Added README.md for text_to_image_reference_using_torch_loadgen

text_to_image_reference_using_torch_loadgen/README.md

@@ -0,0 +1,43 @@
+
+# MLPerf Inference - Text to Image - ifferent backend(pytorch, pytorch-dist, debug)
+This implementation runs text to image model with the different backend.
+
+Currently it supports the following models:
+- stable-diffusion-xl
+
+## Accuracy Full Run(5000 images):
+- SingleStream
+Accuracy Command:
+```
+axs byquery loadgen_output,task=text_to_image,framework=torch,profile=stable-diffusion-xl-pytorch,dtype=fp16,loadgen_scenario=SingleStream,accuracy+
+
+```
+Accuracy:
+```
+axs byquery loadgen_output,task=text_to_image,framework=torch,profile=stable-diffusion-xl-pytorch,dtype=fp16,loadgen_scenario=SingleStream,accuracy+ , get accuracy_dict
+{'fid': 23.61146873078252, 'clip': 31.749096275568007}
+````
+
+```
+axs byquery loadgen_output,task=text_to_image,framework=torch,profile=stable-diffusion-xl-pytorch,dtype=fp16,loadgen_scenario=SingleStream,accuracy+ , get fid
+23.61146873078252
+``
+
+```
+axs byquery loadgen_output,task=text_to_image,framework=torch,profile=stable-diffusion-xl-pytorch,dtype=fp16,loadgen_scenario=SingleStream,accuracy+ , get clip
+31.749096275568007
+
+```
+
+## Perfomance Full Run
+- SingleStream
+Performance Command
+```
+axs byquery loadgen_output,task=text_to_image,framework=torch,profile=stable-diffusion-xl-pytorch,dtype=fp16,loadgen_scenario=SingleStream,loadgen_target_latency=6,loadgen_mode=PerformanceOnly,sut_name=7920t-kilt-onnxruntime_gpu
+```
+Performance
+```
+axs byquery loadgen_output,task=text_to_image,framework=torch,profile=stable-diffusion-xl-pytorch,dtype=fp16,loadgen_scenario=SingleStream,loadgen_target_latency=6,loadgen_mode=PerformanceOnly,sut_name=7920t-kilt-onnxruntime_gpu , get performance
+VALID : _Early_stopping_90th_percentile_estimate=4621.172 (milliseconds)
+
+```