Move llama_transformer to extension/llm

Author: jackzhxng

backends/arm/test/models/test_llama.py

@@ -22,13 +22,13 @@
     TosaPipelineBI,
     TosaPipelineMI,
 )
-
-from executorch.extension.llm.export.config.llm_config import LlmConfig
 from executorch.examples.models.llama.export_llama_lib import (
     build_args_parser,
     get_llama_model,
 )
 
+from executorch.extension.llm.export.config.llm_config import LlmConfig
+
 input_t = Tuple[torch.Tensor]
 
 # Add project dir to sys path to workaround importlib.import_module() conditions in model_factory.py

examples/apple/mps/scripts/mps_example.py

@@ -19,8 +19,6 @@
 from executorch.devtools.bundled_program.serialize import (
     serialize_from_bundled_program_to_flatbuffer,
 )
-
-from executorch.extension.llm.export.config.llm_config import LlmConfig
 from executorch.exir import (
     EdgeCompileConfig,
     EdgeProgramManager,
@@ -31,6 +29,8 @@
 from executorch.exir.capture._config import ExecutorchBackendConfig
 from executorch.extension.export_util.utils import export_to_edge, save_pte_program
 
+from executorch.extension.llm.export.config.llm_config import LlmConfig
+
 from ....models import MODEL_NAME_TO_MODEL
 from ....models.model_factory import EagerModelFactory
 

examples/models/llama/TARGETS

@@ -13,10 +13,6 @@ runtime.python_library(
     name = "llama_transformer",
     srcs = [
         "llama_transformer.py",
-        "rope.py",
-        "attention.py",
-        "model_args.py",
-        "norm.py",
     ],
     _is_external_target = True,
     base_module = "executorch.examples.models.llama",
@@ -26,23 +22,21 @@ runtime.python_library(
     ],
     deps = [
         "//caffe2:torch",
+        "//executorch/extension/llm/modeling/text_decoder:text_decoder_attention",
+        "//executorch/extension/llm/modeling/text_decoder:text_decoder_model_args",
+        "//executorch/extension/llm/modeling/text_decoder:text_decoder_norm",
+        "//executorch/extension/llm/modeling/text_decoder:text_decoder_rope",
     ],
 )
 
 runtime.python_library(
     name = "static_attention",
-    srcs = [
-        "static_attention.py",
-    ],
-    _is_external_target = True,
-    base_module = "executorch.examples.models.llama",
     visibility = [
         "//executorch/...",
         "@EXECUTORCH_CLIENTS",
     ],
     deps = [
-        ":llama_transformer",
-        "//caffe2:torch",
+        "//executorch/extension/llm/modeling/text_decoder:text_decoder_static_attention",
     ],
 )