diff --git a/pyproject.toml b/pyproject.toml
index 1b3be23dc19..3f1e1b64fc5 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -110,6 +110,7 @@ dev = [
     "quart",
     "openai[aiohttp]",
     "orjson",
+    "fast-hadamard-transform",
 ]
 
 ### 'lts' extra is deprecated                                                   ###
@@ -180,6 +181,7 @@ no-build-isolation-package = [
     "mamba-ssm",
     "transformer-engine",
     "transformer-engine-torch",
+    "fast-hadamard-transform",
 ]
 link-mode = "copy"
 # We don't define override-dependencies globally but rather locally where we need it.
@@ -191,14 +193,20 @@ override-dependencies = [
     "triton; sys_platform == 'never'",
 ]
 
+[[tool.uv.dependency-metadata]]
+name = "fast-hadamard-transform"
+version = "1.0.4.post1"
+requires-dist = ["torch", "packaging", "ninja"]
+
 [tool.uv.sources]
 
 flash_mla = [
-    { git = "https://github.com/deepseek-ai/FlashMLA", rev = "9edee0c022cd0938148a18e334203b0aab43aa19" },
+    { git = "https://github.com/deepseek-ai/FlashMLA", rev = "nv_dev" },
 ]
 transformer-engine = { git = "https://github.com/NVIDIA/TransformerEngine.git", rev = "4220403e831d29e93868f7793693ea83f6b8b05b" }
 nemo-run = { git = "https://github.com/NVIDIA-NeMo/Run.git", rev = "17ae86b64d7f75653351664f5d8c9e466faede00" }
 emerging_optimizers = { git = "https://github.com/NVIDIA-NeMo/Emerging-Optimizers.git", rev = "v0.2.0" }
+fast-hadamard-transform = { git = "https://github.com/Dao-AILab/fast-hadamard-transform.git", rev = "f134af63deb2df17e1171a9ec1ea4a7d8604d5ca" }
 
 [tool.isort]
 profile = "black"                                                          # black-compatible
diff --git a/uv.lock b/uv.lock
index de6edb9d9bd..7af5f7007ac 100644
--- a/uv.lock
+++ b/uv.lock
@@ -1,5 +1,5 @@
 version = 1
-revision = 2
+revision = 3
 requires-python = ">=3.12"
 resolution-markers = [
     "python_full_version >= '3.14' and sys_platform == 'win32'",
@@ -20,6 +20,11 @@ overrides = [
     { name = "triton", marker = "sys_platform == 'never'" },
 ]
 
+[[manifest.dependency-metadata]]
+name = "fast-hadamard-transform"
+version = "1.0.4.post1"
+requires-dist = ["torch", "packaging", "ninja"]
+
 [[package]]
 name = "absl-py"
 version = "2.4.0"
@@ -1122,6 +1127,16 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/37/f9/f8497ef8b873a8bb2a750ee2a6c5f0fc22258e1acb6245fd237042a6c279/fabric-3.2.3-py3-none-any.whl", hash = "sha256:ce61917f4f398018337ce279b357650a3a74baecf3fdd53a5839013944af965e", size = 59502, upload-time = "2026-04-06T00:00:10.176Z" },
 ]
 
+[[package]]
+name = "fast-hadamard-transform"
+version = "1.0.4.post1"
+source = { git = "https://github.com/Dao-AILab/fast-hadamard-transform.git?rev=f134af63deb2df17e1171a9ec1ea4a7d8604d5ca#f134af63deb2df17e1171a9ec1ea4a7d8604d5ca" }
+dependencies = [
+    { name = "ninja" },
+    { name = "packaging" },
+    { name = "torch", marker = "sys_platform == 'never'" },
+]
+
 [[package]]
 name = "fastapi"
 version = "0.138.0"
@@ -1213,8 +1228,8 @@ wheels = [
 
 [[package]]
 name = "flash-mla"
-version = "1.0.0+9edee0c"
-source = { git = "https://github.com/deepseek-ai/FlashMLA?rev=9edee0c022cd0938148a18e334203b0aab43aa19#9edee0c022cd0938148a18e334203b0aab43aa19" }
+version = "1.0.0+b7643bd"
+source = { git = "https://github.com/deepseek-ai/FlashMLA?rev=nv_dev#b7643bd54521f563b839b98289b5cd048c062ba2" }
 
 [[package]]
 name = "flashinfer-python"
@@ -2172,6 +2187,7 @@ dev = [
     { name = "datasets" },
     { name = "einops" },
     { name = "emerging-optimizers" },
+    { name = "fast-hadamard-transform" },
     { name = "fastapi" },
     { name = "flash-linear-attention" },
     { name = "flashinfer-python" },
@@ -2275,6 +2291,7 @@ requires-dist = [
     { name = "datasets", marker = "extra == 'dev'" },
     { name = "einops", marker = "extra == 'dev'", specifier = "~=0.8" },
     { name = "emerging-optimizers", marker = "extra == 'dev'", git = "https://github.com/NVIDIA-NeMo/Emerging-Optimizers.git?rev=v0.2.0" },
+    { name = "fast-hadamard-transform", marker = "extra == 'dev'", git = "https://github.com/Dao-AILab/fast-hadamard-transform.git?rev=f134af63deb2df17e1171a9ec1ea4a7d8604d5ca" },
     { name = "fastapi", marker = "extra == 'dev'", specifier = "~=0.50" },
     { name = "flash-linear-attention", marker = "extra == 'dev'", specifier = "~=0.4.0" },
     { name = "flashinfer-python", marker = "extra == 'dev'", specifier = ">=0.5.0,<0.7.0" },
@@ -2343,7 +2360,7 @@ linting = [
 ]
 no-pypi-wheels = [
     { name = "emerging-optimizers", git = "https://github.com/NVIDIA-NeMo/Emerging-Optimizers.git?rev=v0.2.0" },
-    { name = "flash-mla", git = "https://github.com/deepseek-ai/FlashMLA?rev=9edee0c022cd0938148a18e334203b0aab43aa19" },
+    { name = "flash-mla", git = "https://github.com/deepseek-ai/FlashMLA?rev=nv_dev" },
 ]
 test = [
     { name = "coverage" },
@@ -5129,14 +5146,14 @@ version = "2.12.1"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
     { name = "cuda-bindings", marker = "sys_platform == 'linux'" },
-    { name = "filelock" },
-    { name = "fsspec" },
-    { name = "jinja2" },
-    { name = "networkx" },
-    { name = "setuptools" },
-    { name = "sympy" },
+    { name = "filelock", marker = "sys_platform != 'emscripten' and sys_platform != 'win32'" },
+    { name = "fsspec", marker = "sys_platform != 'emscripten' and sys_platform != 'win32'" },
+    { name = "jinja2", marker = "sys_platform != 'emscripten' and sys_platform != 'win32'" },
+    { name = "networkx", marker = "sys_platform != 'emscripten' and sys_platform != 'win32'" },
+    { name = "setuptools", marker = "sys_platform != 'emscripten' and sys_platform != 'win32'" },
+    { name = "sympy", marker = "sys_platform != 'emscripten' and sys_platform != 'win32'" },
     { name = "triton", marker = "sys_platform == 'never'" },
-    { name = "typing-extensions" },
+    { name = "typing-extensions", marker = "sys_platform != 'emscripten' and sys_platform != 'win32'" },
 ]
 
 [[package]]