From c47205dce9b7a64a47fc4cd3cd55f2fd6926f98b Mon Sep 17 00:00:00 2001 From: TangJingqi Date: Thu, 15 Aug 2024 11:25:12 +0800 Subject: [PATCH] fix name --- .../optimize/optimize_rules/DeepSeek-V2-Chat-multi-gpu-4.yaml | 2 +- .../optimize/optimize_rules/DeepSeek-V2-Chat-multi-gpu.yaml | 2 +- ktransformers/optimize/optimize_rules/DeepSeek-V2-Chat.yaml | 2 +- .../optimize_rules/DeepSeek-V2-Lite-Chat-multi-gpu.yaml | 2 +- .../optimize_rules/Qwen2-57B-A14B-Instruct-multi-gpu.yaml | 2 +- .../optimize/optimize_rules/Qwen2-57B-A14B-Instruct.yaml | 2 +- 6 files changed, 6 insertions(+), 6 deletions(-) diff --git a/ktransformers/optimize/optimize_rules/DeepSeek-V2-Chat-multi-gpu-4.yaml b/ktransformers/optimize/optimize_rules/DeepSeek-V2-Chat-multi-gpu-4.yaml index 5f3b780..d7adfa2 100644 --- a/ktransformers/optimize/optimize_rules/DeepSeek-V2-Chat-multi-gpu-4.yaml +++ b/ktransformers/optimize/optimize_rules/DeepSeek-V2-Chat-multi-gpu-4.yaml @@ -190,7 +190,7 @@ - match: name: "^model$" replace: - class: "ktransformers.operators.layer_wise_prefill.KDeepseekV2Model" + class: "ktransformers.operators.models.KDeepseekV2Model" kwargs: per_layer_prefill_intput_threshold: 0 # 0 is close layer wise prefill transfer_map: diff --git a/ktransformers/optimize/optimize_rules/DeepSeek-V2-Chat-multi-gpu.yaml b/ktransformers/optimize/optimize_rules/DeepSeek-V2-Chat-multi-gpu.yaml index 882c75f..a21b22d 100644 --- a/ktransformers/optimize/optimize_rules/DeepSeek-V2-Chat-multi-gpu.yaml +++ b/ktransformers/optimize/optimize_rules/DeepSeek-V2-Chat-multi-gpu.yaml @@ -103,7 +103,7 @@ - match: name: "^model$" replace: - class: "ktransformers.operators.layer_wise_prefill.KDeepseekV2Model" + class: "ktransformers.operators.models.KDeepseekV2Model" kwargs: per_layer_prefill_intput_threshold: 0 # 0 is close layer wise prefill transfer_map: diff --git a/ktransformers/optimize/optimize_rules/DeepSeek-V2-Chat.yaml b/ktransformers/optimize/optimize_rules/DeepSeek-V2-Chat.yaml index 85d2e82..9d029a9 100644 --- a/ktransformers/optimize/optimize_rules/DeepSeek-V2-Chat.yaml +++ b/ktransformers/optimize/optimize_rules/DeepSeek-V2-Chat.yaml @@ -55,7 +55,7 @@ - match: name: "^model$" replace: - class: "ktransformers.operators.layer_wise_prefill.KDeepseekV2Model" + class: "ktransformers.operators.models.KDeepseekV2Model" kwargs: generate_device: "cuda" prefill_device: "cuda" diff --git a/ktransformers/optimize/optimize_rules/DeepSeek-V2-Lite-Chat-multi-gpu.yaml b/ktransformers/optimize/optimize_rules/DeepSeek-V2-Lite-Chat-multi-gpu.yaml index 4115592..cfd77dc 100644 --- a/ktransformers/optimize/optimize_rules/DeepSeek-V2-Lite-Chat-multi-gpu.yaml +++ b/ktransformers/optimize/optimize_rules/DeepSeek-V2-Lite-Chat-multi-gpu.yaml @@ -103,7 +103,7 @@ - match: name: "^model$" replace: - class: "ktransformers.operators.layer_wise_prefill.KDeepseekV2Model" + class: "ktransformers.operators.models.KDeepseekV2Model" kwargs: per_layer_prefill_intput_threshold: 0 # 0 is close layer wise prefill transfer_map: diff --git a/ktransformers/optimize/optimize_rules/Qwen2-57B-A14B-Instruct-multi-gpu.yaml b/ktransformers/optimize/optimize_rules/Qwen2-57B-A14B-Instruct-multi-gpu.yaml index 37c8a36..bfa60b7 100644 --- a/ktransformers/optimize/optimize_rules/Qwen2-57B-A14B-Instruct-multi-gpu.yaml +++ b/ktransformers/optimize/optimize_rules/Qwen2-57B-A14B-Instruct-multi-gpu.yaml @@ -89,7 +89,7 @@ - match: name: "^model$" replace: - class: "ktransformers.operators.layer_wise_prefill.KQwen2MoeModel" + class: "ktransformers.operators.models.KQwen2MoeModel" kwargs: per_layer_prefill_intput_threshold: 0 # 0 is close layer wise prefill transfer_map: diff --git a/ktransformers/optimize/optimize_rules/Qwen2-57B-A14B-Instruct.yaml b/ktransformers/optimize/optimize_rules/Qwen2-57B-A14B-Instruct.yaml index a44c750..073332c 100644 --- a/ktransformers/optimize/optimize_rules/Qwen2-57B-A14B-Instruct.yaml +++ b/ktransformers/optimize/optimize_rules/Qwen2-57B-A14B-Instruct.yaml @@ -38,7 +38,7 @@ - match: name: "^model$" replace: - class: "ktransformers.operators.layer_wise_prefill.KQwen2MoeModel" + class: "ktransformers.operators.models.KQwen2MoeModel" kwargs: per_layer_prefill_intput_threshold: 0 # 0 is close layer wise prefill - match: