Add mamba and megatron related build packages to no-build-isolation. (#579)

* Add mamba and megatron related build packages to `no-build-isolation`.

* Update `CONTRIBUTING.md` to show how to fix cloud/local CI fingerpint mismatch.

Author: FurtherAI

CONTRIBUTING.md

@@ -41,7 +41,7 @@ These checks are automatically run in CI for all pull requests. If your PR fails
 
 The PR `prek` workflow uses a prebuilt full `uv` cache (stored as a GitHub release asset) to avoid rebuilding heavy dependencies on every run.
 
-To refresh the cache after dependency changes, run:
+To refresh the cache after dependency changes, ensure your branch is rebased or merged with main, or checkout the PR merge branch, then run:
 
 ```bash
 bash scripts/ci/build_and_push_uv_cache.sh

pyproject.toml

@@ -124,14 +124,14 @@ required-version = ">=0.6.15"
 # numpy>=2 on Python 3.9+, but megatron-core requires numpy<2.
 override-dependencies = ["transformer-engine>=2.11.0", "numpy<2"]
 exclude-dependencies = ["pynvml"]
-no-build-isolation-package = ["apex", "transformer-engine-torch", "nv-grouped-gemm"]
+no-build-isolation-package = ["apex", "transformer-engine", "transformer-engine-cu12", "transformer-engine-torch", "megatron-core", "megatron-bridge", "nv-grouped-gemm", "mamba-ssm", "causal-conv1d"]
 
 [tool.uv.extra-build-dependencies]
 apex = ["torch>=2.8.0"]
 transformer-engine-torch = ["torch>=2.8.0"]
 
 [tool.uv.extra-build-variables]
-apex = { APEX_CPP_EXT = "1", APEX_CUDA_EXT = "1", APEX_FAST_LAYER_NORM = "1", APEX_PARALLEL_BUILD = "1", NVCC_APPEND_FLAGS = "--threads 1" }
+apex = { APEX_CPP_EXT = "1", APEX_CUDA_EXT = "1", APEX_FAST_LAYER_NORM = "1", APEX_PARALLEL_BUILD = "16", NVCC_APPEND_FLAGS = "--threads 4" }
 
 [tool.ty.environment]
 python-version = "3.11"