Qualcomm AI Engine Direct - GA Model Enablement (Roberta) (#11354)

### Summary
- support e2e script / test case for GA RoBERTa model
- perf: 16a8w 79.2ms/inf (SM8750) (seq_len=107)
- acc: ~= 50%

### Test plan
```
python backends/qualcomm/tests/test_qnn_delegate.py -k TestExampleOssScript.test_roberta --device <device_id> --host <host> --model <soc_model> --build_folder build-android --executorch_root . --sentence_dataset wikisent2.txt --artifact roberta_artifact
```

Author: chenweng-quic

backends/qualcomm/_passes/i64_to_i32.py

@@ -28,8 +28,10 @@ class I64toI32(ExportPass):
     I64_OPS = {
         exir_ops.edge.aten.argmin.default,
         exir_ops.edge.aten.arange.start_step,
+        exir_ops.edge.aten.cumsum.default,
         exir_ops.edge.aten.full.default,
         exir_ops.edge.aten.scalar_tensor.default,
+        exir_ops.edge.dim_order_ops._to_dim_order_copy.default,
     }
     # This dict is to ensure that the input of the OPs are int64 due to Pytorch restrictions.
     # For example, scatter op can only accept args[2], the index, as int64.

backends/qualcomm/_passes/lift_constant_scalar_operands.py

@@ -86,7 +86,8 @@ def _build_tensor_constant(
             dtype=(
                 node.args[0].meta["val"].dtype
                 if not is_float_tensor(node)
-                and not SCALAR_OPS.get(node.target).use_self_dtype
+                and (info := SCALAR_OPS.get(node.target))
+                and not info.use_self_dtype
                 else node.meta["val"].dtype
             ),
             device=node.meta["val"].device,

backends/qualcomm/_passes/replace_inf_values.py

@@ -30,6 +30,12 @@ def call(self, graph_module: torch.fx.GraphModule):
                     arg_list[index] = torch.finfo(torch.float32).min
                 elif arg == float("inf"):
                     arg_list[index] = torch.finfo(torch.float32).max
+
+            if node.target == torch.ops.aten.masked_fill.Scalar:
+                if arg_list[2] == torch.finfo(torch.float32).max:
+                    arg_list[2] = 255
+                elif arg_list[2] == torch.finfo(torch.float32).min:
+                    arg_list[2] = -255
             node.args = tuple(arg_list)
 
         graph_module.recompile()

backends/qualcomm/builders/op_cum_sum.py

@@ -51,6 +51,8 @@ def define_node(
         dim = self.get_param(node, input_tensor)
 
         output_tensor = self.get_tensor(node, node)
+        if output_tensor.dtype == torch.int64:
+            output_tensor = output_tensor.to(torch.int32)
         output_tensor_wrapper = self.define_tensor(
             node,
             node,

backends/qualcomm/tests/models.py

@@ -1101,6 +1101,16 @@ def forward(self, x):
         return torch.mean(x, (-1, -2))
 
 
+class MaskedFill(torch.nn.Module):
+    def __init__(self):
+        super().__init__()
+
+    def forward(self, attn_mask):
+        return attn_mask.masked_fill(attn_mask != 0, float(-100.0)).masked_fill(
+            attn_mask == 0, float(0.0)
+        )
+
+
 class Maximum(torch.nn.Module):
     def __init__(self):
         super().__init__()
@@ -1751,16 +1761,6 @@ def forward(self, x):
         )
 
 
-class MaskedFill(torch.nn.Module):
-    def __init__(self):
-        super().__init__()
-
-    def forward(self, attn_mask):
-        return attn_mask.masked_fill(attn_mask != 0, float(-100.0)).masked_fill(
-            attn_mask == 0, float(0.0)
-        )
-
-
 # Mimi Decoder has 0D tensor which QNN cannot handle.
 class ZeroDimTensor(torch.nn.Module):
     def __init__(self):

backends/qualcomm/tests/test_qnn_delegate.py

@@ -272,9 +272,24 @@ def test_qnn_backend_cos(self):
         self.lower_module_and_test_output(module, sample_input)
 
     def test_qnn_backend_cumsum(self):
-        module = CumSum()  # noqa: F405
-        sample_input = (torch.randn(4),)
-        self.lower_module_and_test_output(module, sample_input)
+        sample_input = ()
+        test_comb = [
+            {
+                QCOM_MODULE: [CumSum()],  # noqa: F405
+                QCOM_SAMPLE_INPUTS: [
+                    (torch.randn(4),),
+                    (torch.randint(0, 10, size=(4,)),),
+                ],
+            }
+        ]
+
+        index = 0
+        for comb in test_comb:
+            for module in comb[QCOM_MODULE]:
+                for sample_input in comb[QCOM_SAMPLE_INPUTS]:
+                    with self.subTest(i=index):
+                        self.lower_module_and_test_output(module, sample_input)
+                        index += 1
 
