Remove cpp, python, and cmake files

Author: winskuo-quic

backends/qualcomm/tests/test_qnn_delegate.py

@@ -4,7 +4,6 @@
 # This source code is licensed under the BSD-style license found in the
 # LICENSE file in the root directory of this source tree.
 import csv
-import io
 import itertools
 import json
 import logging
@@ -8517,207 +8516,6 @@ def test_whisper(self):
                 self.assertLessEqual(msg["wer"], 0.25)
 
 
-class TestExampleQaihubScript(TestQNN):
-    def test_utils_export(self):
-        with tempfile.TemporaryDirectory() as tmp_dir:
-            module = ContextBinaryExample()  # noqa: F405
-            generate_context_binary(
-                module=module,
-                inputs=module.example_inputs(),
-                quantized=True,
-                artifact_dir=tmp_dir,
-            )
-            ctx_path = f"{tmp_dir}/model_ctx.bin"
-            fpath = f"{self.executorch_root}/examples/qualcomm/qaihub_scripts/utils/export.py"
-
-            # do compilation
-            compile_cmds = [
-                "python",
-                fpath,
-                "compile",
-                "-a",
-                ctx_path,
-                "-m",
-                self.soc_model,
-                "-l",
-                "False",
-                "-b",
-                self.build_folder,
-                "-o",
-                f"{tmp_dir}/output_pte",
-            ]
-            compile_process = subprocess.Popen(
-                compile_cmds, stdout=subprocess.DEVNULL, cwd=self.executorch_root
-            )
-            output_pte_dir = f"{tmp_dir}/output_pte/model_ctx"
-            compile_process.communicate()
-
-            # check artifacts are correctly generated
-            self.assertTrue(
-                all(
-                    [
-                        Path(output_pte_dir).exists(),
-                        Path(f"{output_pte_dir}/model_ctx.json").exists(),
-                        Path(f"{output_pte_dir}/model_ctx.svg").exists(),
-                    ]
-                )
-            )
-
-            # prepare input files
-            input_list, inputs = [], module.example_inputs()
-            for name, tensor in inputs.items():
-                tensor_path = f"{output_pte_dir}/{name}.pt"
-                torch.save(tensor, tensor_path)
-                input_list.append(tensor_path)
-
-            # do execution
-            output_data_dir = f"{tmp_dir}/output_data"
-            execute_cmds = [
-                "python",
-                fpath,
-                "execute",
-                "-p",
-                output_pte_dir,
-                "-i",
-                *input_list,
-                "-s",
-                self.device,
-                "-z",
-                "-b",
-                self.build_folder,
-                "-o",
-                output_data_dir,
-            ]
-            if self.host is not None:
-                execute_cmds.append(f"-H {self.host}")
-            execute_process = subprocess.Popen(execute_cmds, cwd=self.executorch_root)
-            execute_process.communicate()
-
-            # read outputs
-            with open(f"{output_pte_dir}/model_ctx.json", "r") as f:
-                graph_info = json.load(f)
-
-            device_output = []
-            for output in graph_info["outputs"]:
-                with open(f"{output_data_dir}/{output['name']}.pt", "rb") as f:
-                    buffer = io.BytesIO(f.read())
-                    device_output.append(torch.load(buffer, weights_only=False))
-
-            # validate outputs
-            golden_output = module.forward(inputs["x"], inputs["y"])
-            self.atol, self.rtol = 1e-1, 1
-            self._assert_outputs_equal(golden_output, device_output)
-
-    def test_llama2_7b(self):
-        if not self.required_envs():
-            self.skipTest("missing required envs")
-
-        prompt = "Explain the rules of baseball"
-        cmds = [
-            "python",
-            f"{self.executorch_root}/examples/qualcomm/qaihub_scripts/llama/llama2/qaihub_llama2_7b.py",
-            "--artifact",
-            self.artifact_dir,
-            "--build_folder",
-            self.build_folder,
-            "--tokenizer_bin",
-            f"{self.artifact_dir}/tokenizer.bin",
-            "--context_binaries",
