convert : allow quantizing lora again (#17453)

diff --git a/convert_hf_to_gguf.py b/convert_hf_to_gguf.py
index 6cbaee03dfdb1f82d187219cc4e5a39847caf653..d24a4682f3d612dc107713aeb738d1853775083a 100755 (executable)
--- a/convert_hf_to_gguf.py
+++ b/convert_hf_to_gguf.py
@@ -565,7 +565,7 @@ class ModelBase:
                             gguf.MODEL_TENSOR.ALTUP_PREDICT_COEF,
                         )
                     )
-                    or not new_name.endswith(".weight")
+                    or new_name[-7:] not in (".weight", ".lora_a", ".lora_b")
                 ):
                     data_qtype = gguf.GGMLQuantizationType.F32
 

diff --git a/convert_lora_to_gguf.py b/convert_lora_to_gguf.py
index 57c6cd0df1d6d989b2edf840284387730ba39132..b0adde8a8b4fa5840c85d7eaab0af6de99f37175 100755 (executable)
--- a/convert_lora_to_gguf.py
+++ b/convert_lora_to_gguf.py
@@ -242,7 +242,7 @@ def parse_args() -> argparse.Namespace:
         help="path to write to; default: based on input. {ftype} will be replaced by the outtype.",
     )
     parser.add_argument(
-        "--outtype", type=str, choices=["f32", "f16", "bf16", "q8_0", "auto"], default="f16",
+        "--outtype", type=str, choices=["f32", "f16", "bf16", "q8_0", "auto"], default="f32",
         help="output format - use f32 for float32, f16 for float16, bf16 for bfloat16, q8_0 for Q8_0, auto for the highest-fidelity 16-bit float type depending on the first loaded tensor type",
     )
     parser.add_argument(