CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-30 08:42:00 +00:00
Code Issues Packages Projects Releases Wiki Activity

convert-hf : set the model name based on cli arg, if present (#7693)

Browse Source 
`--model-name` argument was added a while ago but did not do anything.
This commit fixes this issue and enables this feature.
This commit is contained in:
sasha0552
2024-06-09 06:39:25 +00:00
committed by GitHub
parent 5795b94182
commit 2decf57bc6
1 changed files with 14 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

26
convert-hf-to-gguf.py
View File

@@ -52,6 +52,7 @@ class Model:
endianess: gguf.GGUFEndian endianess: gguf.GGUFEndian
use_temp_file: bool use_temp_file: bool
lazy: bool lazy: bool
model_name: str | None
part_names: list[str] part_names: list[str]
is_safetensors: bool is_safetensors: bool
hparams: dict[str, Any] hparams: dict[str, Any]
@@ -64,7 +65,7 @@ class Model:
# subclasses should define this! # subclasses should define this!
model_arch: gguf.MODEL_ARCH model_arch: gguf.MODEL_ARCH
def __init__(self, dir_model: Path, ftype: gguf.LlamaFileType, fname_out: Path, is_big_endian: bool, use_temp_file: bool, eager: bool): def __init__(self, dir_model: Path, ftype: gguf.LlamaFileType, fname_out: Path, is_big_endian: bool, use_temp_file: bool, eager: bool, model_name: str | None):
if type(self) is Model: if type(self) is Model:
raise TypeError(f"{type(self).__name__!r} should not be directly instantiated") raise TypeError(f"{type(self).__name__!r} should not be directly instantiated")
self.dir_model = dir_model self.dir_model = dir_model
@@ -73,6 +74,7 @@ class Model:
self.endianess = gguf.GGUFEndian.BIG if is_big_endian else gguf.GGUFEndian.LITTLE self.endianess = gguf.GGUFEndian.BIG if is_big_endian else gguf.GGUFEndian.LITTLE
self.use_temp_file = use_temp_file self.use_temp_file = use_temp_file
self.lazy = not eager self.lazy = not eager
self.model_name = model_name
self.part_names = Model.get_model_part_names(self.dir_model, "model", ".safetensors") self.part_names = Model.get_model_part_names(self.dir_model, "model", ".safetensors")
self.is_safetensors = len(self.part_names) > 0 self.is_safetensors = len(self.part_names) > 0
if not self.is_safetensors: if not self.is_safetensors:
@@ -182,7 +184,7 @@ class Model:
return new_name return new_name
def set_gguf_parameters(self): def set_gguf_parameters(self):
self.gguf_writer.add_name(self.dir_model.name) self.gguf_writer.add_name(self.dir_model.name if self.model_name is None else self.model_name)
self.gguf_writer.add_block_count(self.block_count) self.gguf_writer.add_block_count(self.block_count)
if (n_ctx := self.find_hparam(["max_position_embeddings", "n_ctx"], optional=True)) is not None: if (n_ctx := self.find_hparam(["max_position_embeddings", "n_ctx"], optional=True)) is not None:
@@ -665,7 +667,7 @@ class GPTNeoXModel(Model):
def set_gguf_parameters(self): def set_gguf_parameters(self):
block_count = self.hparams["num_hidden_layers"] block_count = self.hparams["num_hidden_layers"]
self.gguf_writer.add_name(self.dir_model.name) self.gguf_writer.add_name(self.dir_model.name if self.model_name is None else self.model_name)
self.gguf_writer.add_context_length(self.hparams["max_position_embeddings"]) self.gguf_writer.add_context_length(self.hparams["max_position_embeddings"])
self.gguf_writer.add_embedding_length(self.hparams["hidden_size"]) self.gguf_writer.add_embedding_length(self.hparams["hidden_size"])
self.gguf_writer.add_block_count(block_count) self.gguf_writer.add_block_count(block_count)
@@ -798,7 +800,7 @@ class MPTModel(Model):
def set_gguf_parameters(self): def set_gguf_parameters(self):
block_count = self.hparams["n_layers"] block_count = self.hparams["n_layers"]
self.gguf_writer.add_name(self.dir_model.name) self.gguf_writer.add_name(self.dir_model.name if self.model_name is None else self.model_name)
self.gguf_writer.add_context_length(self.hparams["max_seq_len"]) self.gguf_writer.add_context_length(self.hparams["max_seq_len"])
self.gguf_writer.add_embedding_length(self.hparams["d_model"]) self.gguf_writer.add_embedding_length(self.hparams["d_model"])
self.gguf_writer.add_block_count(block_count) self.gguf_writer.add_block_count(block_count)
@@ -850,7 +852,7 @@ class OrionModel(Model):
raise ValueError("gguf: can not find ctx length parameter.") raise ValueError("gguf: can not find ctx length parameter.")
self.gguf_writer.add_file_type(self.ftype) self.gguf_writer.add_file_type(self.ftype)
self.gguf_writer.add_name(self.dir_model.name) self.gguf_writer.add_name(self.dir_model.name if self.model_name is None else self.model_name)
self.gguf_writer.add_source_hf_repo(hf_repo) self.gguf_writer.add_source_hf_repo(hf_repo)
self.gguf_writer.add_tensor_data_layout("Meta AI original pth") self.gguf_writer.add_tensor_data_layout("Meta AI original pth")
