From 0d857442053218f89ec29d511c761e3659bbd870 Mon Sep 17 00:00:00 2001
From: "dependabot[bot]" <49699333+dependabot[bot]@users.noreply.github.com>
Date: Mon, 20 Oct 2025 20:36:55 -0300
Subject: [PATCH 1/6] Bump triton-windows in /requirements/full (#7274)

---
 requirements/full/requirements.txt                 | 2 +-
 requirements/full/requirements_amd.txt             | 2 +-
 requirements/full/requirements_amd_noavx2.txt      | 2 +-
 requirements/full/requirements_apple_intel.txt     | 2 +-
 requirements/full/requirements_apple_silicon.txt   | 2 +-
 requirements/full/requirements_cpu_only.txt        | 2 +-
 requirements/full/requirements_cpu_only_noavx2.txt | 2 +-
 requirements/full/requirements_noavx2.txt          | 2 +-
 requirements/full/requirements_nowheels.txt        | 2 +-
 9 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/requirements/full/requirements.txt b/requirements/full/requirements.txt
index 0609f69b..1cd1f86e 100644
--- a/requirements/full/requirements.txt
+++ b/requirements/full/requirements.txt
@@ -25,7 +25,7 @@ scipy
 sentencepiece
 tensorboard
 transformers==4.57.*
-triton-windows==3.3.1.post19; platform_system == "Windows"
+triton-windows==3.5.0.post21; platform_system == "Windows"
 tqdm
 wandb
 
diff --git a/requirements/full/requirements_amd.txt b/requirements/full/requirements_amd.txt
index 73d35185..7bf075b8 100644
--- a/requirements/full/requirements_amd.txt
+++ b/requirements/full/requirements_amd.txt
@@ -23,7 +23,7 @@ scipy
 sentencepiece
 tensorboard
 transformers==4.57.*
-triton-windows==3.2.0.post19; platform_system == "Windows"
+triton-windows==3.5.0.post21; platform_system == "Windows"
 tqdm
 wandb
 
diff --git a/requirements/full/requirements_amd_noavx2.txt b/requirements/full/requirements_amd_noavx2.txt
index dad354e6..b84d5993 100644
--- a/requirements/full/requirements_amd_noavx2.txt
+++ b/requirements/full/requirements_amd_noavx2.txt
@@ -23,7 +23,7 @@ scipy
 sentencepiece
 tensorboard
 transformers==4.57.*
-triton-windows==3.2.0.post19; platform_system == "Windows"
+triton-windows==3.5.0.post21; platform_system == "Windows"
 tqdm
 wandb
 
diff --git a/requirements/full/requirements_apple_intel.txt b/requirements/full/requirements_apple_intel.txt
index ab48c839..f9860021 100644
--- a/requirements/full/requirements_apple_intel.txt
+++ b/requirements/full/requirements_apple_intel.txt
@@ -23,7 +23,7 @@ scipy
 sentencepiece
 tensorboard
 transformers==4.57.*
-triton-windows==3.2.0.post19; platform_system == "Windows"
+triton-windows==3.5.0.post21; platform_system == "Windows"
 tqdm
 wandb
 
diff --git a/requirements/full/requirements_apple_silicon.txt b/requirements/full/requirements_apple_silicon.txt
index 3a341b8a..14016bcb 100644
--- a/requirements/full/requirements_apple_silicon.txt
+++ b/requirements/full/requirements_apple_silicon.txt
@@ -23,7 +23,7 @@ scipy
 sentencepiece
 tensorboard
 transformers==4.57.*
-triton-windows==3.2.0.post19; platform_system == "Windows"
+triton-windows==3.5.0.post21; platform_system == "Windows"
 tqdm
 wandb
 
diff --git a/requirements/full/requirements_cpu_only.txt b/requirements/full/requirements_cpu_only.txt
index 3765d329..5ed55a60 100644
--- a/requirements/full/requirements_cpu_only.txt
+++ b/requirements/full/requirements_cpu_only.txt
@@ -23,7 +23,7 @@ scipy
 sentencepiece
 tensorboard
 transformers==4.57.*
-triton-windows==3.2.0.post19; platform_system == "Windows"
+triton-windows==3.5.0.post21; platform_system == "Windows"
 tqdm
 wandb
 
diff --git a/requirements/full/requirements_cpu_only_noavx2.txt b/requirements/full/requirements_cpu_only_noavx2.txt
index a3e4e516..5e7ac929 100644
--- a/requirements/full/requirements_cpu_only_noavx2.txt
+++ b/requirements/full/requirements_cpu_only_noavx2.txt
@@ -23,7 +23,7 @@ scipy
 sentencepiece
 tensorboard
 transformers==4.57.*
-triton-windows==3.2.0.post19; platform_system == "Windows"
+triton-windows==3.5.0.post21; platform_system == "Windows"
 tqdm
 wandb
 
diff --git a/requirements/full/requirements_noavx2.txt b/requirements/full/requirements_noavx2.txt
index abdd262e..787e9058 100644
--- a/requirements/full/requirements_noavx2.txt
+++ b/requirements/full/requirements_noavx2.txt
@@ -25,7 +25,7 @@ scipy
 sentencepiece
 tensorboard
 transformers==4.57.*
-triton-windows==3.3.1.post19; platform_system == "Windows"
+triton-windows==3.5.0.post21; platform_system == "Windows"
 tqdm
 wandb
 
diff --git a/requirements/full/requirements_nowheels.txt b/requirements/full/requirements_nowheels.txt
index 47b0dcbd..2bb65a1d 100644
--- a/requirements/full/requirements_nowheels.txt
+++ b/requirements/full/requirements_nowheels.txt
@@ -23,7 +23,7 @@ scipy
 sentencepiece
 tensorboard
 transformers==4.57.*
-triton-windows==3.2.0.post19; platform_system == "Windows"
+triton-windows==3.5.0.post21; platform_system == "Windows"
 tqdm
 wandb
 

