CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-27 08:21:30 +00:00
Code Issues Packages Projects Releases Wiki Activity

First draft

Browse Source
This commit is contained in:
Piotr Wilkin
2025-09-18 00:21:17 +02:00
parent cd08fc3ecc
commit 344331c2b6
11 changed files with 856 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
convert_hf_to_gguf.py
View File

@@ -3741,6 +3741,28 @@ class Qwen3MoeModel(Qwen2MoeModel):
super().set_vocab()
@ModelBase.register("Qwen3NextForCausalLM")
class Qwen3NextModel(Qwen3MoeModel):
model_arch = gguf.MODEL_ARCH.QWEN3NEXT
def set_gguf_parameters(self):
super().set_gguf_parameters()
self.gguf_writer.add_ssm_conv_kernel(self.find_hparam(["linear_conv_kernel_dim"]))
self.gguf_writer.add_ssm_state_size(self.find_hparam(["linear_key_head_dim"]))
self.gguf_writer.add_ssm_group_count(self.find_hparam(["linear_num_key_heads"]))
self.gguf_writer.add_ssm_time_step_rank(self.find_hparam(["linear_num_value_heads"]))
self.gguf_writer.add_ssm_inner_size(self.find_hparam(["hidden_size"]) * (self.find_hparam(["linear_num_value_heads"]) // self.find_hparam(["linear_num_key_heads"])))
def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]:
if name.endswith(".A_log"):
data_torch = -torch.exp(data_torch)
elif name.endswith(".dt_bias"):
name = name.rpartition(".dt_bias")[0] + ".dt_proj.bias"
elif "conv1d" in name:
data_torch = data_torch.squeeze()
return Qwen2MoeModel.modify_tensors(self, data_torch, name, bid)
@ModelBase.register("GPT2LMHeadModel")
class GPT2Model(TextModel):