Skip to main content

ShardConfig

Configuration for model sharding.

Attributes

AttributeTypeDescription
engineLiteral["vllm"] = vllmThe sharding engine to use (currently only "vllm" is supported).
args[VLLMShardArgs](vllmshardargs.md?sid=flyte_prefetch__hf_model_vllmshardargs) = VLLMShardArgs()Arguments for the sharding engine.

Constructor

Signature

def ShardConfig(
engine: Literal["vllm"] = "vllm",
args: [VLLMShardArgs](vllmshardargs.md?sid=flyte_prefetch__hf_model_vllmshardargs) = VLLMShardArgs()
) - > null

Parameters

NameTypeDescription
engineLiteral["vllm"] = "vllm"The sharding engine to use (currently only "vllm" is supported).
args[VLLMShardArgs](vllmshardargs.md?sid=flyte_prefetch__hf_model_vllmshardargs) = VLLMShardArgs()Arguments for the sharding engine.

Signature

def ShardConfig(
engine: Literal["vllm"] = vllm,
args: [VLLMShardArgs](vllmshardargs.md?sid=flyte_prefetch__hf_model_vllmshardargs) = VLLMShardArgs()
) - > null

Parameters

NameTypeDescription
engineLiteral["vllm"] = vllmThe sharding engine to use for model distribution, currently restricted to "vllm".
args[VLLMShardArgs](vllmshardargs.md?sid=flyte_prefetch__hf_model_vllmshardargs) = VLLMShardArgs()A collection of engine-specific arguments and settings used to configure the sharding process.