From 9476123ee69999249ddebe5feb57e9838d7e9f10 Mon Sep 17 00:00:00 2001
From: oobabooga <112222186+oobabooga@users.noreply.github.com>
Date: Mon, 20 Oct 2025 19:43:26 -0700
Subject: [PATCH 2/6] Update llama.cpp

---
 requirements/full/requirements.txt                     | 4 ++--
 requirements/full/requirements_amd.txt                 | 4 ++--
 requirements/full/requirements_amd_noavx2.txt          | 4 ++--
 requirements/full/requirements_apple_intel.txt         | 4 ++--
 requirements/full/requirements_apple_silicon.txt       | 6 +++---
 requirements/full/requirements_cpu_only.txt            | 4 ++--
 requirements/full/requirements_cpu_only_noavx2.txt     | 4 ++--
 requirements/full/requirements_noavx2.txt              | 4 ++--
 requirements/portable/requirements.txt                 | 4 ++--
 requirements/portable/requirements_apple_intel.txt     | 6 +++---
 requirements/portable/requirements_apple_silicon.txt   | 6 +++---
 requirements/portable/requirements_cpu_only.txt        | 4 ++--
 requirements/portable/requirements_cpu_only_noavx2.txt | 4 ++--
 requirements/portable/requirements_noavx2.txt          | 4 ++--
 requirements/portable/requirements_vulkan.txt          | 4 ++--
 requirements/portable/requirements_vulkan_noavx2.txt   | 4 ++--
 16 files changed, 35 insertions(+), 35 deletions(-)

diff --git a/requirements/full/requirements.txt b/requirements/full/requirements.txt
index 0609f69b..553f9c97 100644
--- a/requirements/full/requirements.txt
+++ b/requirements/full/requirements.txt
@@ -39,8 +39,8 @@ sse-starlette==1.6.5
 tiktoken
 
 # CUDA wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0+cu124-py3-none-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0+cu124-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cu124-py3-none-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cu124-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
 https://github.com/turboderp-org/exllamav3/releases/download/v0.0.10/exllamav3-0.0.10+cu128.torch2.7.0-cp311-cp311-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
 https://github.com/turboderp-org/exllamav3/releases/download/v0.0.10/exllamav3-0.0.10+cu128.torch2.7.0-cp311-cp311-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
 https://github.com/turboderp-org/exllamav2/releases/download/v0.3.2/exllamav2-0.3.2+cu128.torch2.7.0-cp311-cp311-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
diff --git a/requirements/full/requirements_amd.txt b/requirements/full/requirements_amd.txt
index 73d35185..85e20209 100644
--- a/requirements/full/requirements_amd.txt
+++ b/requirements/full/requirements_amd.txt
@@ -37,7 +37,7 @@ sse-starlette==1.6.5
 tiktoken
 
 # AMD wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0+vulkan-py3-none-win_amd64.whl; platform_system == "Windows"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0+vulkan-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+vulkan-py3-none-win_amd64.whl; platform_system == "Windows"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+vulkan-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
 https://github.com/turboderp-org/exllamav2/releases/download/v0.3.2/exllamav2-0.3.2+rocm6.2.4.torch2.6.0-cp311-cp311-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
 https://github.com/turboderp-org/exllamav2/releases/download/v0.3.2/exllamav2-0.3.2-py3-none-any.whl; platform_system != "Darwin" and platform_machine != "x86_64"
diff --git a/requirements/full/requirements_amd_noavx2.txt b/requirements/full/requirements_amd_noavx2.txt
index dad354e6..4922d692 100644
--- a/requirements/full/requirements_amd_noavx2.txt
+++ b/requirements/full/requirements_amd_noavx2.txt
@@ -37,7 +37,7 @@ sse-starlette==1.6.5
 tiktoken
 
 # AMD wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0+vulkanavx-py3-none-win_amd64.whl; platform_system == "Windows"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0+vulkanavx-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+vulkanavx-py3-none-win_amd64.whl; platform_system == "Windows"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+vulkanavx-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
 https://github.com/turboderp-org/exllamav2/releases/download/v0.3.2/exllamav2-0.3.2+rocm6.2.4.torch2.6.0-cp311-cp311-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
 https://github.com/turboderp-org/exllamav2/releases/download/v0.3.2/exllamav2-0.3.2-py3-none-any.whl; platform_system != "Darwin" and platform_machine != "x86_64"
diff --git a/requirements/full/requirements_apple_intel.txt b/requirements/full/requirements_apple_intel.txt
index ab48c839..b7c6ff14 100644
--- a/requirements/full/requirements_apple_intel.txt
+++ b/requirements/full/requirements_apple_intel.txt
@@ -37,5 +37,5 @@ sse-starlette==1.6.5
 tiktoken
 
