Skip to content

Commit a6a836c

Browse files
committed
Fix loading non-llama LLM via ios benchmark app
1 parent d12c6f6 commit a6a836c

File tree

2 files changed

+7
-3
lines changed

2 files changed

+7
-3
lines changed

.github/workflows/apple-perf.yml

Lines changed: 6 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -209,6 +209,8 @@ jobs:
209209
210210
# Convert HF checkpoint to ET via etLLM path
211211
if [[ "$HF_MODEL_REPO" == meta-llama/* ]]; then
212+
# The benchmark app replies on the _llm suffix to determine whether the model is a LLM or not
213+
OUT_ET_MODEL_NAME=${OUT_ET_MODEL_NAME}_llm
212214
# Llama models on Hugging Face
213215
if [[ ${{ matrix.config }} == "llama3_spinquant" ]]; then
214216
# SpinQuant
@@ -311,6 +313,7 @@ jobs:
311313
ls -lh "${OUT_ET_MODEL_NAME}.pte"
312314
fi
313315
elif [[ "$HF_MODEL_REPO" == "Qwen/Qwen3-0.6B" ]]; then
316+
OUT_ET_MODEL_NAME=${OUT_ET_MODEL_NAME}_llm
314317
if [[ ${{ matrix.config }} == "et_xnnpack_custom_spda_kv_cache_8da4w" ]]; then
315318
DOWNLOADED_PATH=$(bash .ci/scripts/download_hf_hub.sh --model_id "${HF_MODEL_REPO}" --subdir "." --files "tokenizer.json")
316319
${CONDA_RUN} python -m examples.models.llama.export_llama \
@@ -367,12 +370,13 @@ jobs:
367370
${CONDA_RUN} optimum-cli export executorch "${ARGS[@]}"
368371
popd
369372
373+
# The benchmark app replies on the _llm suffix to determine whether the model is a LLM or not
374+
OUT_ET_MODEL_NAME=${OUT_ET_MODEL_NAME}_llm
370375
mv model.pte ${OUT_ET_MODEL_NAME}.pte
371376
ls -lh "${OUT_ET_MODEL_NAME}.pte"
372377
fi
373378
374-
# zip -j model.zip ${OUT_ET_MODEL_NAME}.pte ${DOWNLOADED_PATH}/tokenizer.*
375-
zip -j model.zip ${OUT_ET_MODEL_NAME}.pte
379+
zip -j model.zip ${OUT_ET_MODEL_NAME}.pte ${DOWNLOADED_PATH}/tokenizer.*
376380
ls -lh model.zip
377381
mkdir -p "${ARTIFACTS_DIR_NAME}"
378382
mv model.zip "${ARTIFACTS_DIR_NAME}"

extension/benchmark/apple/Benchmark/Tests/LLaMA/LLaMATests.mm

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -60,7 +60,7 @@ @implementation LLaMATests
6060
+ (NSDictionary<NSString *, BOOL (^)(NSString *)> *)predicates {
6161
return @{
6262
@"model" : ^BOOL(NSString *filename){
63-
return [filename hasSuffix:@".pte"] && [filename.lowercaseString containsString:@"llama"];
63+
return [filename hasSuffix:@".pte"] && [filename.lowercaseString containsString:@"llm"];
6464
},
6565
@"tokenizer" : ^BOOL(NSString *filename) {
6666
return [filename isEqual:@"tokenizer.bin"] || [filename isEqual:@"tokenizer.model"] || [filename isEqual:@"tokenizer.json"];

0 commit comments

Comments
 (0)