     def test_qnn_backend_einsum_outer_product(self):
         module = EinsumOuterProduct()  # noqa: F405
@@ -311,6 +326,12 @@ def test_qnn_backend_element_wise_add(self):
                 QCOM_MODULE: [AddConstantFloat()],  # noqa: F405
                 QCOM_SAMPLE_INPUTS: [(torch.randn(2, 5, 1, 3),)],
             },
+            {
+                QCOM_MODULE: [
+                    AddConstantLong(),  # noqa: F405
+                ],
+                QCOM_SAMPLE_INPUTS: [(torch.randint(0, 10, size=(2, 3)),)],
+            },
         ]
 
         index = 0
@@ -4526,6 +4547,40 @@ def test_retinanet(self):
             else:
                 self.assertGreaterEqual(msg["mAP"], 0.6)
 
+    def test_roberta(self):
+        if not self.required_envs([self.sentence_dataset]):
+            self.skipTest("missing required envs")
+        cmds = [
+            "python",
+            f"{self.executorch_root}/examples/qualcomm/oss_scripts/roberta.py",
+            "--dataset",
+            self.sentence_dataset,
+            "--artifact",
+            self.artifact_dir,
+            "--build_folder",
+            self.build_folder,
+            "--device",
+            self.device,
+            "--model",
+            self.model,
+            "--ip",
+            self.ip,
+            "--port",
+            str(self.port),
+        ]
+        if self.host:
+            cmds.extend(["--host", self.host])
+
+        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:
+                self.assertGreaterEqual(msg["accuracy"], 0.5)
+
     def test_squeezenet(self):
         if not self.required_envs([self.image_dataset]):
             self.skipTest("missing required envs")
@@ -5344,6 +5399,11 @@ def setup_environment():
         help="Location for imagenet dataset",
         type=str,
     )
+    parser.add_argument(
+        "--sentence_dataset",
+        help="Location for sentence dataset",
+        type=str,
+    )
     parser.add_argument(
         "-p",
         "--pretrained_weight",
@@ -5402,6 +5462,7 @@ def setup_environment():
     TestQNN.executorch_root = args.executorch_root
     TestQNN.artifact_dir = args.artifact_dir
     TestQNN.image_dataset = args.image_dataset
+    TestQNN.sentence_dataset = args.sentence_dataset
     TestQNN.pretrained_weight = args.pretrained_weight
     TestQNN.model_name = args.model_name
     TestQNN.online_prepare = args.online_prepare

backends/qualcomm/tests/utils.py

@@ -183,6 +183,7 @@ class TestQNN(unittest.TestCase):
     executorch_root: str = ""
     artifact_dir: str = ""
     image_dataset: str = ""
+    sentence_dataset: str = ""
     pretrained_weight: str = ""
     enable_profile: bool = False
     op_package_dir: str = ""