 # Mac wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0-py3-none-macosx_15_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "24.0.0" and platform_release < "25.0.0" and python_version == "3.11"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0-py3-none-macosx_14_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "23.0.0" and platform_release < "24.0.0" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_15_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "24.0.0" and platform_release < "25.0.0" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_14_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "23.0.0" and platform_release < "24.0.0" and python_version == "3.11"
diff --git a/requirements/full/requirements_apple_silicon.txt b/requirements/full/requirements_apple_silicon.txt
index 3a341b8a..ccedec86 100644
--- a/requirements/full/requirements_apple_silicon.txt
+++ b/requirements/full/requirements_apple_silicon.txt
@@ -37,6 +37,6 @@ sse-starlette==1.6.5
 tiktoken
 
 # Mac wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0-py3-none-macosx_15_0_arm64.whl; platform_system == "Darwin" and platform_release >= "24.0.0" and platform_release < "25.0.0" and python_version == "3.11"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0-py3-none-macosx_14_0_arm64.whl; platform_system == "Darwin" and platform_release >= "23.0.0" and platform_release < "24.0.0" and python_version == "3.11"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0-py3-none-macosx_13_0_arm64.whl; platform_system == "Darwin" and platform_release >= "22.0.0" and platform_release < "23.0.0" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_15_0_arm64.whl; platform_system == "Darwin" and platform_release >= "24.0.0" and platform_release < "25.0.0" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_14_0_arm64.whl; platform_system == "Darwin" and platform_release >= "23.0.0" and platform_release < "24.0.0" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_13_0_arm64.whl; platform_system == "Darwin" and platform_release >= "22.0.0" and platform_release < "23.0.0" and python_version == "3.11"
diff --git a/requirements/full/requirements_cpu_only.txt b/requirements/full/requirements_cpu_only.txt
index 3765d329..da477a21 100644
--- a/requirements/full/requirements_cpu_only.txt
+++ b/requirements/full/requirements_cpu_only.txt
@@ -37,5 +37,5 @@ sse-starlette==1.6.5
 tiktoken
 
 # llama.cpp (CPU only, AVX2)
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0+cpuavx2-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0+cpuavx2-py3-none-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cpuavx2-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cpuavx2-py3-none-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
diff --git a/requirements/full/requirements_cpu_only_noavx2.txt b/requirements/full/requirements_cpu_only_noavx2.txt
index a3e4e516..0ecbf6c1 100644
--- a/requirements/full/requirements_cpu_only_noavx2.txt
+++ b/requirements/full/requirements_cpu_only_noavx2.txt
@@ -37,5 +37,5 @@ sse-starlette==1.6.5
 tiktoken
 
 # llama.cpp (CPU only, no AVX2)
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0+cpuavx-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0+cpuavx-py3-none-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cpuavx-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cpuavx-py3-none-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
diff --git a/requirements/full/requirements_noavx2.txt b/requirements/full/requirements_noavx2.txt
index abdd262e..11ef84eb 100644
--- a/requirements/full/requirements_noavx2.txt
+++ b/requirements/full/requirements_noavx2.txt
@@ -39,8 +39,8 @@ sse-starlette==1.6.5
 tiktoken
 
 # CUDA wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0+cu124avx-py3-none-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0+cu124avx-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cu124avx-py3-none-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cu124avx-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
 https://github.com/turboderp-org/exllamav3/releases/download/v0.0.10/exllamav3-0.0.10+cu128.torch2.7.0-cp311-cp311-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
 https://github.com/turboderp-org/exllamav3/releases/download/v0.0.10/exllamav3-0.0.10+cu128.torch2.7.0-cp311-cp311-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
 https://github.com/turboderp-org/exllamav2/releases/download/v0.3.2/exllamav2-0.3.2+cu128.torch2.7.0-cp311-cp311-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
diff --git a/requirements/portable/requirements.txt b/requirements/portable/requirements.txt
index d43007d0..97fa21b8 100644
--- a/requirements/portable/requirements.txt
+++ b/requirements/portable/requirements.txt
@@ -22,5 +22,5 @@ sse-starlette==1.6.5
 tiktoken
 
 # CUDA wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0+cu124-py3-none-win_amd64.whl; platform_system == "Windows"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0+cu124-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cu124-py3-none-win_amd64.whl; platform_system == "Windows"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cu124-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
diff --git a/requirements/portable/requirements_apple_intel.txt b/requirements/portable/requirements_apple_intel.txt
index 77a66dd8..318a7dfa 100644
--- a/requirements/portable/requirements_apple_intel.txt
+++ b/requirements/portable/requirements_apple_intel.txt
@@ -22,6 +22,6 @@ sse-starlette==1.6.5
 tiktoken
 
 # Mac wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0-py3-none-macosx_15_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "24.0.0" and platform_release < "25.0.0"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0-py3-none-macosx_14_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "23.0.0" and platform_release < "24.0.0"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0-py3-none-macosx_13_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "22.0.0" and platform_release < "23.0.0"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_15_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "24.0.0" and platform_release < "25.0.0"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_14_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "23.0.0" and platform_release < "24.0.0"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_13_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "22.0.0" and platform_release < "23.0.0"
diff --git a/requirements/portable/requirements_apple_silicon.txt b/requirements/portable/requirements_apple_silicon.txt
index a2479e98..d26a85c3 100644
--- a/requirements/portable/requirements_apple_silicon.txt
+++ b/requirements/portable/requirements_apple_silicon.txt
@@ -22,6 +22,6 @@ sse-starlette==1.6.5
 tiktoken
 
