[tracer] Add config for PyTorch2.4 (#356)

* [tracer]add PyTorch 2.4 config

* [quantizer] skip quantization for RMSNorm

* [quantizer] update quantization_support doc

docs/quantization_support.md

@@ -22,6 +22,7 @@ Quantized OPs that are natively not supported by PyTorch (and possibly TFLite).
 | `log_softmax` | / |
 | `matmul` | / |
 | `mm` | / |
+| `norm` | / |
 | `pad` | 1.7.0 |
 | `pow` | / |
 | `prelu` | / |
@@ -46,6 +47,7 @@ Quantized OPs that are natively not supported by PyTorch (and possibly TFLite).
 | `torch.nn.LayerNorm` | / |
 | `torch.nn.LogSoftmax` | / |
 | `torch.nn.PReLU` | / |
+| `torch.nn.RMSNorm` | / |
 | `torch.nn.RNN` | / |
 | `torch.nn.SiLU` | / |
 | `torch.nn.Softmax` | / |

tinynn/graph/configs/gen_funcs_yml.py

@@ -160,7 +160,7 @@ def get_scope(ns):
 ver = '_'.join(ver.split('.')[:2])
 
 # Stage 7: Functions in new versions may exist in current version
-latest = '2_0'
+latest = '2_4'
 if ver != latest:
     with open(f'torch_func_override_{latest}.yml', 'r') as f:
         d = yaml.load(f, yaml.SafeLoader)