Quantizer supported by the Arm Ethos-U backend. .. warning:: Setting ``use_composable_quantizer=True`` enables an experimental API surface that may change without notice. Args: compile_spec (EthosUCompileSpec): Backend compile specification for Ethos-U t
| 1236 | |
| 1237 | |
| 1238 | class EthosUQuantizer(TOSAQuantizer): |
| 1239 | """Quantizer supported by the Arm Ethos-U backend. |
| 1240 | |
| 1241 | .. warning:: |
| 1242 | Setting ``use_composable_quantizer=True`` enables an experimental API |
| 1243 | surface that may change without notice. |
| 1244 | |
| 1245 | Args: |
| 1246 | compile_spec (EthosUCompileSpec): Backend compile specification for |
| 1247 | Ethos-U targets. |
| 1248 | use_composable_quantizer (bool): Whether to use the composable quantizer implementation. See https://github.com/pytorch/executorch/issues/17701" for details. |
| 1249 | |
| 1250 | """ |
| 1251 | |
| 1252 | def __init__( |
| 1253 | self, |
| 1254 | compile_spec: EthosUCompileSpec, |
| 1255 | use_composable_quantizer: bool = False, |
| 1256 | ) -> None: |
| 1257 | super().__init__(compile_spec, use_composable_quantizer) |
| 1258 | |
| 1259 | |
| 1260 | class VgfQuantizer(TOSAQuantizer): |
no outgoing calls