CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-31 08:51:55 +00:00
Code Issues Packages Projects Releases Wiki Activity

gguf-py : allow previewing reflinked size on non-Linux platforms

Browse Source
This commit is contained in:
Francis Couture-Harpin
2025-09-05 10:45:41 -04:00
parent 8ef4136b20
commit 2499e47cfd
3 changed files with 2 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
convert_hf_to_gguf.py
View File

@@ -738,7 +738,6 @@ class TextModel(ModelBase):
super().prepare_metadata(vocab_only=vocab_only) super().prepare_metadata(vocab_only=vocab_only)
total_params = self.gguf_writer.get_total_parameter_count()[0] total_params = self.gguf_writer.get_total_parameter_count()[0]
# Extract the encoding scheme from the file type name. e.g. 'gguf.LlamaFileType.MOSTLY_Q8_0' --> 'Q8_0' # Extract the encoding scheme from the file type name. e.g. 'gguf.LlamaFileType.MOSTLY_Q8_0' --> 'Q8_0'
output_type: str = self.ftype.name.partition("_")[2] output_type: str = self.ftype.name.partition("_")[2]

2
ggml/src/gguf.cpp
View File

@@ -624,7 +624,7 @@ struct gguf_context * gguf_init_from_file_impl(FILE * file, struct gguf_init_par
ctx->size = 0; ctx->size = 0;
for (size_t i = 0; i < ctx->info.size(); ++i) { for (size_t i = 0; i < ctx->info.size(); ++i) {
const gguf_tensor_info & ti = ctx->info[i]; const gguf_tensor_info & ti = ctx->info[i];
// alignment offset is only necessary for GGUF converted with reflinks // alignment offset only exists for GGUF converted with reflinks
const size_t align_offset = ti.offset % ctx->alignment; const size_t align_offset = ti.offset % ctx->alignment;
if (ti.offset - align_offset != ctx->size) { if (ti.offset - align_offset != ctx->size) {
GGML_LOG_ERROR("%s: tensor '%s' has offset %" PRIu64 ", expected %zu\n", GGML_LOG_ERROR("%s: tensor '%s' has offset %" PRIu64 ", expected %zu\n",

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

@@ -93,7 +93,7 @@ class GGUFWriter:
self.arch = arch self.arch = arch
self.endianess = endianess self.endianess = endianess
self.data_alignment = GGUF_DEFAULT_ALIGNMENT self.data_alignment = GGUF_DEFAULT_ALIGNMENT
self.use_reflinks = use_reflinks and hasattr(os, "copy_file_range") self.use_reflinks = use_reflinks
self.use_temp_file = False if self.use_reflinks else use_temp_file self.use_temp_file = False if self.use_reflinks else use_temp_file
self.temp_file = None self.temp_file = None
self.tensors = [{}] self.tensors = [{}]