ci : use -no-cnv in gguf-split tests (#11254)

* ci : use -no-cnv in gguf-split tests

ggml-ci

* ci : use -no-cnv in requantize tests

ggml-ci

* scripts : fix [no ci]
This commit is contained in:
Georgi Gerganov 2025-01-15 18:28:35 +02:00 committed by GitHub
parent 1d8504338e
commit f11cfdfd7f
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
4 changed files with 10 additions and 122 deletions

View file

@ -80,18 +80,18 @@ run_conversion_and_inference_lora() {
# Run inference
echo -e "\n\n---------------------------\n\n"
echo "Running llama-cli without lora for $model_name with hidden_size $hidden_size..."
OUTPUT_BASE=$(./llama-cli -m $MODELS_REPO/$model_name/hidden_size=$hidden_size/base/Base-F32.gguf \
OUTPUT_BASE=$(./llama-cli -no-cnv -m $MODELS_REPO/$model_name/hidden_size=$hidden_size/base/Base-F32.gguf \
-p "$EXPECTED_BASE_FIRST_WORD" -n 50 --seed 42 --temp 0)
echo -e "\n\n---------------------------\n\n"
echo "Running llama-cli with hot lora for $model_name with hidden_size $hidden_size..."
OUTPUT_LORA_HOT=$(./llama-cli -m $MODELS_REPO/$model_name/hidden_size=$hidden_size/base/Base-F32.gguf \
OUTPUT_LORA_HOT=$(./llama-cli -no-cnv -m $MODELS_REPO/$model_name/hidden_size=$hidden_size/base/Base-F32.gguf \
--lora $MODELS_REPO/$model_name/hidden_size=$hidden_size/lora/Lora-F32-LoRA.gguf \
-p "$EXPECTED_LORA_FIRST_WORD" -n 50 --seed 42 --temp 0)
echo -e "\n\n---------------------------\n\n"
echo "Running llama-cli with merged lora for $model_name with hidden_size $hidden_size..."
OUTPUT_LORA_MERGED=$(./llama-cli -m $MODELS_REPO/$model_name/hidden_size=$hidden_size/base/Base-F32-lora-merged.gguf \
OUTPUT_LORA_MERGED=$(./llama-cli -no-cnv -m $MODELS_REPO/$model_name/hidden_size=$hidden_size/base/Base-F32-lora-merged.gguf \
-p "$EXPECTED_LORA_FIRST_WORD" -n 50 --seed 42 --temp 0)
# Remove any initial white space