self.gguf_writer.add_context_length(ctx_length) self.gguf_writer.add_context_length(ctx_length)
@@ -887,7 +889,7 @@ class BaichuanModel(Model):
else: else:
raise ValueError("gguf: can not find ctx length parameter.") raise ValueError("gguf: can not find ctx length parameter.")
self.gguf_writer.add_name(self.dir_model.name) self.gguf_writer.add_name(self.dir_model.name if self.model_name is None else self.model_name)
self.gguf_writer.add_source_hf_repo(hf_repo) self.gguf_writer.add_source_hf_repo(hf_repo)
self.gguf_writer.add_tensor_data_layout("Meta AI original pth") self.gguf_writer.add_tensor_data_layout("Meta AI original pth")
self.gguf_writer.add_context_length(ctx_length) self.gguf_writer.add_context_length(ctx_length)
@@ -1010,7 +1012,7 @@ class XverseModel(Model):
else: else:
raise ValueError("gguf: can not find ctx length parameter.") raise ValueError("gguf: can not find ctx length parameter.")
self.gguf_writer.add_name(self.dir_model.name) self.gguf_writer.add_name(self.dir_model.name if self.model_name is None else self.model_name)
self.gguf_writer.add_source_hf_repo(hf_repo) self.gguf_writer.add_source_hf_repo(hf_repo)
self.gguf_writer.add_tensor_data_layout("Meta AI original pth") self.gguf_writer.add_tensor_data_layout("Meta AI original pth")
self.gguf_writer.add_context_length(ctx_length) self.gguf_writer.add_context_length(ctx_length)
@@ -1206,7 +1208,7 @@ class StableLMModel(Model):
hparams = self.hparams hparams = self.hparams
block_count = hparams["num_hidden_layers"] block_count = hparams["num_hidden_layers"]
self.gguf_writer.add_name(self.dir_model.name) self.gguf_writer.add_name(self.dir_model.name if self.model_name is None else self.model_name)
self.gguf_writer.add_context_length(hparams["max_position_embeddings"]) self.gguf_writer.add_context_length(hparams["max_position_embeddings"])
self.gguf_writer.add_embedding_length(hparams["hidden_size"]) self.gguf_writer.add_embedding_length(hparams["hidden_size"])
self.gguf_writer.add_block_count(block_count) self.gguf_writer.add_block_count(block_count)
@@ -1681,7 +1683,7 @@ class GPT2Model(Model):
model_arch = gguf.MODEL_ARCH.GPT2 model_arch = gguf.MODEL_ARCH.GPT2
def set_gguf_parameters(self): def set_gguf_parameters(self):
self.gguf_writer.add_name(self.dir_model.name) self.gguf_writer.add_name(self.dir_model.name if self.model_name is None else self.model_name)
self.gguf_writer.add_block_count(self.hparams["n_layer"]) self.gguf_writer.add_block_count(self.hparams["n_layer"])
self.gguf_writer.add_context_length(self.hparams["n_ctx"]) self.gguf_writer.add_context_length(self.hparams["n_ctx"])
self.gguf_writer.add_embedding_length(self.hparams["n_embd"]) self.gguf_writer.add_embedding_length(self.hparams["n_embd"])
@@ -2248,7 +2250,7 @@ class GemmaModel(Model):
hparams = self.hparams hparams = self.hparams
block_count = hparams["num_hidden_layers"] block_count = hparams["num_hidden_layers"]
self.gguf_writer.add_name(self.dir_model.name) self.gguf_writer.add_name(self.dir_model.name if self.model_name is None else self.model_name)
self.gguf_writer.add_context_length(hparams["max_position_embeddings"]) self.gguf_writer.add_context_length(hparams["max_position_embeddings"])
self.gguf_writer.add_embedding_length(hparams["hidden_size"]) self.gguf_writer.add_embedding_length(hparams["hidden_size"])
self.gguf_writer.add_block_count(block_count) self.gguf_writer.add_block_count(block_count)
@@ -2348,7 +2350,7 @@ class MambaModel(Model):
# Fail early for models which don't have a block expansion factor of 2 # Fail early for models which don't have a block expansion factor of 2
assert d_inner == 2 * d_model assert d_inner == 2 * d_model
self.gguf_writer.add_name(self.dir_model.name) self.gguf_writer.add_name(self.dir_model.name if self.model_name is None else self.model_name)
self.gguf_writer.add_context_length(2**20) # arbitrary value; for those who use the default self.gguf_writer.add_context_length(2**20) # arbitrary value; for those who use the default
self.gguf_writer.add_embedding_length(d_model) self.gguf_writer.add_embedding_length(d_model)
self.gguf_writer.add_feed_forward_length(0) # unused, but seemingly required when loading self.gguf_writer.add_feed_forward_length(0) # unused, but seemingly required when loading
@@ -2852,7 +2854,7 @@ def main() -> None:
logger.error(f"Model {hparams['architectures'][0]} is not supported") logger.error(f"Model {hparams['architectures'][0]} is not supported")
sys.exit(1) sys.exit(1)
model_instance = model_class(dir_model, ftype_map[args.outtype], fname_out, args.bigendian, args.use_temp_file, args.no_lazy) model_instance = model_class(dir_model, ftype_map[args.outtype], fname_out, args.bigendian, args.use_temp_file, args.no_lazy, args.model_name)
logger.info("Set model parameters") logger.info("Set model parameters")
model_instance.set_gguf_parameters() model_instance.set_gguf_parameters()