diff --git a/Cargo.lock b/Cargo.lock
index 5b49a52..904194a 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -852,13 +852,13 @@ dependencies = [
 [[package]]
 name = "candle-core"
 version = "0.9.1"
-source = "git+https://github.com/huggingface/candle.git?branch=main#be4f920f35b99af5919a86c0a2e08a78f3f63567"
+source = "git+https://github.com/huggingface/candle.git?branch=main#9fe623237a515f95f70d117c0a6da610c28a5ecd"
 dependencies = [
  "byteorder",
  "candle-kernels",
  "cudarc",
- "float8",
- "gemm 0.17.1",
+ "float8 0.4.2",
+ "gemm",
  "half",
  "memmap2 0.9.7",
  "num-traits",
@@ -877,7 +877,7 @@ dependencies = [
 [[package]]
 name = "candle-flash-attn"
 version = "0.9.1"
-source = "git+https://github.com/huggingface/candle.git?branch=main#be4f920f35b99af5919a86c0a2e08a78f3f63567"
+source = "git+https://github.com/huggingface/candle.git?branch=main#9fe623237a515f95f70d117c0a6da610c28a5ecd"
 dependencies = [
  "anyhow",
  "bindgen_cuda",
@@ -888,7 +888,7 @@ dependencies = [
 [[package]]
 name = "candle-kernels"
 version = "0.9.1"
-source = "git+https://github.com/huggingface/candle.git?branch=main#be4f920f35b99af5919a86c0a2e08a78f3f63567"
+source = "git+https://github.com/huggingface/candle.git?branch=main#9fe623237a515f95f70d117c0a6da610c28a5ecd"
 dependencies = [
  "bindgen_cuda",
 ]
@@ -896,10 +896,11 @@ dependencies = [
 [[package]]
 name = "candle-nn"
 version = "0.9.1"
-source = "git+https://github.com/huggingface/candle.git?branch=main#be4f920f35b99af5919a86c0a2e08a78f3f63567"
+source = "git+https://github.com/huggingface/candle.git?branch=main#9fe623237a515f95f70d117c0a6da610c28a5ecd"
 dependencies = [
  "candle-core",
  "half",
+ "libc",
  "num-traits",
  "rayon",
  "safetensors",
@@ -910,7 +911,7 @@ dependencies = [
 [[package]]
 name = "candle-transformers"
 version = "0.9.1"
-source = "git+https://github.com/huggingface/candle.git?branch=main#be4f920f35b99af5919a86c0a2e08a78f3f63567"
+source = "git+https://github.com/huggingface/candle.git?branch=main#9fe623237a515f95f70d117c0a6da610c28a5ecd"
 dependencies = [
  "byteorder",
  "candle-core",
@@ -1531,10 +1532,11 @@ checksum = "1f791803201ab277ace03903de1594460708d2d54df6053f2d9e82f592b19e3b"
 
 [[package]]
 name = "cudarc"
-version = "0.16.6"
+version = "0.17.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "17200eb07e7d85a243aa1bf4569a7aa998385ba98d14833973a817a63cc86e92"
+checksum = "72ba848ae5c6f3cb36e71eab5f268763e3fabcabe3f7bc683e16f7fa3d46281e"
 dependencies = [
+ "float8 0.3.0",
  "half",
  "libloading",
 ]
@@ -1872,16 +1874,6 @@ dependencies = [
  "linux-raw-sys 0.6.5",
 ]
 
-[[package]]
-name = "dyn-stack"
-version = "0.10.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "56e53799688f5632f364f8fb387488dd05db9fe45db7011be066fc20e7027f8b"
-dependencies = [
- "bytemuck",
- "reborrow",
-]
-
 [[package]]
 name = "dyn-stack"
 version = "0.13.0"
@@ -2169,8 +2161,18 @@ dependencies = [
 
