init

Author: metascroy

.ci/scripts/test_llama_torchao_lowbit.sh

@@ -78,7 +78,6 @@ ${PYTHON_EXECUTABLE} -m examples.models.llama.export_llama \
     -qmode "torchao:8da${QLINEAR_BITWIDTH}w" \
     --group_size ${QLINEAR_GROUP_SIZE} \
     -E "torchao:${QEMBEDDING_BITWIDTH},${QEMBEDDING_GROUP_SIZE}" \
-    --disable_dynamic_shape \
     -d fp32
 
 # Test run

examples/models/llama/README.md

@@ -382,7 +382,7 @@ Please refer to [this tutorial](https://pytorch.org/executorch/main/llm/llama-de
 
 ## Running with low-bit kernels
 
-We now give instructions for quantizating and running your model with low-bit kernels.  These are still experimental, and require you do development on an Arm-based Mac.  Also note that low-bit quantization often requires QAT (quantization-aware training) to give good quality results.  Currently dynamic shapes must be disabled when exporting a model with these kernels.
+We now give instructions for quantizating and running your model with low-bit kernels.  These are still experimental, and require you do development on an Arm-based Mac.  Also note that low-bit quantization often requires QAT (quantization-aware training) to give good quality results.
 
 First export your model for lowbit quantization (step 2 above):
 
@@ -408,7 +408,6 @@ python -m examples.models.llama.export_llama \
   -qmode "torchao:8da${QLINEAR_BITWIDTH}w" \
   --group_size ${QLINEAR_GROUP_SIZE} \
   -E "torchao:${QEMBEDDING_BITWIDTH},${QEMBEDDING_GROUP_SIZE}" \
-  --disable_dynamic_shape \
   -d fp32
 ```