fix llama BC checking

Author: DannyYuyang-quic

.ci/scripts/test_qnn_static_llama.sh

@@ -33,12 +33,12 @@ echo "Creating tokenizer.bin"
 $PYTHON_EXECUTABLE -m pytorch_tokenizers.tools.llama2c.convert -t tokenizer.model -o tokenizer.bin
 
 set +e
-# Compile only as weight sharing is not applicable on x86
-$PYTHON_EXECUTABLE backends/qualcomm/tests/test_qnn_delegate.py -k TestExampleLLMScript.test_llama_stories_110m --model SM8650 --build_folder build-android/ --executorch_root . --artifact_dir . --llama_artifacts . --compile_only
+# Compile only as weight sharing is not applicable on x86.
+$PYTHON_EXECUTABLE backends/qualcomm/tests/test_qnn_delegate.py -k TestExampleLLMScript.test_llama_stories_110m --model SM8650 --build_folder build-android/ --executorch_root . --artifact_dir ./stories_110m_pte_size --llama_artifacts . --compile_only
 exit_code1=$?
 
 # Checks accuracy with weight sharing disabled since x86 does not support weight sharing.
-$PYTHON_EXECUTABLE backends/qualcomm/tests/test_qnn_delegate.py -k TestExampleLLMScript.test_llama_stories_110m --model SM8650 --build_folder build-x86/ --executorch_root . --artifact_dir . --llama_artifacts . --enable_x86_64
+$PYTHON_EXECUTABLE backends/qualcomm/tests/test_qnn_delegate.py -k TestExampleLLMScript.test_llama_stories_110m --model SM8650 --build_folder build-x86/ --executorch_root . --artifact_dir ./stories_110m_accuracy --llama_artifacts . --enable_x86_64
 exit_code2=$?
 
 # Check BC

backends/qualcomm/bc/test_qnn_static_llama_bc.sh

@@ -13,15 +13,17 @@ fi
 which "${PYTHON_EXECUTABLE}"
 
 
-llama_artifacts="."
+llama_artifacts="260k_stories"
 PTE_ARTIFACT="examples/qualcomm/oss_scripts/llama/artifacts"
 
+mkdir ${llama_artifacts}
 # Download stories260K.pt and tokenizer from Github
-curl -Ls "https://huggingface.co/karpathy/tinyllamas/resolve/main/stories260K/stories260K.pt" --output stories260K.pt
-curl -Ls "https://huggingface.co/karpathy/tinyllamas/resolve/main/stories260K/tok512.model" --output tokenizer.model
+curl -Ls "https://huggingface.co/karpathy/tinyllamas/resolve/main/stories260K/stories260K.pt" --output ${llama_artifacts}/stories260K.pt
+curl -Ls "https://huggingface.co/karpathy/tinyllamas/resolve/main/stories260K/tok512.model" --output ${llama_artifacts}/tokenizer.model
+$PYTHON_EXECUTABLE -m pytorch_tokenizers.tools.llama2c.convert -t ${llama_artifacts}/tokenizer.model -o ${llama_artifacts}/tokenizer.bin
 # Create params.json file
-touch params.json
-echo '{"dim": 64, "n_layers": 5, "n_heads": 8, "n_kv_heads": 4, "vocab_size": 512, "multiple_of": 4, "max_seq_len": 512}' > params.json
+touch ${llama_artifacts}/params.json
+echo '{"dim": 64, "n_layers": 5, "n_heads": 8, "n_kv_heads": 4, "vocab_size": 512, "multiple_of": 4, "max_seq_len": 512}' > ${llama_artifacts}/params.json
 
 # Checks e2e accuracy
 expected=$($PYTHON_EXECUTABLE backends/qualcomm/tests/test_qnn_delegate.py -k TestExampleLLMScript.test_llama_stories_260k --model SM8650 --build_folder build-x86/ --executorch_root . --artifact_dir . --llama_artifacts $llama_artifacts --enable_x86_64 | grep "Model CI result:")

examples/qualcomm/oss_scripts/llama/llama.py

@@ -362,7 +362,6 @@ def compile(args, pte_filename, tokenizer):
     kv_config.use_kv_cache = True
     kv_config.enable_masked_softmax = args.enable_masked_softmax
     kv_config.enable_r3 = args.r3
-    kv_config.base_model_name_or_path = args.decoder_model
 
     prefill_config = copy.copy(kv_config)
     prefill_config.use_kv_cache = (
@@ -566,7 +565,6 @@ def permute(w, heads):
         llama_instance_list[i] = SingleLlama(
             llama_instance_list[i].eval(), pte_filename
         )
-
         if args.embedding_quantize:
             llama_instance_list[i].passes_job[I64toI32][
                 QCOM_PASS_ARGS_KWARGS_DEFAULTS_KEY