examples/qualcomm/oss_scripts/roberta.py

@@ -0,0 +1,163 @@
+# Copyright (c) Qualcomm Innovation Center, Inc.
+# All rights reserved
+#
+# This source code is licensed under the BSD-style license found in the
+# LICENSE file in the root directory of this source tree.
+
+import getpass
+import json
+import os
+from multiprocessing.connection import Client
+
+import evaluate
+import numpy as np
+import torch
+
+from executorch.backends.qualcomm._passes.qnn_pass_manager import (
+    get_capture_program_passes,
+)
+from executorch.backends.qualcomm.quantizer.quantizer import QuantDtype
+
+from executorch.examples.qualcomm.utils import (
+    build_executorch_binary,
+    get_masked_language_model_dataset,
+    make_output_dir,
+    parse_skip_delegation_node,
+    setup_common_args_and_variables,
+    SimpleADB,
+)
+from transformers import AutoModelForMaskedLM, AutoTokenizer
+
+
+def get_instance(args):
+    module = AutoModelForMaskedLM.from_pretrained("xlm-roberta-base").eval()
+    return module
+
+
+def main(args):
+    skip_node_id_set, skip_node_op_set = parse_skip_delegation_node(args)
+
+    os.makedirs(args.artifact, exist_ok=True)
+    data_size = 100
+
+    tokenizer = AutoTokenizer.from_pretrained("xlm-roberta-base")
+    inputs, targets, input_list = get_masked_language_model_dataset(
+        args.dataset, tokenizer, data_size
+    )
+
+    # Get the Roberta model.
+    model = get_instance(args)
+    pte_filename = "roberta_qnn"
+
+    # lower to QNN
+    passes_job = get_capture_program_passes()
+    build_executorch_binary(
+        model,
+        inputs[0],
+        args.model,
+        f"{args.artifact}/{pte_filename}",
+        dataset=inputs,
+        skip_node_id_set=skip_node_id_set,
+        skip_node_op_set=skip_node_op_set,
+        quant_dtype=QuantDtype.use_16a8w,
+        passes_job=passes_job,
+        shared_buffer=args.shared_buffer,
+    )
+
+    if args.compile_only:
+        return
+
+    workspace = f"/data/local/tmp/{getpass.getuser()}/executorch/{pte_filename}"
+    pte_path = f"{args.artifact}/{pte_filename}.pte"
+
+    adb = SimpleADB(
+        qnn_sdk=os.getenv("QNN_SDK_ROOT"),
+        build_path=f"{args.build_folder}",
+        pte_path=pte_path,
+        workspace=workspace,
+        device_id=args.device,
+        host_id=args.host,
+        soc_model=args.model,
+    )
+    output_data_folder = f"{args.artifact}/outputs"
+    make_output_dir(output_data_folder)
+
+    # demo
+    mask_token = tokenizer.mask_token
+    text = f"Hello I'm a {mask_token} model."
+    sample_input = tokenizer(
+        text,
+        return_tensors="pt",
+        padding="max_length",
+        max_length=inputs[0][0].shape[1],
+    )
+    sample_input["input_ids"] = sample_input["input_ids"].to(torch.int32)
+    sample_input["attention_mask"] = sample_input["attention_mask"].to(torch.float32)
+    sample_input = tuple(sample_input.values())
+    golden = model(*sample_input)[0]
+    adb.push(inputs=[sample_input], input_list="input_0_0.raw input_0_1.raw\n")
+    adb.execute()
+    adb.pull(output_path=args.artifact)
+
+    print(f"input: {tokenizer.batch_decode(sample_input[0])}")
+    print(f"golden output: {tokenizer.batch_decode(golden.argmax(axis=2))}")
+    predictions = np.fromfile(
+        os.path.join(output_data_folder, "output_0_0.raw"), dtype=np.float32
+    ).reshape([1, inputs[0][0].shape[1], -1])
+    print(f"QNN output: {tokenizer.batch_decode(predictions.argmax(axis=2))}")
+
+    # accuracy analysis
+    adb.push(inputs=inputs, input_list=input_list)
+    adb.execute()
+    adb.pull(output_path=args.artifact)
+    goldens, predictions = [], []
+    for i in range(len(inputs)):
+        indice = [i for i, x in enumerate(targets[i]) if x != -100]
+        goldens.extend(targets[i][indice].tolist())
+        prediction = (
+            np.fromfile(
+                os.path.join(output_data_folder, f"output_{i}_0.raw"), dtype=np.float32
+            )
+            .reshape([1, inputs[0][0].shape[1], -1])
+            .argmax(axis=-1)
+        )
+        predictions.extend(prediction[0, indice].tolist())
+    metric = evaluate.load("accuracy")
+    results = metric.compute(predictions=predictions, references=goldens)
+    if args.ip and args.port != -1:
+        with Client((args.ip, args.port)) as conn:
+            conn.send(json.dumps({"accuracy": results["accuracy"]}))
+    else:
+        print(f"accuracy: {results['accuracy']}")
+
+
+if __name__ == "__main__":
+    parser = setup_common_args_and_variables()
+    parser.add_argument(
+        "-a",
+        "--artifact",
+        help="path for storing generated artifacts and output by this example. Default ./Roberta_qnn",
+        default="./Roberta_qnn",
+        type=str,
+    )
+    parser.add_argument(
+        "-d",
+        "--dataset",
+        help=(
+            "path to the validation text. "
+            "e.g. --dataset wikisent2.txt "
+            "for https://www.kaggle.com/datasets/mikeortman/wikipedia-sentences"
+        ),
+        type=str,
+        required=True,
+    )
+
+    args = parser.parse_args()
+    try:
+        main(args)
+    except Exception as e:
+        if args.ip and args.port != -1:
+            with Client((args.ip, args.port)) as conn:
+                conn.send(json.dumps({"Error": str(e)}))
+        else:
+            raise Exception(e)