 # Mac wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0-py3-none-macosx_15_0_arm64.whl; platform_system == "Darwin" and platform_release >= "24.0.0" and platform_release < "25.0.0"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0-py3-none-macosx_14_0_arm64.whl; platform_system == "Darwin" and platform_release >= "23.0.0" and platform_release < "24.0.0"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0-py3-none-macosx_13_0_arm64.whl; platform_system == "Darwin" and platform_release >= "22.0.0" and platform_release < "23.0.0"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_15_0_arm64.whl; platform_system == "Darwin" and platform_release >= "24.0.0" and platform_release < "25.0.0"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_14_0_arm64.whl; platform_system == "Darwin" and platform_release >= "23.0.0" and platform_release < "24.0.0"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_13_0_arm64.whl; platform_system == "Darwin" and platform_release >= "22.0.0" and platform_release < "23.0.0"
diff --git a/requirements/portable/requirements_cpu_only.txt b/requirements/portable/requirements_cpu_only.txt
index 8d8e5233..af73d19f 100644
--- a/requirements/portable/requirements_cpu_only.txt
+++ b/requirements/portable/requirements_cpu_only.txt
@@ -22,5 +22,5 @@ sse-starlette==1.6.5
 tiktoken
 
 # llama.cpp (CPU only, AVX2)
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0+cpuavx2-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0+cpuavx2-py3-none-win_amd64.whl; platform_system == "Windows"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cpuavx2-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cpuavx2-py3-none-win_amd64.whl; platform_system == "Windows"
diff --git a/requirements/portable/requirements_cpu_only_noavx2.txt b/requirements/portable/requirements_cpu_only_noavx2.txt
index 08b7240c..8330aa34 100644
--- a/requirements/portable/requirements_cpu_only_noavx2.txt
+++ b/requirements/portable/requirements_cpu_only_noavx2.txt
@@ -22,5 +22,5 @@ sse-starlette==1.6.5
 tiktoken
 
 # llama.cpp (CPU only, no AVX2)
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0+cpuavx-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0+cpuavx-py3-none-win_amd64.whl; platform_system == "Windows"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cpuavx-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cpuavx-py3-none-win_amd64.whl; platform_system == "Windows"
diff --git a/requirements/portable/requirements_noavx2.txt b/requirements/portable/requirements_noavx2.txt
index 857c5e66..85386971 100644
--- a/requirements/portable/requirements_noavx2.txt
+++ b/requirements/portable/requirements_noavx2.txt
@@ -22,5 +22,5 @@ sse-starlette==1.6.5
 tiktoken
 
 # CUDA wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0+cu124avx-py3-none-win_amd64.whl; platform_system == "Windows"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0+cu124avx-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cu124avx-py3-none-win_amd64.whl; platform_system == "Windows"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cu124avx-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
diff --git a/requirements/portable/requirements_vulkan.txt b/requirements/portable/requirements_vulkan.txt
index d2e66763..72387a15 100644
--- a/requirements/portable/requirements_vulkan.txt
+++ b/requirements/portable/requirements_vulkan.txt
@@ -22,5 +22,5 @@ sse-starlette==1.6.5
 tiktoken
 
 # CUDA wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0+vulkan-py3-none-win_amd64.whl; platform_system == "Windows"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0+vulkan-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+vulkan-py3-none-win_amd64.whl; platform_system == "Windows"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+vulkan-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
diff --git a/requirements/portable/requirements_vulkan_noavx2.txt b/requirements/portable/requirements_vulkan_noavx2.txt
index 34927294..86d3caaf 100644
--- a/requirements/portable/requirements_vulkan_noavx2.txt
+++ b/requirements/portable/requirements_vulkan_noavx2.txt
@@ -22,5 +22,5 @@ sse-starlette==1.6.5
 tiktoken
 
 # CUDA wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0+vulkanavx-py3-none-win_amd64.whl; platform_system == "Windows"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.53.0/llama_cpp_binaries-0.53.0+vulkanavx-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+vulkanavx-py3-none-win_amd64.whl; platform_system == "Windows"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+vulkanavx-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"

From 24fd2b4dece947ad82770044fdc595dfaa20c0f0 Mon Sep 17 00:00:00 2001
From: oobabooga <112222186+oobabooga@users.noreply.github.com>
Date: Tue, 21 Oct 2025 07:26:38 -0700
Subject: [PATCH 3/6] Update exllamav3 to 0.0.11

