vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

convert-hf: add missing ftype

Browse source
This commit is contained in:
Francis Couture-Harpin 2024-05-11 23:18:30 -04:00
parent d7e199e444
commit 2b1e5ea37b
2 changed files with 10 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

4
convert-hf-to-gguf.py
View file

@ -1202,6 +1202,7 @@ class StableLMModel(Model):
self.gguf_writer.add_head_count_kv(hparams["num_key_value_heads"])
self.gguf_writer.add_parallel_residual(hparams["use_parallel_residual"] if "use_parallel_residual" in hparams else True)
self.gguf_writer.add_layer_norm_eps(self.find_hparam(["layer_norm_eps", "norm_eps"]))
self.gguf_writer.add_file_type(self.ftype)
_q_norms: list[dict[str, Tensor]] | None = None
_k_norms: list[dict[str, Tensor]] | None = None
@ -1578,6 +1579,7 @@ class QwenModel(Model):
self.gguf_writer.add_rope_dimension_count(self.hparams["hidden_size"] // self.hparams["num_attention_heads"])
self.gguf_writer.add_head_count(self.hparams["num_attention_heads"])
self.gguf_writer.add_layer_norm_rms_eps(self.hparams["layer_norm_epsilon"])
self.gguf_writer.add_file_type(self.ftype)
@Model.register("Qwen2ForCausalLM")
@ -1815,6 +1817,7 @@ class PlamoModel(Model):
self.gguf_writer.add_head_count(hparams["num_attention_heads"])
self.gguf_writer.add_head_count_kv(5) # hparams["num_key_value_heads"]) is wrong
self.gguf_writer.add_layer_norm_rms_eps(hparams["rms_norm_eps"])
self.gguf_writer.add_file_type(self.ftype)
def shuffle_attn_q_weight(self, data_torch):
assert data_torch.size() == (5120, 5120)
@ -1994,6 +1997,7 @@ in chat mode so that the conversation can end normally.")
self.gguf_writer.add_head_count(self.hparams["num_attention_heads"])
self.gguf_writer.add_layer_norm_rms_eps(self.hparams["rms_norm_eps"])
self.gguf_writer.add_head_count_kv(self.hparams["num_key_value_heads"])
self.gguf_writer.add_file_type(self.ftype)
def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]:
num_heads = self.hparams["num_attention_heads"]