mirror of
https://github.com/kvcache-ai/ktransformers.git
synced 2025-09-05 12:09:48 +00:00
45 lines
1.4 KiB
YAML
45 lines
1.4 KiB
YAML
- match:
|
|
name: "^model\\.layers\\..*\\."
|
|
replace:
|
|
class: "default"
|
|
kwargs:
|
|
generate_device: "cuda"
|
|
prefill_device: "cuda"
|
|
- match:
|
|
class: ktransformers.models.modeling_mixtral.MixtralRotaryEmbedding
|
|
replace:
|
|
class: ktransformers.operators.RoPE.RotaryEmbedding
|
|
- match:
|
|
name: "^model\\.layers\\..*$"
|
|
class: torch.nn.Linear # only match modules matching name and class simultaneously
|
|
replace:
|
|
class: ktransformers.operators.linear.KTransformerLinear # optimized Kernel on quantized data types
|
|
kwargs:
|
|
generate_device: "cuda"
|
|
prefill_device: "cuda"
|
|
generate_op: "QuantizedLinearMarlin"
|
|
prefill_op: "QuantizedLinearTorch"
|
|
- match:
|
|
name: "^model\\.layers\\..*\\.block_sparse_moe$"
|
|
class: ktransformers.models.modeling_mixtral.MixtralSparseMoeBlock
|
|
replace:
|
|
class: ktransformers.operators.experts.MisrtalSparseMoEBlockInjected
|
|
- match:
|
|
name: "^model\\.layers\\..*\\.block_sparse_moe\\.experts$"
|
|
replace:
|
|
class: ktransformers.operators.experts.KTransformersMLPExpert
|
|
kwargs:
|
|
prefill_device: "cuda"
|
|
prefill_mlp_type: "MLPExpertsTorch"
|
|
generate_device: "cpu"
|
|
generate_mlp_type: "MLPCPUExperts"
|
|
out_device: "cuda"
|
|
recursive: False # don't recursively inject submodules of this module
|
|
|
|
- match:
|
|
name: "^model.embed_tokens"
|
|
replace:
|
|
class: "default"
|
|
kwargs:
|
|
generate_device: "cpu"
|
|
prefill_device: "cpu"
|