---
 requirements/full/requirements.txt        | 4 ++--
 requirements/full/requirements_noavx2.txt | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/requirements/full/requirements.txt b/requirements/full/requirements.txt
index 95f2e18c..bd592024 100644
--- a/requirements/full/requirements.txt
+++ b/requirements/full/requirements.txt
@@ -41,8 +41,8 @@ tiktoken
 # CUDA wheels
 https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cu124-py3-none-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
 https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cu124-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
-https://github.com/turboderp-org/exllamav3/releases/download/v0.0.10/exllamav3-0.0.10+cu128.torch2.7.0-cp311-cp311-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
-https://github.com/turboderp-org/exllamav3/releases/download/v0.0.10/exllamav3-0.0.10+cu128.torch2.7.0-cp311-cp311-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
+https://github.com/turboderp-org/exllamav3/releases/download/v0.0.11/exllamav3-0.0.11+cu128.torch2.7.0-cp311-cp311-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
+https://github.com/turboderp-org/exllamav3/releases/download/v0.0.11/exllamav3-0.0.11+cu128.torch2.7.0-cp311-cp311-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
 https://github.com/turboderp-org/exllamav2/releases/download/v0.3.2/exllamav2-0.3.2+cu128.torch2.7.0-cp311-cp311-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
 https://github.com/turboderp-org/exllamav2/releases/download/v0.3.2/exllamav2-0.3.2+cu128.torch2.7.0-cp311-cp311-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
 https://github.com/turboderp-org/exllamav2/releases/download/v0.3.2/exllamav2-0.3.2-py3-none-any.whl; platform_system == "Linux" and platform_machine != "x86_64"
diff --git a/requirements/full/requirements_noavx2.txt b/requirements/full/requirements_noavx2.txt
index 6dd7383b..c0507248 100644
--- a/requirements/full/requirements_noavx2.txt
+++ b/requirements/full/requirements_noavx2.txt
@@ -41,8 +41,8 @@ tiktoken
 # CUDA wheels
 https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cu124avx-py3-none-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
 https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cu124avx-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
-https://github.com/turboderp-org/exllamav3/releases/download/v0.0.10/exllamav3-0.0.10+cu128.torch2.7.0-cp311-cp311-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
-https://github.com/turboderp-org/exllamav3/releases/download/v0.0.10/exllamav3-0.0.10+cu128.torch2.7.0-cp311-cp311-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
+https://github.com/turboderp-org/exllamav3/releases/download/v0.0.11/exllamav3-0.0.11+cu128.torch2.7.0-cp311-cp311-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
+https://github.com/turboderp-org/exllamav3/releases/download/v0.0.11/exllamav3-0.0.11+cu128.torch2.7.0-cp311-cp311-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
 https://github.com/turboderp-org/exllamav2/releases/download/v0.3.2/exllamav2-0.3.2+cu128.torch2.7.0-cp311-cp311-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
 https://github.com/turboderp-org/exllamav2/releases/download/v0.3.2/exllamav2-0.3.2+cu128.torch2.7.0-cp311-cp311-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
 https://github.com/turboderp-org/exllamav2/releases/download/v0.3.2/exllamav2-0.3.2-py3-none-any.whl; platform_system == "Linux" and platform_machine != "x86_64"

From 138cc654c4330a01bd35fdfcaa6b98d18943cb28 Mon Sep 17 00:00:00 2001
From: reksarka <reksarka@gmail.com>
Date: Wed, 22 Oct 2025 20:55:17 +0300
Subject: [PATCH 4/6] Make it possible to run a portable Web UI build via a
 symlink (#7277)

---
 start_linux.sh | 2 +-
 start_macos.sh | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/start_linux.sh b/start_linux.sh
index a991e1bc..fac8df3b 100755
--- a/start_linux.sh
+++ b/start_linux.sh
@@ -5,7 +5,7 @@ export PYTHONNOUSERSITE=1
 unset PYTHONPATH
 unset PYTHONHOME
 
-cd "$(dirname "${BASH_SOURCE[0]}")"
+cd "$(dirname "$(readlink -f "${BASH_SOURCE[0]}")")"
 
 # Portable install case
 if [ -d "portable_env" ]; then
diff --git a/start_macos.sh b/start_macos.sh
index 6426229a..6329cc0c 100755
--- a/start_macos.sh
+++ b/start_macos.sh
@@ -5,7 +5,7 @@ export PYTHONNOUSERSITE=1
 unset PYTHONPATH
 unset PYTHONHOME
 
-cd "$(dirname "${BASH_SOURCE[0]}")"
+cd "$(dirname "$(readlink -f "${BASH_SOURCE[0]}")")"
 
 # Portable install case
 if [ -d "portable_env" ]; then

From 9a84a828fc4b0e5cc632c98736f2b31adedf2106 Mon Sep 17 00:00:00 2001
From: Immanuel <drieschel@yahoo.de>
Date: Wed, 22 Oct 2025 19:59:27 +0200
Subject: [PATCH 5/6] Fixed python requirements for apple devices with macos
 tahoe (#7273)

---
 requirements/full/requirements_apple_intel.txt       | 4 ++--
 requirements/full/requirements_apple_silicon.txt     | 2 +-
 requirements/portable/requirements_apple_intel.txt   | 2 +-
 requirements/portable/requirements_apple_silicon.txt | 2 +-
 4 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/requirements/full/requirements_apple_intel.txt b/requirements/full/requirements_apple_intel.txt
index efad6b25..05ac2e46 100644
--- a/requirements/full/requirements_apple_intel.txt
+++ b/requirements/full/requirements_apple_intel.txt
@@ -37,5 +37,5 @@ sse-starlette==1.6.5
 tiktoken
 
 # Mac wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_15_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "24.0.0" and platform_release < "25.0.0" and python_version == "3.11"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_14_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "23.0.0" and platform_release < "24.0.0" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_15_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "24.0.0" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_14_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "23.0.0" and platform_release < "24.0.0" and python_version == "3.11"
\ No newline at end of file
diff --git a/requirements/full/requirements_apple_silicon.txt b/requirements/full/requirements_apple_silicon.txt
index c3faff2d..3919f0cc 100644
--- a/requirements/full/requirements_apple_silicon.txt
+++ b/requirements/full/requirements_apple_silicon.txt
@@ -37,6 +37,6 @@ sse-starlette==1.6.5
 tiktoken
 