 [[package]]
 name = "float8"
-version = "0.2.1"
-source = "git+https://github.com/zackangelo/float8?branch=cudarc_0_16#03c1f5fe7cdb2f9cb690823fdd40593be57c408f"
+version = "0.3.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0f498aec3b227cd892ce18967f4033d9d397d28a80a7ab67e9f6b0176a79654e"
+dependencies = [
+ "half",
+]
+
+[[package]]
+name = "float8"
+version = "0.4.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4203231de188ebbdfb85c11f3c20ca2b063945710de04e7b59268731e728b462"
 dependencies = [
  "cudarc",
  "half",
@@ -2451,58 +2453,23 @@ dependencies = [
  "slab",
 ]
 
-[[package]]
-name = "gemm"
-version = "0.17.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6ab24cc62135b40090e31a76a9b2766a501979f3070fa27f689c27ec04377d32"
-dependencies = [
- "dyn-stack 0.10.0",
- "gemm-c32 0.17.1",
- "gemm-c64 0.17.1",
- "gemm-common 0.17.1",
- "gemm-f16 0.17.1",
- "gemm-f32 0.17.1",
- "gemm-f64 0.17.1",
- "num-complex",
- "num-traits",
- "paste",
- "raw-cpuid 10.7.0",
- "seq-macro",
-]
-
 [[package]]
 name = "gemm"
 version = "0.18.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "ab96b703d31950f1aeddded248bc95543c9efc7ac9c4a21fda8703a83ee35451"
 dependencies = [
- "dyn-stack 0.13.0",
- "gemm-c32 0.18.2",
- "gemm-c64 0.18.2",
- "gemm-common 0.18.2",
- "gemm-f16 0.18.2",
- "gemm-f32 0.18.2",
- "gemm-f64 0.18.2",
+ "dyn-stack",
+ "gemm-c32",
+ "gemm-c64",
+ "gemm-common",
+ "gemm-f16",
+ "gemm-f32",
+ "gemm-f64",
  "num-complex",
  "num-traits",
  "paste",
- "raw-cpuid 11.5.0",
- "seq-macro",
-]
-
-[[package]]
-name = "gemm-c32"
-version = "0.17.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b9c030d0b983d1e34a546b86e08f600c11696fde16199f971cd46c12e67512c0"
-dependencies = [
- "dyn-stack 0.10.0",
- "gemm-common 0.17.1",
- "num-complex",
- "num-traits",
- "paste",
- "raw-cpuid 10.7.0",
+ "raw-cpuid",
  "seq-macro",
 ]
 
@@ -2512,27 +2479,12 @@ version = "0.18.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "f6db9fd9f40421d00eea9dd0770045a5603b8d684654816637732463f4073847"
 dependencies = [
- "dyn-stack 0.13.0",
- "gemm-common 0.18.2",
- "num-complex",
- "num-traits",
- "paste",
- "raw-cpuid 11.5.0",
- "seq-macro",
-]
-
-[[package]]
-name = "gemm-c64"
-version = "0.17.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "fbb5f2e79fefb9693d18e1066a557b4546cd334b226beadc68b11a8f9431852a"
-dependencies = [
- "dyn-stack 0.10.0",
- "gemm-common 0.17.1",
+ "dyn-stack",
+ "gemm-common",
  "num-complex",
  "num-traits",
  "paste",
- "raw-cpuid 10.7.0",
+ "raw-cpuid",
  "seq-macro",
 ]
 
@@ -2542,35 +2494,15 @@ version = "0.18.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "dfcad8a3d35a43758330b635d02edad980c1e143dc2f21e6fd25f9e4eada8edf"
 dependencies = [
- "dyn-stack 0.13.0",
- "gemm-common 0.18.2",
+ "dyn-stack",
+ "gemm-common",
  "num-complex",
  "num-traits",
  "paste",
- "raw-cpuid 11.5.0",
+ "raw-cpuid",
  "seq-macro",
 ]
 
