add tgi

app.py

@@ -8,20 +8,21 @@ from run import run_benchmark
 from config_store import (
     get_training_config,
     get_inference_config,
+    get_text_generation_inference_config,
     get_neural_compressor_config,
     get_onnxruntime_config,
     get_openvino_config,
     get_pytorch_config,
 )
 
-BACKENDS = ["pytorch", "onnxruntime", "openvino", "neural-compressor"]
+BACKENDS = ["pytorch", "onnxruntime", "openvino", "neural-compressor", "text-generation-inference"]
 BENCHMARKS = ["inference", "training"]
 DEVICES = ["cpu", "cuda"]
 
 
 with gr.Blocks() as demo:
     # title text
-    gr.HTML("<h1 style='text-align: center'>🤗 Optimum-Benchmark UI   🏋️</h1>")
+    gr.HTML("<h1 style='text-align: center'>🤗 Optimum-Benchmark UI 🏋️</h1>")
     # explanation text
     gr.Markdown(
         "This is a demo space of [Optimum-Benchmark](https://github.com/huggingface/optimum-benchmark.git):"
@@ -72,6 +73,8 @@ with gr.Blocks() as demo:
                     openvino_config = get_openvino_config()
                 with gr.Accordion(label="Neural Compressor Config", open=False, visible=False):
                     neural_compressor_config = get_neural_compressor_config()
+                with gr.Accordion(label="Text Generation Inference Config", open=False, visible=False):
+                    text_generation_inference_config = get_text_generation_inference_config()
 
         # hide backend configs based on backend
         backend.change(

config_store.py

@@ -114,9 +114,7 @@ def get_pytorch_config():
 
 def get_onnxruntime_config():
     return get_base_backend_config(backend_name="onnxruntime")
-        # no_weights
-
-
+    # no_weights
 
     # no_weights: bool = False
 
@@ -169,12 +167,17 @@ def get_onnxruntime_config():
     # peft_strategy: Optional[str] = None
     # peft_config: Dict[str, Any] = field(default_factory=dict)
 
+
 def get_openvino_config():
     return get_base_backend_config(backend_name="openvino")
 
 
 def get_neural_compressor_config():
-    return get_base_backend_config(backend_name="neural_compressor")
+    return get_base_backend_config(backend_name="neural-compressor")
+
+
+def get_text_generation_inference_config():
+    return get_base_backend_config(backend_name="text-generation-inference")
 
 
 def get_inference_config():

requirements.txt

@@ -1,3 +1,4 @@
 gradio==3.41
+docker==6.1.3
 ansi2html==1.8.0
 optimum-benchmark[onnxruntime,openvino,neural-compressor,diffusers,peft]@git+https://github.com/huggingface/optimum-benchmark.git