transformers

9b322e72 - Fix torch_float casting to int instead of float (#47468)

Fix torch_float casting to int instead of float (#47468) The non-tracing branches used int(x) (copy-paste from torch_int), so the result was truncated instead of returned as a float.