-[[package]]
-name = "gemm-common"
-version = "0.17.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a2e7ea062c987abcd8db95db917b4ffb4ecdfd0668471d8dc54734fdff2354e8"
-dependencies = [
- "bytemuck",
- "dyn-stack 0.10.0",
- "half",
- "num-complex",
- "num-traits",
- "once_cell",
- "paste",
- "pulp 0.18.22",
- "raw-cpuid 10.7.0",
- "rayon",
- "seq-macro",
- "sysctl 0.5.5",
-]
-
 [[package]]
 name = "gemm-common"
 version = "0.18.2"
@@ -2578,36 +2510,18 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "a352d4a69cbe938b9e2a9cb7a3a63b7e72f9349174a2752a558a8a563510d0f3"
 dependencies = [
  "bytemuck",
- "dyn-stack 0.13.0",
+ "dyn-stack",
  "half",
  "libm",
  "num-complex",
  "num-traits",
  "once_cell",
  "paste",
- "pulp 0.21.5",
- "raw-cpuid 11.5.0",
- "rayon",
- "seq-macro",
- "sysctl 0.6.0",
-]
-
-[[package]]
-name = "gemm-f16"
-version = "0.17.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7ca4c06b9b11952071d317604acb332e924e817bd891bec8dfb494168c7cedd4"
-dependencies = [
- "dyn-stack 0.10.0",
- "gemm-common 0.17.1",
- "gemm-f32 0.17.1",
- "half",
- "num-complex",
- "num-traits",
- "paste",
- "raw-cpuid 10.7.0",
+ "pulp",
+ "raw-cpuid",
  "rayon",
  "seq-macro",
+ "sysctl",
 ]
 
 [[package]]
@@ -2616,60 +2530,30 @@ version = "0.18.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "cff95ae3259432f3c3410eaa919033cd03791d81cebd18018393dc147952e109"
 dependencies = [
- "dyn-stack 0.13.0",
- "gemm-common 0.18.2",
- "gemm-f32 0.18.2",
+ "dyn-stack",
+ "gemm-common",
+ "gemm-f32",
  "half",
  "num-complex",
  "num-traits",
  "paste",
- "raw-cpuid 11.5.0",
+ "raw-cpuid",
  "rayon",
  "seq-macro",
 ]
 
-[[package]]
-name = "gemm-f32"
-version = "0.17.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e9a69f51aaefbd9cf12d18faf273d3e982d9d711f60775645ed5c8047b4ae113"
-dependencies = [
- "dyn-stack 0.10.0",
- "gemm-common 0.17.1",
- "num-complex",
- "num-traits",
- "paste",
- "raw-cpuid 10.7.0",
- "seq-macro",
-]
-
 [[package]]
 name = "gemm-f32"
 version = "0.18.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "bc8d3d4385393304f407392f754cd2dc4b315d05063f62cf09f47b58de276864"
 dependencies = [
- "dyn-stack 0.13.0",
- "gemm-common 0.18.2",
+ "dyn-stack",
+ "gemm-common",
  "num-complex",
  "num-traits",
  "paste",
- "raw-cpuid 11.5.0",
- "seq-macro",
-]
-
-[[package]]
-name = "gemm-f64"
-version = "0.17.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "aa397a48544fadf0b81ec8741e5c0fba0043008113f71f2034def1935645d2b0"
-dependencies = [
- "dyn-stack 0.10.0",
- "gemm-common 0.17.1",
- "num-complex",
- "num-traits",
- "paste",
- "raw-cpuid 10.7.0",
+ "raw-cpuid",
  "seq-macro",
 ]
 
@@ -2679,12 +2563,12 @@ version = "0.18.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "35b2a4f76ce4b8b16eadc11ccf2e083252d8237c1b589558a49b0183545015bd"
 dependencies = [
- "dyn-stack 0.13.0",
- "gemm-common 0.18.2",
+ "dyn-stack",
+ "gemm-common",
  "num-complex",
  "num-traits",
  "paste",
- "raw-cpuid 11.5.0",
+ "raw-cpuid",
  "seq-macro",
 ]
 
