Better dyn atten sdpa usage estimation logic

This commit is contained in:
Disty0
2025-01-26 16:57:26 +03:00
parent a21bc9d9cd
commit b0ecdf2f1c
3 changed files with 40 additions and 57 deletions
+2 -2
View File
@@ -514,8 +514,8 @@ options_templates.update(options_section(('cuda', "Compute Settings"), {
"cross_attention_optimization": OptionInfo(startup_cross_attention, "Attention optimization method", gr.Radio, lambda: {"choices": shared_items.list_crossattention(native)}),
"sdp_options": OptionInfo(startup_sdp_options, "SDP options", gr.CheckboxGroup, {"choices": ['Flash attention', 'Memory attention', 'Math attention', 'Dynamic attention', 'Sage attention'], "visible": native}),
"xformers_options": OptionInfo(['Flash attention'], "xFormers options", gr.CheckboxGroup, {"choices": ['Flash attention'] }),
"dynamic_attention_slice_rate": OptionInfo(4, "Dynamic Attention slicing rate in GB", gr.Slider, {"minimum": 0.1, "maximum": gpu_memory, "step": 0.1, "visible": native}),
"dynamic_attention_trigger_rate": OptionInfo(6, "Dynamic Attention trigger rate in GB", gr.Slider, {"minimum": 0.1, "maximum": gpu_memory*2, "step": 0.1, "visible": native}),
"dynamic_attention_slice_rate": OptionInfo(2, "Dynamic Attention slicing rate in GB", gr.Slider, {"minimum": 0.1, "maximum": gpu_memory, "step": 0.1, "visible": native}),
"dynamic_attention_trigger_rate": OptionInfo(3, "Dynamic Attention trigger rate in GB", gr.Slider, {"minimum": 0.1, "maximum": gpu_memory*2, "step": 0.1, "visible": native}),
"sub_quad_sep": OptionInfo("<h3>Sub-quadratic options</h3>", "", gr.HTML, {"visible": not native}),
"sub_quad_q_chunk_size": OptionInfo(512, "Attention query chunk size", gr.Slider, {"minimum": 16, "maximum": 8192, "step": 8, "visible": not native}),
"sub_quad_kv_chunk_size": OptionInfo(512, "Attention kv chunk size", gr.Slider, {"minimum": 0, "maximum": 8192, "step": 8, "visible": not native}),