Skip to content

Commit

Permalink
Fix preci failures
Browse files Browse the repository at this point in the history
  • Loading branch information
z5269887 committed Apr 23, 2024
1 parent 141eb51 commit e0a3679
Show file tree
Hide file tree
Showing 2 changed files with 5 additions and 5 deletions.
6 changes: 3 additions & 3 deletions examples/quantize/test.sh
Original file line number Diff line number Diff line change
Expand Up @@ -42,7 +42,7 @@ echo PASS
echo

# 3. Requant model with '--keep_split'
$QUANTIZE --allow-requantize --keep_split $WORK_PATH/ggml-model-split-00001-of-00006.gguf $WORK_PATH/ggml-model-requant.gguf Q4_K
$QUANTIZE --allow-requantize --keep_split $WORK_PATH/ggml-model-split-00001-of-00006.gguf $WORK_PATH/ggml-model-requant.gguf Q4_K
echo PASS
echo

Expand All @@ -52,7 +52,7 @@ echo PASS
echo

# 4. Requant mode without '--keep_split'
$QUANTIZE --allow-requantize $WORK_PATH/ggml-model-split-00001-of-00006.gguf $WORK_PATH/ggml-model-requant-merge.gguf Q4_K
$QUANTIZE --allow-requantize $WORK_PATH/ggml-model-split-00001-of-00006.gguf $WORK_PATH/ggml-model-requant-merge.gguf Q4_K
echo PASS
echo

Expand All @@ -62,4 +62,4 @@ echo PASS
echo

# Clean up
rm -f $WORK_PATH/ggml-model-split*.gguf $WORK_PATH/ggml-model-requant*.gguf
rm -f $WORK_PATH/ggml-model-split*.gguf $WORK_PATH/ggml-model-requant*.gguf
4 changes: 2 additions & 2 deletions llama.cpp
Original file line number Diff line number Diff line change
Expand Up @@ -13618,13 +13618,13 @@ static void llama_model_quantize_internal(const std::string & fname_inp, const s

// Set split info if needed
if (n_split > 1) {
for (int i = 0; i < ctx_outs.size(); ++i) {
for (size_t i = 0; i < ctx_outs.size(); ++i) {
gguf_set_val_u16(ctx_outs[i], ml.llm_kv(LLM_KV_SPLIT_NO).c_str(), i);
gguf_set_val_u16(ctx_outs[i], ml.llm_kv(LLM_KV_SPLIT_COUNT).c_str(), n_split);
gguf_set_val_i32(ctx_outs[i], ml.llm_kv(LLM_KV_SPLIT_TENSORS_COUNT).c_str(), ml.n_tensors);
}
}

int cur_split = -1;
std::ofstream fout;
auto close_ofstream = [&]() {
Expand Down

0 comments on commit e0a3679

Please sign in to comment.