 # Mac wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_15_0_arm64.whl; platform_system == "Darwin" and platform_release >= "24.0.0" and platform_release < "25.0.0" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_15_0_arm64.whl; platform_system == "Darwin" and platform_release >= "24.0.0" and python_version == "3.11"
 https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_14_0_arm64.whl; platform_system == "Darwin" and platform_release >= "23.0.0" and platform_release < "24.0.0" and python_version == "3.11"
 https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_13_0_arm64.whl; platform_system == "Darwin" and platform_release >= "22.0.0" and platform_release < "23.0.0" and python_version == "3.11"
diff --git a/requirements/portable/requirements_apple_intel.txt b/requirements/portable/requirements_apple_intel.txt
index 318a7dfa..7933188a 100644
--- a/requirements/portable/requirements_apple_intel.txt
+++ b/requirements/portable/requirements_apple_intel.txt
@@ -22,6 +22,6 @@ sse-starlette==1.6.5
 tiktoken
 
 # Mac wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_15_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "24.0.0" and platform_release < "25.0.0"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_15_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "24.0.0"
 https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_14_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "23.0.0" and platform_release < "24.0.0"
 https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_13_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "22.0.0" and platform_release < "23.0.0"
diff --git a/requirements/portable/requirements_apple_silicon.txt b/requirements/portable/requirements_apple_silicon.txt
index d26a85c3..538fbc14 100644
--- a/requirements/portable/requirements_apple_silicon.txt
+++ b/requirements/portable/requirements_apple_silicon.txt
@@ -22,6 +22,6 @@ sse-starlette==1.6.5
 tiktoken
 
 # Mac wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_15_0_arm64.whl; platform_system == "Darwin" and platform_release >= "24.0.0" and platform_release < "25.0.0"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_15_0_arm64.whl; platform_system == "Darwin" and platform_release >= "24.0.0"
 https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_14_0_arm64.whl; platform_system == "Darwin" and platform_release >= "23.0.0" and platform_release < "24.0.0"
 https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_13_0_arm64.whl; platform_system == "Darwin" and platform_release >= "22.0.0" and platform_release < "23.0.0"

From f4c9e671558bcdddf566f47bdad5f9e38e4018ae Mon Sep 17 00:00:00 2001
From: oobabooga <112222186+oobabooga@users.noreply.github.com>
Date: Thu, 23 Oct 2025 08:19:32 -0700
Subject: [PATCH 6/6] Update llama.cpp

---
 requirements/full/requirements.txt                     | 4 ++--
 requirements/full/requirements_amd.txt                 | 4 ++--
 requirements/full/requirements_amd_noavx2.txt          | 4 ++--
 requirements/full/requirements_apple_intel.txt         | 4 ++--
 requirements/full/requirements_apple_silicon.txt       | 6 +++---
 requirements/full/requirements_cpu_only.txt            | 4 ++--
 requirements/full/requirements_cpu_only_noavx2.txt     | 4 ++--
 requirements/full/requirements_noavx2.txt              | 4 ++--
 requirements/portable/requirements.txt                 | 4 ++--
 requirements/portable/requirements_apple_intel.txt     | 6 +++---
 requirements/portable/requirements_apple_silicon.txt   | 6 +++---
 requirements/portable/requirements_cpu_only.txt        | 4 ++--
 requirements/portable/requirements_cpu_only_noavx2.txt | 4 ++--
 requirements/portable/requirements_noavx2.txt          | 4 ++--
 requirements/portable/requirements_vulkan.txt          | 4 ++--
 requirements/portable/requirements_vulkan_noavx2.txt   | 4 ++--
 16 files changed, 35 insertions(+), 35 deletions(-)

diff --git a/requirements/full/requirements.txt b/requirements/full/requirements.txt
index bd592024..6b579bc4 100644
--- a/requirements/full/requirements.txt
+++ b/requirements/full/requirements.txt
@@ -39,8 +39,8 @@ sse-starlette==1.6.5
 tiktoken
 
 # CUDA wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cu124-py3-none-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cu124-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0+cu124-py3-none-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0+cu124-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
 https://github.com/turboderp-org/exllamav3/releases/download/v0.0.11/exllamav3-0.0.11+cu128.torch2.7.0-cp311-cp311-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
 https://github.com/turboderp-org/exllamav3/releases/download/v0.0.11/exllamav3-0.0.11+cu128.torch2.7.0-cp311-cp311-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
 https://github.com/turboderp-org/exllamav2/releases/download/v0.3.2/exllamav2-0.3.2+cu128.torch2.7.0-cp311-cp311-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
diff --git a/requirements/full/requirements_amd.txt b/requirements/full/requirements_amd.txt
index a694fd8b..277beb86 100644
--- a/requirements/full/requirements_amd.txt
+++ b/requirements/full/requirements_amd.txt
@@ -37,7 +37,7 @@ sse-starlette==1.6.5
 tiktoken
 
 # AMD wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+vulkan-py3-none-win_amd64.whl; platform_system == "Windows"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+vulkan-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0+vulkan-py3-none-win_amd64.whl; platform_system == "Windows"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0+vulkan-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
 https://github.com/turboderp-org/exllamav2/releases/download/v0.3.2/exllamav2-0.3.2+rocm6.2.4.torch2.6.0-cp311-cp311-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
 https://github.com/turboderp-org/exllamav2/releases/download/v0.3.2/exllamav2-0.3.2-py3-none-any.whl; platform_system != "Darwin" and platform_machine != "x86_64"
diff --git a/requirements/full/requirements_amd_noavx2.txt b/requirements/full/requirements_amd_noavx2.txt
index 73a54f3b..4637702e 100644
--- a/requirements/full/requirements_amd_noavx2.txt
+++ b/requirements/full/requirements_amd_noavx2.txt
@@ -37,7 +37,7 @@ sse-starlette==1.6.5
 tiktoken
 
