For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /get-started.md).
Python function
apply_fused_kernel_flags
apply_fused_kernel_flags()
max.pipelines.weights.apply_fused_kernel_flags(config, state_dict)
Sets the fused-kernel eligibility flags on a freshly built QuantConfig.
The last step of parse_quant_config(), split out for callers that
build a config through one of the format-specific builders. Skipping it
leaves both flags off, silently dropping the fused MoE path.
-
Parameters:
-
- config (QuantConfig) – The config to finish, mutated in place.
- state_dict (Mapping[str, WeightData]) – The checkpoint weights, read to decide MLP fusion.
-
Returns:
-
config, for chaining. -
Return type: