support mimi model export

Summary: Compare the the xnnpack and eager model

Reviewed By: iseeyuan

Differential Revision: D71634148

Author: billmguo

examples/models/moshi/mimi/test_mimi.py

@@ -13,15 +13,26 @@
     get_symmetric_quantization_config,
     XNNPACKQuantizer,
 )
+from executorch.devtools.backend_debug import print_delegation_info
 from executorch.exir import to_edge_transform_and_lower
+from executorch.runtime import Runtime
 
 from huggingface_hub import hf_hub_download
 from moshi.models import loaders
-from torch.ao.quantization.quantize_pt2e import (
-    convert_pt2e,
-    prepare_pt2e,
-)
+from torch.ao.quantization.quantize_pt2e import convert_pt2e, prepare_pt2e
 from torch.export import export, ExportedProgram
+from torch.utils._pytree import tree_flatten
+
+
+def compute_sqnr(x: torch.Tensor, y: torch.Tensor) -> float:
+    assert x.shape == y.shape, "Tensor shapes do not match"
+    x = x.float()
+    y = y.float()
+    error = x - y
+    original_power = torch.mean(torch.pow(x, 2))
+    error_power = torch.mean(torch.pow(error, 2))
+    sqnr = 10 * torch.log10(original_power / error_power)
+    return sqnr.item()
 
 
 def read_mp3_from_url(url):
@@ -189,6 +200,59 @@ def forward(self, x):
         ep_encode_output = exported_encode.module()(chunk)
         self.assertTrue(torch.allclose(ep_encode_output, ref_encode_output, atol=1e-6))
 
+    def test_exported_decoder_xnnpack(self):
+        class MimiDecode(nn.Module):
+            def __init__(self, mimi: nn.Module):
+                super().__init__()
+                self.mimi_model = mimi
+
+            def forward(self, x):
+                return self.mimi_model.decode(x)
+
+        sample_pcm = torch.tensor(self.sample_pcm, device=self.device)[None]
+        pcm_chunk_size = int(self.mimi.sample_rate / self.mimi.frame_rate)
+        chunk = sample_pcm[..., 0:pcm_chunk_size]
+        input = self.mimi.encode(chunk)
+
+        mimi_decode = MimiDecode(self.mimi)
+        exported_decode: ExportedProgram = export(mimi_decode, (input,), strict=False)
+        quantization_config = get_symmetric_quantization_config(
+            is_per_channel=True,
+            is_dynamic=True,
+        )
+        quantizer = XNNPACKQuantizer()
+        quantizer.set_global(quantization_config)
+        m = exported_decode.module()
+        m = prepare_pt2e(m, quantizer)
+        m(input)
+        m = convert_pt2e(m)
+        print("quantized graph:")
+        print(m.graph)
+        # Export quantized module
+        exported_decode: ExportedProgram = export(m, (input,), strict=False)
+        # Lower
+        edge_manager = to_edge_transform_and_lower(
+            exported_decode,
+            partitioner=[XnnpackPartitioner()],
+        )
+        print("delegate graph:")
+        print_delegation_info(edge_manager.exported_program().graph_module)
+        exec_prog = edge_manager.to_executorch()
+        output_file = "/tmp/mimi_decode.pte"
+        with open(output_file, "wb") as file:
+            exec_prog.write_to_file(file)
+
+        eager_res = mimi_decode(input)
+        runtime = Runtime.get()
+        program = runtime.load_program(output_file)
+        method = program.load_method("forward")
+        flattened_x = tree_flatten(input)[0]
+        res = method.execute(flattened_x)
+        # Compare results
+        sqnr = compute_sqnr(eager_res, res[0])
+        print(f"SQNR: {sqnr}")
+        torch.testing.assert_close(eager_res, res[0], atol=1e-3, rtol=1e-3)
+
 
 if __name__ == "__main__":
     unittest.main()