convert-hf : use zero value of sliding_window to distinguish Phi-4 from other PHI3 models

This commit is contained in:
Stanisław Szymczyk 2024-12-14 11:59:59 +01:00
parent 520e8a0377
commit c7fdbd3735

View file

@ -2254,8 +2254,10 @@ class Phi3MiniModel(Model):
self.gguf_writer.add_rope_dimension_count(rope_dims) self.gguf_writer.add_rope_dimension_count(rope_dims)
self.gguf_writer.add_rope_freq_base(self.find_hparam(["rope_theta"])) self.gguf_writer.add_rope_freq_base(self.find_hparam(["rope_theta"]))
self.gguf_writer.add_file_type(self.ftype) self.gguf_writer.add_file_type(self.ftype)
# handle null value of sliding_window (Phi-4 model) sliding_window = self.hparams.get("sliding_window")
if (sliding_window := self.hparams.get("sliding_window")) is not None: # use zero value of sliding_window to distinguish Phi-4 from other PHI3 models
if sliding_window is None:
sliding_window = 0
self.gguf_writer.add_sliding_window(sliding_window) self.gguf_writer.add_sliding_window(sliding_window)
def generate_extra_tensors(self) -> Iterable[tuple[str, Tensor]]: def generate_extra_tensors(self) -> Iterable[tuple[str, Tensor]]: