root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-11-14 23:09:53 +00:00
Code Issues Actions 14 Packages Projects Releases Wiki Activity

Merge a279f17815 into c02e5ab2a6

Browse Source
This commit is contained in:
pancake 2024-10-31 21:55:53 +00:00 committed by GitHub
commit 8d2b6381b8
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 8 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
convert_hf_to_gguf.py
View File

@ -357,6 +357,10 @@ class Model:
data_qtype = gguf.GGMLQuantizationType.TQ1_0 data_qtype = gguf.GGMLQuantizationType.TQ1_0
elif self.ftype == gguf.LlamaFileType.MOSTLY_TQ2_0: elif self.ftype == gguf.LlamaFileType.MOSTLY_TQ2_0:
data_qtype = gguf.GGMLQuantizationType.TQ2_0 data_qtype = gguf.GGMLQuantizationType.TQ2_0
elif self.ftype == gguf.LlamaFileType.MOSTLY_Q4_0:
data_qtype = gguf.GGMLQuantizationType.Q4_0
elif self.ftype == gguf.LlamaFileType.MOSTLY_Q4_1:
data_qtype = gguf.GGMLQuantizationType.Q4_1
else: else:
raise ValueError(f"Unknown file type: {self.ftype.name}") raise ValueError(f"Unknown file type: {self.ftype.name}")
@ -4296,8 +4300,8 @@ def parse_args() -> argparse.Namespace:
help="path to write to; default: based on input. {ftype} will be replaced by the outtype.", help="path to write to; default: based on input. {ftype} will be replaced by the outtype.",
) )
parser.add_argument( parser.add_argument(
"--outtype", type=str, choices=["f32", "f16", "bf16", "q8_0", "tq1_0", "tq2_0", "auto"], default="f16", "--outtype", type=str, choices=["f32", "f16", "bf16", "q4_0", "q4_1", "q8_0", "tq1_0", "tq2_0", "auto"], default="f16",
help="output format - use f32 for float32, f16 for float16, bf16 for bfloat16, q8_0 for Q8_0, tq1_0 or tq2_0 for ternary, and auto for the highest-fidelity 16-bit float type depending on the first loaded tensor type", help="output format - use f32 for float32, f16 for float16, bf16 for bfloat16, q4_0, q4_1 , q8_0 for Q8_0, tq1_0 or tq2_0 for ternary, and auto for the highest-fidelity 16-bit float type depending on the first loaded tensor type",
) )
parser.add_argument( parser.add_argument(
"--bigendian", action="store_true", "--bigendian", action="store_true",
@ -4383,6 +4387,8 @@ def main() -> None:
"f32": gguf.LlamaFileType.ALL_F32, "f32": gguf.LlamaFileType.ALL_F32,
"f16": gguf.LlamaFileType.MOSTLY_F16, "f16": gguf.LlamaFileType.MOSTLY_F16,
"bf16": gguf.LlamaFileType.MOSTLY_BF16, "bf16": gguf.LlamaFileType.MOSTLY_BF16,
"q4_0": gguf.LlamaFileType.MOSTLY_Q4_0,
"q4_1": gguf.LlamaFileType.MOSTLY_Q4_1,
"q8_0": gguf.LlamaFileType.MOSTLY_Q8_0, "q8_0": gguf.LlamaFileType.MOSTLY_Q8_0,
"tq1_0": gguf.LlamaFileType.MOSTLY_TQ1_0, "tq1_0": gguf.LlamaFileType.MOSTLY_TQ1_0,
"tq2_0": gguf.LlamaFileType.MOSTLY_TQ2_0, "tq2_0": gguf.LlamaFileType.MOSTLY_TQ2_0,