@@ -3877,7 +3761,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "07033963ba89ebaf1584d767badaa2e8fcec21aedea6b8c0346d487d49c28667"
 dependencies = [
  "cfg-if",
- "windows-targets 0.53.3",
+ "windows-targets 0.48.5",
 ]
 
 [[package]]
@@ -4466,7 +4350,7 @@ version = "0.7.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "77e878c846a8abae00dd069496dbe8751b16ac1c3d6bd2a7283a938e8228f90d"
 dependencies = [
- "proc-macro-crate 3.3.0",
+ "proc-macro-crate 1.3.1",
  "proc-macro2",
  "quote",
  "syn 2.0.106",
@@ -5277,18 +5161,6 @@ version = "1.0.17"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "3eb8486b569e12e2c32ad3e204dbaba5e4b5b216e9367044f25f1dba42341773"
 
-[[package]]
-name = "pulp"
-version = "0.18.22"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a0a01a0dc67cf4558d279f0c25b0962bd08fc6dec0137699eae304103e882fe6"
-dependencies = [
- "bytemuck",
- "libm",
- "num-complex",
- "reborrow",
-]
-
 [[package]]
 name = "pulp"
 version = "0.21.5"
@@ -5463,15 +5335,6 @@ version = "1.6.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "f93e7e49bb0bf967717f7bd674458b3d6b0c5f48ec7e3038166026a69fc22223"
 
-[[package]]
-name = "raw-cpuid"
-version = "10.7.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6c297679cb867470fa8c9f67dbba74a78d78e3e98d7cf2b08d6d71540f797332"
-dependencies = [
- "bitflags 1.3.2",
-]
-
 [[package]]
 name = "raw-cpuid"
 version = "11.5.0"
@@ -5871,7 +5734,7 @@ dependencies = [
  "errno",
  "libc",
  "linux-raw-sys 0.4.15",
- "windows-sys 0.59.0",
+ "windows-sys 0.52.0",
 ]
 
 [[package]]
@@ -6755,20 +6618,6 @@ dependencies = [
  "libc",
 ]
 
-[[package]]
-name = "sysctl"
-version = "0.5.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ec7dddc5f0fee506baf8b9fdb989e242f17e4b11c61dfbb0635b705217199eea"
-dependencies = [
- "bitflags 2.9.2",
- "byteorder",
- "enum-as-inner",
- "libc",
- "thiserror 1.0.69",
- "walkdir",
-]
-
 [[package]]
 name = "sysctl"
 version = "0.6.0"
@@ -6826,7 +6675,7 @@ dependencies = [
  "getrandom 0.3.3",
  "once_cell",
  "rustix 1.0.8",
- "windows-sys 0.59.0",
+ "windows-sys 0.52.0",
 ]
 
 [[package]]
@@ -7301,11 +7150,11 @@ dependencies = [
 
 [[package]]
 name = "ug"
-version = "0.4.0"
+version = "0.5.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "90b70b37e9074642bc5f60bb23247fd072a84314ca9e71cdf8527593406a0dd3"
+checksum = "76b761acf8af3494640d826a8609e2265e19778fb43306c7f15379c78c9b05b0"
 dependencies = [
- "gemm 0.18.2",
+ "gemm",
  "half",
  "libloading",
  "memmap2 0.9.7",
@@ -7322,9 +7171,9 @@ dependencies = [
 
 [[package]]
 name = "ug-cuda"
-version = "0.4.0"
+version = "0.5.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "14053653d0b7fa7b21015aa9a62edc8af2f60aa6f9c54e66386ecce55f22ed29"
+checksum = "9f0a1fa748f26166778c33b8498255ebb7c6bffb472bcc0a72839e07ebb1d9b5"
 dependencies = [
  "cudarc",
  "half",
@@ -7982,7 +7831,7 @@ version = "0.1.9"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "cf221c93e13a30d793f7645a0e7762c55d169dbb0a49671918a2319d289b10bb"
 dependencies = [
- "windows-sys 0.59.0",
+ "windows-sys 0.48.0",
 ]
 
 [[package]]