Upload folder using huggingface_hub

README.md

@@ -7,7 +7,7 @@ sdk: gradio
 sdk_version: 6.15.2
 python_version: '3.12'
 app_file: app.py
-pinned: true
+pinned: false
 license: apache-2.0
 short_description: Space for quantizing models for use in ComfyUI
 ---

app.py

@@ -44,6 +44,12 @@ def run_quantization(
             quant_args["int8"] = True
             quant_args["scaling_mode"] = "row"
             suffix = "-int8mixedrow-simple"
+        elif quant_format == "int8-convrot":
+            quant_args["int8"] = True
+            quant_args["scaling_mode"] = "row"
+            quant_args["convrot"] = True
+            quant_args["convrot_group_size"] = 256
+            suffix = "-int8-convrot-simple"
         elif quant_format == "mxfp8":
             quant_args["mxfp8"] = True
             suffix = "-mxfp8mixed-simple"
@@ -119,7 +125,7 @@ with gr.Blocks() as demo:
             target_file_base = gr.Textbox(label="Target Filename Base (e.g. model-quant)")
 
             gr.Markdown("### Quantization Options")
-            quant_format = gr.Radio(["fp8", "int8 rowwise", "mxfp8"], value="fp8", label="Format")
+            quant_format = gr.Radio(["fp8", "int8 rowwise", "int8-convrot", "mxfp8"], value="fp8", label="Format")
 
             layer_filter = gr.Dropdown(
                 ["None", "Anima", "Microsoft Lens", "Flux2", "Chroma", "Radiance", "WAN", "LTX-2.x", "Qwen Image", "Z-Image"],

plan.md

@@ -19,6 +19,12 @@ int8 rowwise(add -int8mixedrow-simple to output model name):
 int8=True
 scaling_mode="row"
 
+int8-convrot(add -int8-convrot-simple to output model name):
+int8=True
+scaling_mode="row"
+convrot=True
+convrot_group_size=256
+
 mxfp8(add -mxfp8mixed-simple to output model name):
 mxfp8=True
 

requirements.txt

@@ -4,5 +4,4 @@ huggingface-hub
 spaces
 prodigy-plus-schedule-free
 convert-to-quant
-gradio
-comfy-kitchen
+gradio