Update phi-3-mini to use the export library

as title.

Note: exporting model from HF is currently blocked by this issue, https://fburl.com/qpaapajr. We need to apply the following trick to export the model.
```
        model = torch.export._trace._export(
            model,
            example_inputs,
            dynamic_shapes=dynamic_shape,
            strict=False,
            pre_dispatch=False,
        )
```

Differential Revision: [D59503255](https://our.internmc.facebook.com/intern/diff/D59503255/)

ghstack-source-id: 233088631
Pull Request resolved: #4190

Author: helunwencser

examples/models/phi-3-mini/export_phi-3-mini.py

@@ -5,65 +5,46 @@
 # LICENSE file in the root directory of this source tree.
 
 import torch
+from executorch.extension.llm.export.builder import DType, LLMEdgeManager
 
-from executorch.backends.transforms.duplicate_dynamic_quant_chain import (
-    DuplicateDynamicQuantChainPass,
-)
-from executorch.backends.xnnpack.partition.xnnpack_partitioner import XnnpackPartitioner
-from executorch.backends.xnnpack.utils.configs import get_xnnpack_edge_compile_config
-from executorch.exir import to_edge
-from torch._export import capture_pre_autograd_graph
-from torch.ao.quantization.quantize_pt2e import convert_pt2e, prepare_pt2e
-
-from torch.ao.quantization.quantizer.xnnpack_quantizer import (
-    get_symmetric_quantization_config,
-    XNNPACKQuantizer,
+from executorch.extension.llm.export.partitioner_lib import get_xnnpack_partitioner
+from executorch.extension.llm.export.quantizer_lib import (
+    DynamicQuantLinearOptions,
+    get_pt2e_quantizers,
+    PT2EQuantOptions,
 )
 
 from transformers import Phi3ForCausalLM
 
 
 def main() -> None:
-    torch.random.manual_seed(0)
+    torch.manual_seed(42)
 
     model = Phi3ForCausalLM.from_pretrained("microsoft/Phi-3-mini-4k-instruct")
 
-    example_inputs = (torch.randint(0, 100, (1, 100), dtype=torch.long),)
-    dynamic_shape = {"input_ids": {1: torch.export.Dim("sequence_length", max=128)}}
-
-    xnnpack_quant_config = get_symmetric_quantization_config(
-        is_per_channel=True, is_dynamic=True
-    )
-    xnnpack_quantizer = XNNPACKQuantizer()
-    xnnpack_quantizer.set_global(xnnpack_quant_config)
-
-    with torch.nn.attention.sdpa_kernel(
-        [torch.nn.attention.SDPBackend.MATH]
-    ), torch.no_grad():
-        model = capture_pre_autograd_graph(
-            model, example_inputs, dynamic_shapes=dynamic_shape
+    modelname = "phi-3-mini"
+
+    (
+        LLMEdgeManager(
+            model=model,
+            modelname=modelname,
+            max_seq_len=128,
+            dtype=DType.fp32,
+            use_kv_cache=False,
+            example_inputs=(torch.randint(0, 100, (1, 100), dtype=torch.long),),
+            enable_dynamic_shape=True,
+            verbose=True,
         )
-        model = prepare_pt2e(model, xnnpack_quantizer)
-        model(*example_inputs)
-        model = convert_pt2e(model, fold_quantize=False)
-        DuplicateDynamicQuantChainPass()(model)
-        # TODO(lunwenh): update it to use export once
-        # https://github.com/pytorch/pytorch/issues/128394 is resolved.
-        model = torch.export._trace._export(
-            model,
-            example_inputs,
-            dynamic_shapes=dynamic_shape,
-            strict=False,
-            pre_dispatch=False,
+        .set_output_dir(".")
+        .capture_pre_autograd_graph()
+        .pt2e_quantize(
+            get_pt2e_quantizers(PT2EQuantOptions(None, DynamicQuantLinearOptions()))
         )
-
-    edge_config = get_xnnpack_edge_compile_config()
-    edge_manager = to_edge(model, compile_config=edge_config)
-    edge_manager = edge_manager.to_backend(XnnpackPartitioner(has_dynamic_shapes=True))
-    et_program = edge_manager.to_executorch()
-
-    with open("phi-3-mini.pte", "wb") as file:
-        file.write(et_program.buffer)
+        .export_to_edge()
+        .to_backend([get_xnnpack_partitioner()])
+        .to_executorch()
+        .save_to_pte(f"{modelname}.pte")
+    )
 
 
 if __name__ == "__main__":