 # AMD wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+vulkanavx-py3-none-win_amd64.whl; platform_system == "Windows"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+vulkanavx-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0+vulkanavx-py3-none-win_amd64.whl; platform_system == "Windows"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0+vulkanavx-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
 https://github.com/turboderp-org/exllamav2/releases/download/v0.3.2/exllamav2-0.3.2+rocm6.2.4.torch2.6.0-cp311-cp311-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
 https://github.com/turboderp-org/exllamav2/releases/download/v0.3.2/exllamav2-0.3.2-py3-none-any.whl; platform_system != "Darwin" and platform_machine != "x86_64"
diff --git a/requirements/full/requirements_apple_intel.txt b/requirements/full/requirements_apple_intel.txt
index 05ac2e46..7784181d 100644
--- a/requirements/full/requirements_apple_intel.txt
+++ b/requirements/full/requirements_apple_intel.txt
@@ -37,5 +37,5 @@ sse-starlette==1.6.5
 tiktoken
 
 # Mac wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_15_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "24.0.0" and python_version == "3.11"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_14_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "23.0.0" and platform_release < "24.0.0" and python_version == "3.11"
\ No newline at end of file
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0-py3-none-macosx_15_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "24.0.0" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0-py3-none-macosx_14_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "23.0.0" and platform_release < "24.0.0" and python_version == "3.11"
\ No newline at end of file
diff --git a/requirements/full/requirements_apple_silicon.txt b/requirements/full/requirements_apple_silicon.txt
index 3919f0cc..88b4effb 100644
--- a/requirements/full/requirements_apple_silicon.txt
+++ b/requirements/full/requirements_apple_silicon.txt
@@ -37,6 +37,6 @@ sse-starlette==1.6.5
 tiktoken
 
 # Mac wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_15_0_arm64.whl; platform_system == "Darwin" and platform_release >= "24.0.0" and python_version == "3.11"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_14_0_arm64.whl; platform_system == "Darwin" and platform_release >= "23.0.0" and platform_release < "24.0.0" and python_version == "3.11"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_13_0_arm64.whl; platform_system == "Darwin" and platform_release >= "22.0.0" and platform_release < "23.0.0" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0-py3-none-macosx_15_0_arm64.whl; platform_system == "Darwin" and platform_release >= "24.0.0" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0-py3-none-macosx_14_0_arm64.whl; platform_system == "Darwin" and platform_release >= "23.0.0" and platform_release < "24.0.0" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0-py3-none-macosx_13_0_arm64.whl; platform_system == "Darwin" and platform_release >= "22.0.0" and platform_release < "23.0.0" and python_version == "3.11"
diff --git a/requirements/full/requirements_cpu_only.txt b/requirements/full/requirements_cpu_only.txt
index 0b0d0dca..aca3e51f 100644
--- a/requirements/full/requirements_cpu_only.txt
+++ b/requirements/full/requirements_cpu_only.txt
@@ -37,5 +37,5 @@ sse-starlette==1.6.5
 tiktoken
 
 # llama.cpp (CPU only, AVX2)
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cpuavx2-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cpuavx2-py3-none-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0+cpuavx2-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0+cpuavx2-py3-none-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
diff --git a/requirements/full/requirements_cpu_only_noavx2.txt b/requirements/full/requirements_cpu_only_noavx2.txt
index f46ecda6..e75df348 100644
--- a/requirements/full/requirements_cpu_only_noavx2.txt
+++ b/requirements/full/requirements_cpu_only_noavx2.txt
@@ -37,5 +37,5 @@ sse-starlette==1.6.5
 tiktoken
 
 # llama.cpp (CPU only, no AVX2)
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cpuavx-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cpuavx-py3-none-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0+cpuavx-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0+cpuavx-py3-none-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
diff --git a/requirements/full/requirements_noavx2.txt b/requirements/full/requirements_noavx2.txt
index c0507248..16dfea5a 100644
--- a/requirements/full/requirements_noavx2.txt
+++ b/requirements/full/requirements_noavx2.txt
@@ -39,8 +39,8 @@ sse-starlette==1.6.5
 tiktoken
 
 # CUDA wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cu124avx-py3-none-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cu124avx-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0+cu124avx-py3-none-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0+cu124avx-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
 https://github.com/turboderp-org/exllamav3/releases/download/v0.0.11/exllamav3-0.0.11+cu128.torch2.7.0-cp311-cp311-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
 https://github.com/turboderp-org/exllamav3/releases/download/v0.0.11/exllamav3-0.0.11+cu128.torch2.7.0-cp311-cp311-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
 https://github.com/turboderp-org/exllamav2/releases/download/v0.3.2/exllamav2-0.3.2+cu128.torch2.7.0-cp311-cp311-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
diff --git a/requirements/portable/requirements.txt b/requirements/portable/requirements.txt
index 97fa21b8..c5885c3e 100644
--- a/requirements/portable/requirements.txt
+++ b/requirements/portable/requirements.txt
@@ -22,5 +22,5 @@ sse-starlette==1.6.5
 tiktoken
 
