v0.3.14-cu124-AVX2-win-20250801

Sync llama.cpp API 20250801

v0.3.14-cu124-AVX2-linux-20250801

Sync llama.cpp API 20250801

v0.3.13-cu126-AVX2-win-20250717

fix memory_seq_rm crash bug

v0.3.13-cu126-AVX2-linux-20250717

fix memory_seq_rm crash bug

v0.3.13-cu124-AVX2-win-20250717

fix memory_seq_rm crash bug

v0.3.13-cu124-AVX2-linux-20250717

fix memory_seq_rm crash bug

v0.3.12-cu126-AVX2-win-20250714

try to use the logit_bias instead of logit_processors in test_llama

Choose the logit_bias sampler instead of the potentially unsafe logits_processor

Update Submodule vendor/llama.cpp 84b396e..55c509d

v0.3.12-cu126-AVX2-linux-20250714

try to use the logit_bias instead of logit_processors in test_llama

Choose the logit_bias sampler instead of the potentially unsafe logits_processor

Update Submodule vendor/llama.cpp 84b396e..55c509d

v0.3.12-cu124-AVX2-win-20250714

try to use the logit_bias instead of logit_processors in test_llama

Choose the logit_bias sampler instead of the potentially unsafe logits_processor

Update Submodule vendor/llama.cpp 84b396e..55c509d

v0.3.12-cu124-AVX2-linux-20250714

try to use the logit_bias instead of logit_processors in test_llama

Choose the logit_bias sampler instead of the potentially unsafe logits_processor

Update Submodule vendor/llama.cpp 84b396e..55c509d