mirror of
https://github.com/kvcache-ai/ktransformers.git
synced 2025-09-09 13:55:27 +00:00
⚡ ready to publish
This commit is contained in:
parent
f892d22849
commit
83401dbb3b
6 changed files with 157 additions and 19 deletions
|
@ -10,7 +10,7 @@
|
|||
name: "^model\\.layers\\.(0|[1-9]|[12][0-9])\\."
|
||||
class: ktransformers.models.modeling_deepseek_v3.DeepseekV3RotaryEmbedding
|
||||
replace:
|
||||
class: ktransformers.operators.RoPE.RotaryEmbeddingV3
|
||||
class: ktransformers.operators.RoPE.YarnRotaryEmbeddingV3
|
||||
kwargs:
|
||||
generate_device: "cuda:0"
|
||||
prefill_device: "cuda:0"
|
||||
|
@ -18,7 +18,7 @@
|
|||
name: "^model\\.layers\\.([3456][0-9])\\."
|
||||
class: ktransformers.models.modeling_deepseek_v3.DeepseekV3RotaryEmbedding
|
||||
replace:
|
||||
class: ktransformers.operators.RoPE.RotaryEmbeddingV3
|
||||
class: ktransformers.operators.RoPE.YarnRotaryEmbeddingV3
|
||||
kwargs:
|
||||
generate_device: "cuda:1"
|
||||
prefill_device: "cuda:1"
|
||||
|
|
|
@ -10,7 +10,7 @@
|
|||
name: "^model\\.layers\\.(0|[1-9]|[12][0-9])\\."
|
||||
class: ktransformers.models.modeling_deepseek_v3.DeepseekV3RotaryEmbedding
|
||||
replace:
|
||||
class: ktransformers.operators.RoPE.RotaryEmbeddingV3
|
||||
class: ktransformers.operators.RoPE.YarnRotaryEmbeddingV3
|
||||
kwargs:
|
||||
generate_device: "cuda:0"
|
||||
prefill_device: "cuda:0"
|
||||
|
@ -18,7 +18,7 @@
|
|||
name: "^model\\.layers\\.([3456][0-9])\\."
|
||||
class: ktransformers.models.modeling_deepseek_v3.DeepseekV3RotaryEmbedding
|
||||
replace:
|
||||
class: ktransformers.operators.RoPE.RotaryEmbeddingV3
|
||||
class: ktransformers.operators.RoPE.YarnRotaryEmbeddingV3
|
||||
kwargs:
|
||||
generate_device: "cuda:1"
|
||||
prefill_device: "cuda:1"
|
||||
|
|
|
@ -1,7 +1,7 @@
|
|||
- match:
|
||||
class: ktransformers.models.modeling_deepseek_v3.DeepseekV3RotaryEmbedding
|
||||
replace:
|
||||
class: ktransformers.operators.RoPE.RotaryEmbeddingV3
|
||||
class: ktransformers.operators.RoPE.YarnRotaryEmbeddingV3
|
||||
kwargs:
|
||||
generate_device: "cuda"
|
||||
prefill_device: "cuda"
|
||||
|
|
Loading…
Add table
Add a link
Reference in a new issue