 # CUDA wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cu124-py3-none-win_amd64.whl; platform_system == "Windows"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cu124-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0+cu124-py3-none-win_amd64.whl; platform_system == "Windows"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0+cu124-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
diff --git a/requirements/portable/requirements_apple_intel.txt b/requirements/portable/requirements_apple_intel.txt
index 7933188a..a50a3e84 100644
--- a/requirements/portable/requirements_apple_intel.txt
+++ b/requirements/portable/requirements_apple_intel.txt
@@ -22,6 +22,6 @@ sse-starlette==1.6.5
 tiktoken
 
 # Mac wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_15_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "24.0.0"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_14_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "23.0.0" and platform_release < "24.0.0"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_13_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "22.0.0" and platform_release < "23.0.0"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0-py3-none-macosx_15_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "24.0.0"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0-py3-none-macosx_14_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "23.0.0" and platform_release < "24.0.0"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0-py3-none-macosx_13_0_x86_64.whl; platform_system == "Darwin" and platform_release >= "22.0.0" and platform_release < "23.0.0"
diff --git a/requirements/portable/requirements_apple_silicon.txt b/requirements/portable/requirements_apple_silicon.txt
index 538fbc14..df0beb22 100644
--- a/requirements/portable/requirements_apple_silicon.txt
+++ b/requirements/portable/requirements_apple_silicon.txt
@@ -22,6 +22,6 @@ sse-starlette==1.6.5
 tiktoken
 
 # Mac wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_15_0_arm64.whl; platform_system == "Darwin" and platform_release >= "24.0.0"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_14_0_arm64.whl; platform_system == "Darwin" and platform_release >= "23.0.0" and platform_release < "24.0.0"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0-py3-none-macosx_13_0_arm64.whl; platform_system == "Darwin" and platform_release >= "22.0.0" and platform_release < "23.0.0"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0-py3-none-macosx_15_0_arm64.whl; platform_system == "Darwin" and platform_release >= "24.0.0"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0-py3-none-macosx_14_0_arm64.whl; platform_system == "Darwin" and platform_release >= "23.0.0" and platform_release < "24.0.0"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0-py3-none-macosx_13_0_arm64.whl; platform_system == "Darwin" and platform_release >= "22.0.0" and platform_release < "23.0.0"
diff --git a/requirements/portable/requirements_cpu_only.txt b/requirements/portable/requirements_cpu_only.txt
index af73d19f..75e60b8c 100644
--- a/requirements/portable/requirements_cpu_only.txt
+++ b/requirements/portable/requirements_cpu_only.txt
@@ -22,5 +22,5 @@ sse-starlette==1.6.5
 tiktoken
 
 # llama.cpp (CPU only, AVX2)
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cpuavx2-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cpuavx2-py3-none-win_amd64.whl; platform_system == "Windows"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0+cpuavx2-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0+cpuavx2-py3-none-win_amd64.whl; platform_system == "Windows"
diff --git a/requirements/portable/requirements_cpu_only_noavx2.txt b/requirements/portable/requirements_cpu_only_noavx2.txt
index 8330aa34..b9835fa1 100644
--- a/requirements/portable/requirements_cpu_only_noavx2.txt
+++ b/requirements/portable/requirements_cpu_only_noavx2.txt
@@ -22,5 +22,5 @@ sse-starlette==1.6.5
 tiktoken
 
 # llama.cpp (CPU only, no AVX2)
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cpuavx-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cpuavx-py3-none-win_amd64.whl; platform_system == "Windows"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0+cpuavx-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0+cpuavx-py3-none-win_amd64.whl; platform_system == "Windows"
diff --git a/requirements/portable/requirements_noavx2.txt b/requirements/portable/requirements_noavx2.txt
index 85386971..7cf4755d 100644
--- a/requirements/portable/requirements_noavx2.txt
+++ b/requirements/portable/requirements_noavx2.txt
@@ -22,5 +22,5 @@ sse-starlette==1.6.5
 tiktoken
 
 # CUDA wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cu124avx-py3-none-win_amd64.whl; platform_system == "Windows"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+cu124avx-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0+cu124avx-py3-none-win_amd64.whl; platform_system == "Windows"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0+cu124avx-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
diff --git a/requirements/portable/requirements_vulkan.txt b/requirements/portable/requirements_vulkan.txt
index 72387a15..035c8d4e 100644
--- a/requirements/portable/requirements_vulkan.txt
+++ b/requirements/portable/requirements_vulkan.txt
@@ -22,5 +22,5 @@ sse-starlette==1.6.5
 tiktoken
 
 # CUDA wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+vulkan-py3-none-win_amd64.whl; platform_system == "Windows"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+vulkan-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0+vulkan-py3-none-win_amd64.whl; platform_system == "Windows"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0+vulkan-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
diff --git a/requirements/portable/requirements_vulkan_noavx2.txt b/requirements/portable/requirements_vulkan_noavx2.txt
index 86d3caaf..bb3bd518 100644
--- a/requirements/portable/requirements_vulkan_noavx2.txt
+++ b/requirements/portable/requirements_vulkan_noavx2.txt
@@ -22,5 +22,5 @@ sse-starlette==1.6.5
 tiktoken
 
 # CUDA wheels
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+vulkanavx-py3-none-win_amd64.whl; platform_system == "Windows"
-https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.54.0/llama_cpp_binaries-0.54.0+vulkanavx-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0+vulkanavx-py3-none-win_amd64.whl; platform_system == "Windows"
+https://github.com/oobabooga/llama-cpp-binaries/releases/download/v0.56.0/llama_cpp_binaries-0.56.0+vulkanavx-py3-none-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64"