Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

feat: sync llama.cpp #13

Merged
merged 6 commits into from
Aug 29, 2023
Merged

feat: sync llama.cpp #13

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
60df14c
feat: sync llama.cpp
jhen0409 Aug 29, 2023
7e2df4c
chore: regen patch
jhen0409 Aug 29, 2023
9d93c56
Merge branch 'main' into sync-llama-cpp
jhen0409 Aug 29, 2023
10c167a
fix: copy build-info.h
jhen0409 Aug 29, 2023
e6f4375
fix: llama_token_to_str -> llama_token_to_piece
jhen0409 Aug 29, 2023
8b19b2e
fix: build-info.h gen
jhen0409 Aug 29, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion android/src/main/jni.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -320,7 +320,7 @@ Java_com_rnllama_LlamaContext_doCompletion(
if (token_with_probs.tok == -1 || llama->multibyte_pending > 0) {
continue;
}
const std::string token_text = llama_token_to_str(llama->ctx, token_with_probs.tok);
const std::string token_text = llama_token_to_piece(llama->ctx, token_with_probs.tok);

size_t pos = std::min(sent_count, llama->generated_text.size());

Expand Down
10 changes: 5 additions & 5 deletions cpp/rn-llama.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -101,7 +101,7 @@ static size_t find_partial_stop_string(const std::string &stop,
// format incomplete utf-8 multibyte character for output
static std::string tokens_to_output_formatted_string(const llama_context *ctx, const llama_token token)
{
std::string out = token == -1 ? "" : llama_token_to_str(ctx, token);
std::string out = token == -1 ? "" : llama_token_to_piece(ctx, token);
// if the size is 1 and first bit is 1, meaning it's a partial character
// (size > 1 meaning it's already a known token)
if (out.size() == 1 && (out[0] & 0x80) == 0x80)
Expand All @@ -120,7 +120,7 @@ static std::string tokens_to_str(llama_context *ctx, Iter begin, Iter end)
std::string ret;
for (; begin != end; ++begin)
{
ret += llama_token_to_str(ctx, *begin);
ret += llama_token_to_piece(ctx, *begin);
}
return ret;
}
Expand Down Expand Up @@ -456,7 +456,7 @@ struct llama_rn_context

if (!embd.empty() && embd.back() == llama_token_eos(ctx))
{
// stopping_word = llama_token_to_str(ctx, embd.back());
// stopping_word = llama_token_to_piece(ctx, embd.back());
has_next_token = false;
stopped_eos = true;
LOG_VERBOSE("eos token found", "");
Expand Down Expand Up @@ -503,7 +503,7 @@ struct llama_rn_context
{
const completion_token_output token_with_probs = nextToken();

const std::string token_text = token_with_probs.tok == -1 ? "" : llama_token_to_str(ctx, token_with_probs.tok);
const std::string token_text = token_with_probs.tok == -1 ? "" : llama_token_to_piece(ctx, token_with_probs.tok);
generated_text += token_text;

if (params.n_probs > 0)
Expand Down Expand Up @@ -551,7 +551,7 @@ struct llama_rn_context
}

LOG_VERBOSE("next token, token: %s, token_text: %s, has_next_token: %d, n_remain: %d, num_tokens_predicted: %d, stopped_eos: %d, stopped_word: %d, stopped_limit: %d, stopping_word: %s",
llama_token_to_str(ctx, token_with_probs.tok),
llama_token_to_piece(ctx, token_with_probs.tok),
tokens_to_output_formatted_string(ctx, token_with_probs.tok).c_str(),
has_next_token,
n_remain,
Expand Down
2 changes: 1 addition & 1 deletion ios/RNLlamaContext.mm
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -207,7 +207,7 @@ - (NSDictionary *)completion:(NSDictionary *)params
if (token_with_probs.tok == -1 || llama->multibyte_pending > 0) {
continue;
}
const std::string token_text = llama_token_to_str(llama->ctx, token_with_probs.tok);
const std::string token_text = llama_token_to_piece(llama->ctx, token_with_probs.tok);

size_t pos = std::min(sent_count, llama->generated_text.size());