-            f"{self.artifact_dir}",
-            "--prompt",
-            f"{prompt}",
-        ]
-        self.add_default_cmds(cmds)
-
-        p = subprocess.Popen(cmds, stdout=subprocess.DEVNULL)
-        with Listener((self.ip, self.port)) as listener:
-            conn = listener.accept()
-            p.communicate()
-            msg = json.loads(conn.recv())
-            if "Error" in msg:
-                self.fail(msg["Error"])
-            else:
-                model_out = msg["result"]
-                self.assertTrue(model_out.startswith(prompt))
-
-    def test_llama3_8b(self):
-        if not self.required_envs():
-            self.skipTest("missing required envs")
-
-        prompt = "Explain the rules of baseball"
-        cmds = [
-            "python",
-            f"{self.executorch_root}/examples/qualcomm/qaihub_scripts/llama/llama3/qaihub_llama3_8b.py",
-            "--artifact",
-            self.artifact_dir,
-            "--build_folder",
-            self.build_folder,
-            "--tokenizer_model",
-            f"{self.artifact_dir}/tokenizer.model",
-            "--context_binaries",
-            f"{self.artifact_dir}",
-            "--prompt",
-            f"{prompt}",
-        ]
-        self.add_default_cmds(cmds)
-
-        p = subprocess.Popen(cmds, stdout=subprocess.DEVNULL)
-        with Listener((self.ip, self.port)) as listener:
-            conn = listener.accept()
-            p.communicate()
-            msg = json.loads(conn.recv())
-            if "Error" in msg:
-                self.fail(msg["Error"])
-            else:
-                model_out = msg["result"]
-                expected_result = (
-                    "<|begin_of_text|><|start_header_id|>user<|end_header_id|>\n\n"
-                    + prompt
-                    + "<|eot_id|><|start_header_id|>assistant<|end_header_id|>"
-                )
-                self.assertTrue(model_out.startswith(expected_result))
-
-    def test_stable_diffusion(self):
-        if not self.required_envs():
-            self.skipTest("missing required envs")
-
-        prompt = "a photo of an astronaut riding a horse on mars"
-        cmds = [
-            "python",
-            f"{self.executorch_root}/examples/qualcomm/qaihub_scripts/stable_diffusion/qaihub_stable_diffusion.py",
-            "--artifact",
-            self.artifact_dir,
-            "--build_folder",
-            self.build_folder,
-            "--text_encoder_bin",
-            f"{self.artifact_dir}/text_encoder.serialized.bin",
-            "--unet_bin",
-            f"{self.artifact_dir}/unet.serialized.bin",
-            "--vae_bin",
-            f"{self.artifact_dir}/vae.serialized.bin",
-            "--vocab_json",
-            f"{self.artifact_dir}/vocab.json",
-            "--num_time_steps",
-            "20",
-            "--prompt",
-            f"{prompt}",
-            "--fix_latents",
-        ]
-        self.add_default_cmds(cmds)
-
-        p = subprocess.Popen(cmds, stdout=subprocess.DEVNULL)
-        with Listener((self.ip, self.port)) as listener:
-            conn = listener.accept()
-            p.communicate()
-            msg = json.loads(conn.recv())
-            if "Error" in msg:
-                self.fail(msg["Error"])
-            else:
-                # For the default settings and prompt, the expected results will be {PSNR: 23.258, SSIM: 0.852}
-                self.assertGreaterEqual(msg["PSNR"], 20)
-                self.assertGreaterEqual(msg["SSIM"], 0.8)
-
-
 class TestExampleScript(TestQNN):
     def test_mobilenet_v2(self):
         if not self.required_envs([self.image_dataset]):

examples/qualcomm/CMakeLists.txt

@@ -80,12 +80,6 @@ add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/oss_scripts/t5)
 # build qnn_whisper_runner for whisper
 add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/oss_scripts/whisper)
 
-# build qaihub_llama2_7b_runner and qaihub_llama3_8b_runner
-add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/qaihub_scripts/llama)
-
-# build qaihub_stable_diffusion_runner
-add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/qaihub_scripts/stable_diffusion)
-
 # direct-mode qnn_executor_direct_runner is build with ndk toolchain, paired
 # with libraries built with hexagon toolchain, communicate through self-defined
 # fastrpc protocol.