp/llama/grpc-server.cpp
View File

@ -275,11 +275,11 @@ struct llama_server_context
if (suff_rm_leading_spc && suffix_tokens[0] == space_token) {
suffix_tokens.erase(suffix_tokens.begin());
}
prefix_tokens.insert(prefix_tokens.begin(), llama_token_prefix(ctx));
prefix_tokens.insert(prefix_tokens.begin(), llama_token_bos(ctx)); // always add BOS
prefix_tokens.insert(prefix_tokens.end(), llama_token_suffix(ctx));
prefix_tokens.insert(prefix_tokens.begin(), llama_token_prefix(model));
prefix_tokens.insert(prefix_tokens.begin(), llama_token_bos(model)); // always add BOS
prefix_tokens.insert(prefix_tokens.end(), llama_token_suffix(model));
prefix_tokens.insert(prefix_tokens.end(), suffix_tokens.begin(), suffix_tokens.end());
prefix_tokens.push_back(llama_token_middle(ctx));
prefix_tokens.push_back(llama_token_middle(model));
auto prompt_tokens = prefix_tokens;
@ -419,7 +419,7 @@ struct llama_server_context
if (params.n_predict == 0)
{
has_next_token = false;
result.tok = llama_token_eos(ctx);
result.tok = llama_token_eos(model);
return result;
}
@ -453,7 +453,7 @@ struct llama_server_context
// decrement remaining sampling budget
--n_remain;
if (!embd.empty() && embd.back() == llama_token_eos(ctx))
if (!embd.empty() && embd.back() == llama_token_eos(model))
{
// stopping_word = llama_token_to_piece(ctx, embd.back());
has_next_token = false;
@ -594,7 +594,7 @@ static void parse_options_completion(bool streaming,const backend::PredictOption
if (predict->ignoreeos())
{
llama.params.sparams.logit_bias[llama_token_eos(llama.ctx)] = -INFINITY;
llama.params.sparams.logit_bias[llama_token_eos(llama.model)] = -INFINITY;
}
// const auto &logit_bias = body.find("logit_bias");
@ -676,7 +676,7 @@ static void params_parse(const backend::ModelOptions* request,
}
static bool is_at_eob(llama_server_context &server_context, const llama_token *tokens, const size_t n_tokens) {
return n_tokens && tokens[n_tokens-1] == llama_token_eos(server_context.ctx);
return n_tokens && tokens[n_tokens-1] == llama_token_eos(server_context.model);
}
// Function matching type llama_beam_search_callback_fn_t.