Misc. bug: Llama-Quantize.exe broken on win11 since b5298 , but works on/earlier b5215

[David-AU-github]

Name and Version

Please note that llama-quantize.exe is failing from version b5298 (perhaps earlier) on windows 11 systems.
I have also tested: b5342 , b5361, b5371

Operating systems

Windows

Which llama.cpp modules do you know to be affected?

llama-quantize

Command line

./llama-quantize E:/main-du.gguf i:/llm/David_AU/testfiles/MN-Dark-Universe-MOE-4X12B-Reasoning-Q2_K.gguf Q2_K 8

Problem description & steps to reproduce

ISSUE:

Example:
./llama-quantize E:/main-du.gguf i:/llm/David_AU/testfiles/MN-Dark-Universe-MOE-4X12B-Reasoning-Q2_K.gguf Q2_K 8

(used in powershell)

Generates:

main: build = 5371 (e5c834f)
main: built with MSVC 19.29.30159.0 for Windows AMD64
main: quantizing 'E:/main-du.gguf' to 'i:/llm/David_AU/testfiles/MN-Dark-Universe-MOE-4X12B-Reasoning2-Q2_K.gguf' as Q2_K using 8 threads
llama_model_loader: loaded meta data with 34 key-value pairs and 403 tensors from E:/main-du.gguf (version GGUF V3 (latest))
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
llama_model_loader: - kv 0: general.architecture str = llama
llama_model_loader: - kv 1: general.type str = model
llama_model_loader: - kv 2: general.name str = MN Dark Universe MOE 4X12B Reasoning
llama_model_loader: - kv 3: general.finetune str = Reasoning
llama_model_loader: - kv 4: general.basename str = MN-Dark-Universe-MOE
llama_model_loader: - kv 5: general.size_label str = 4X12B
llama_model_loader: - kv 6: llama.block_count u32 = 40
llama_model_loader: - kv 7: llama.context_length u32 = 1024000
llama_model_loader: - kv 8: llama.embedding_length u32 = 5120
llama_model_loader: - kv 9: llama.feed_forward_length u32 = 14336
llama_model_loader: - kv 10: llama.attention.head_count u32 = 32
llama_model_loader: - kv 11: llama.attention.head_count_kv u32 = 8
llama_model_loader: - kv 12: llama.rope.freq_base f32 = 1000000.000000
llama_model_loader: - kv 13: llama.attention.layer_norm_rms_epsilon f32 = 0.000010
llama_model_loader: - kv 14: llama.expert_count u32 = 4
llama_model_loader: - kv 15: llama.expert_used_count u32 = 4
llama_model_loader: - kv 16: llama.attention.key_length u32 = 128
llama_model_loader: - kv 17: llama.attention.value_length u32 = 128
llama_model_loader: - kv 18: general.file_type u32 = 32
llama_model_loader: - kv 19: llama.vocab_size u32 = 131072
llama_model_loader: - kv 20: llama.rope.dimension_count u32 = 128
llama_model_loader: - kv 21: general.quantization_version u32 = 2
llama_model_loader: - kv 22: tokenizer.ggml.model str = gpt2
llama_model_loader: - kv 23: tokenizer.ggml.pre str = tekken
llama_model_loader: - kv 24: tokenizer.ggml.tokens arr[str,131072] = ["", "", "", "<|im_start|...
llama_model_loader: - kv 25: tokenizer.ggml.token_type arr[i32,131072] = [3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, ...
llama_model_loader: - kv 26: tokenizer.ggml.merges arr[str,269443] = ["─á ─á", "─á t", "e r", "i n", "─á ─...
llama_model_loader: - kv 27: tokenizer.ggml.bos_token_id u32 = 1
llama_model_loader: - kv 28: tokenizer.ggml.eos_token_id u32 = 4
llama_model_loader: - kv 29: tokenizer.ggml.unknown_token_id u32 = 0
llama_model_loader: - kv 30: tokenizer.ggml.padding_token_id u32 = 1
llama_model_loader: - kv 31: tokenizer.ggml.add_bos_token bool = false
llama_model_loader: - kv 32: tokenizer.ggml.add_eos_token bool = false
llama_model_loader: - kv 33: tokenizer.ggml.add_space_prefix bool = false
llama_model_loader: - type f32: 121 tensors
llama_model_loader: - type bf16: 282 tensors

-> DIES here , windows error:
An unhandled win32 exception occured in [5900] llama-quantize.exe

Additional details:
I am downloading the pre-made "exes" from releases.
I tried Cuda 12.4, 11.7, and X64 Cpu.
All same results.

Issue seems to be with loading the source files themselves.
Note that versions b5215 and earlier - no issues.

Also:
./llama-quantize --help
works - no issue.

First Bad Commit

b5298 (perhaps earlier)

Relevant log output