Expand Down
2 changes: 1 addition & 1 deletion llama.cpp
Open in desktop
Submodule llama.cpp updated 49 files
+35 −9 .devops/llama-cpp-clblast.srpm.spec
+25 −1 .devops/llama-cpp-cublas.srpm.spec
+36 −9 .devops/llama-cpp.srpm.spec
+4 −1 .gitignore
+10 −4 Makefile
+4 −0 README.md
+118 −22 ci/run.sh
+351 −12 common/common.cpp
+39 −1 common/common.h
+3 −3 examples/beam_search/beam_search.cpp
+2 −6 examples/convert-llama2c-to-ggml/README.md
+218 −123 examples/convert-llama2c-to-ggml/convert-llama2c-to-ggml.cpp
+1 −1 examples/embd-input/embd-input-lib.cpp
+1 −4 examples/embedding/embedding.cpp
+5 −0 examples/gguf/CMakeLists.txt
+3 −0 examples/gguf/gguf.cpp
+6 −1 examples/llama-bench/llama-bench.cpp
+82 −14 examples/main/main.cpp
+121 −28 examples/perplexity/perplexity.cpp
+2 −2 examples/quantize/quantize.cpp
+2 −2 examples/save-load-state/save-load-state.cpp
+8 −10 examples/server/server.cpp
+2 −2 examples/simple/simple.cpp
+7 −7 examples/train-text-from-scratch/README.md
+492 −0 examples/train-text-from-scratch/convert-train-checkpoint-to-gguf.py
+1,152 −2,250 examples/train-text-from-scratch/train-text-from-scratch.cpp
+7 −3 ggml-alloc.c
+23 −11 ggml-cuda.cu
+1 −0 ggml-metal.h
+81 −19 ggml-metal.m
+315 −206 ggml.c
+41 −18 ggml.h
+29 −7 gguf-py/gguf/gguf.py
+2 −2 k_quants.c
+120 −48 llama.cpp
+10 −5 llama.h
+140 −0 run_with_preset.py
+26 −0 scripts/convert-gg.sh
+2 −0 scripts/qnt-all.sh
+2 −0 scripts/run-all-perf.sh
+2 −0 scripts/run-all-ppl.sh
+4 −2 tests/CMakeLists.txt
+37 −17 tests/test-grad0.cpp
+178 −0 tests/test-tokenizer-0-falcon.cpp
+83 −0 tests/test-tokenizer-0-falcon.py
+182 −0 tests/test-tokenizer-0-llama.cpp
+95 −0 tests/test-tokenizer-0-llama.py
+0 −141 tests/test-tokenizer-0.cpp
+3 −11 tests/test-tokenizer-1.cpp
5 changes: 5 additions & 0 deletions scripts/bootstrap.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,6 +3,11 @@
git submodule init
git submodule update --recursive

cd llama.cpp
./scripts/build-info.sh > build-info.h
cd -

cp ./llama.cpp/build-info.h ./cpp/build-info.h
cp ./llama.cpp/ggml.h ./cpp/ggml.h
cp ./llama.cpp/ggml.c ./cpp/ggml.c
cp ./llama.cpp/ggml-metal.h ./cpp/ggml-metal.h
Expand Down
8 changes: 4 additions & 4 deletions scripts/ggml-metal.m.patch
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
--- ggml-metal-orig.m 2023-08-27 09:44:37
+++ ggml-metal.m 2023-08-27 09:46:07
@@ -226,13 +226,13 @@
--- ggml-metal-orig.m 2023-08-29 10:32:31
+++ ggml-metal.m 2023-08-29 10:32:32
@@ -230,13 +230,13 @@
#undef LM_GGML_METAL_ADD_KERNEL
}

Expand All @@ -20,7 +20,7 @@

return ctx;
}
@@ -374,15 +374,15 @@
@@ -436,15 +436,15 @@
}
}

Expand Down
Loading