happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

fixed flake8 lint issues

This commit is contained in:
Saba Fallah 2025-12-05 12:24:10 +01:00
parent 5f2ee1aecf
commit 1c88647ec6
3 changed files with 15 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
convert_hf_to_gguf.py
View File

@ -6005,6 +6005,7 @@ class Gemma3VisionModel(MmprojModel):
return [] # skip other tensors return [] # skip other tensors
@ModelBase.register("DeepseekOCRForCausalLM") @ModelBase.register("DeepseekOCRForCausalLM")
class DeepseekOCRVisionModel(MmprojModel): class DeepseekOCRVisionModel(MmprojModel):
def set_gguf_parameters(self): def set_gguf_parameters(self):
@ -6044,7 +6045,6 @@ class DeepseekOCRVisionModel(MmprojModel):
return vision_config return vision_config
def tensor_force_quant(self, name, new_name, bid, n_dims): def tensor_force_quant(self, name, new_name, bid, n_dims):
# TODO: increase numercial stability. maybe delete later. # TODO: increase numercial stability. maybe delete later.
return gguf.GGMLQuantizationType.F32 return gguf.GGMLQuantizationType.F32
@ -7244,7 +7244,7 @@ class DeepseekV2Model(TextModel):
def __init__(self, *args, **kwargs): def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs) super().__init__(*args, **kwargs)
vision_config = self.hparams.get('vision_config', {}).get('width', {}) vision_config = self.hparams.get('vision_config', {}).get('width', {})
if 'clip-l-14-224' in vision_config and 'sam_vit_b' in vision_config: if 'clip-l-14-224' in vision_config and 'sam_vit_b' in vision_config:
self.model_arch = gguf.MODEL_ARCH.DEEPSEEK2OCR self.model_arch = gguf.MODEL_ARCH.DEEPSEEK2OCR
self.gguf_writer.arch = gguf.MODEL_ARCH_NAMES[self.model_arch] self.gguf_writer.arch = gguf.MODEL_ARCH_NAMES[self.model_arch]
@ -7354,8 +7354,12 @@ class DeepseekV2Model(TextModel):
def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]: def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]:
# skip vision tensors and remove "language_model." for Kimi-VL # skip vision tensors and remove "language_model." for Kimi-VL
if "vision_" in name or "multi_modal_projector" in name \ if ("vision_" in name
or "image_newline" in name or "model.projector" in name or "sam_model" in name or "view_seperator" in name: or "multi_modal_projector" in name
or "image_newline" in name
or "model.projector" in name
or "sam_model" in name
or "view_seperator" in name):
return [] return []
if name.startswith("language_model."): if name.startswith("language_model."):
@ -7435,6 +7439,7 @@ class DeepseekV2Model(TextModel):
if len(experts) > 0: if len(experts) > 0:
raise ValueError(f"Unprocessed experts: {experts}") raise ValueError(f"Unprocessed experts: {experts}")
@ModelBase.register("MiniMaxM2ForCausalLM") @ModelBase.register("MiniMaxM2ForCausalLM")
class MiniMaxM2Model(TextModel): class MiniMaxM2Model(TextModel):
model_arch = gguf.MODEL_ARCH.MINIMAXM2 model_arch = gguf.MODEL_ARCH.MINIMAXM2

2
gguf-py/gguf/gguf_writer.py
View File

@ -1129,7 +1129,7 @@ class GGUFWriter:
def add_vision_sam_layers_count(self, value: int) -> None: def add_vision_sam_layers_count(self, value: int) -> None:
self.add_uint32(Keys.ClipVision.SAM.BLOCK_COUNT, value) self.add_uint32(Keys.ClipVision.SAM.BLOCK_COUNT, value)
def add_vision_sam_embedding_length(self, value: int) -> None: def add_vision_sam_embedding_length(self, value: int) -> None:
self.add_uint32(Keys.ClipVision.SAM.EMBEDDING_LENGTH, value) self.add_uint32(Keys.ClipVision.SAM.EMBEDDING_LENGTH, value)

10
gguf-py/gguf/tensor_mapping.py
View File

@ -1238,15 +1238,15 @@ class TensorNameMap:
"vision_model.positional_embedding_vlm", # llama 4 "vision_model.positional_embedding_vlm", # llama 4
"vision_tower.patch_embed.pos_emb", # kimi-vl "vision_tower.patch_embed.pos_emb", # kimi-vl
"visual.pos_embed", # qwen3vl "visual.pos_embed", # qwen3vl
"model.vision.patch_embedding.position_embedding", # cogvlm "model.vision.patch_embedding.position_embedding", # cogvlm
), ),
MODEL_TENSOR.V_ENC_EMBD_IMGNL: ( MODEL_TENSOR.V_ENC_EMBD_IMGNL: (
"model.image_newline", # Deepseek-OCR "model.image_newline", # Deepseek-OCR
), ),
MODEL_TENSOR.V_ENC_EMBD_VSEP: ( MODEL_TENSOR.V_ENC_EMBD_VSEP: (
"model.view_seperator", # Deepseek-OCR "model.view_seperator", # Deepseek-OCR
), ),
MODEL_TENSOR.V_ENC_ATTN_QKV: ( MODEL_TENSOR.V_ENC_ATTN_QKV: (