add models

serve/gradio_web_server.py

@@ -25,7 +25,7 @@ from serve import INTERACTION_MODES_MAP, MODEL_ID_TO_NAME
 
 LOGDIR = "/logs"
 
-logger = build_logger("gradio_web_server", "gradio_web_server.log")
+# logger = build_logger("gradio_web_server", "gradio_web_server.log")
 
 headers = {"User-Agent": "PrismaticVLMs Client"}
 
@@ -48,7 +48,7 @@ def get_model_list():
     models = sorted(
         models, key=lambda x: list(MODEL_ID_TO_NAME.values()).index(x) if x in MODEL_ID_TO_NAME.values() else len(models)
     )
-    logger.info(f"Models: {models}")
+    # logger.info(f"Models: {models}")
     return models
 
 
@@ -63,7 +63,7 @@ function() {
 
 
 def load_demo(url_params, request: gr.Request):
-    logger.info(f"load_demo. ip: {request.client.host}. params: {url_params}")
+    # logger.info(f"load_demo. ip: {request.client.host}. params: {url_params}")
 
     dropdown_update = gr.Dropdown.update(visible=True)
     if "model" in url_params:
@@ -76,7 +76,7 @@ def load_demo(url_params, request: gr.Request):
 
 
 def load_demo_refresh_model_list(request: gr.Request):
-    logger.info(f"load_demo. ip: {request.client.host}")
+    # logger.info(f"load_demo. ip: {request.client.host}")
     models = get_model_list()
     state = default_conversation.copy()
     dropdown_update = gr.Dropdown.update(choices=models, value=models[0] if len(models) > 0 else "")
@@ -84,19 +84,20 @@ def load_demo_refresh_model_list(request: gr.Request):
 
 
 def vote_last_response(state, vote_type, model_selector, request: gr.Request):
-    with open(get_conv_log_filename(), "a") as fout:
-        data = {
-            "tstamp": round(time.time(), 4),
-            "type": vote_type,
-            "model": model_selector,
-            "state": state.dict(),
-            "ip": request.client.host,
-        }
-        fout.write(json.dumps(data) + "\n")
+    pass
+    # with open(get_conv_log_filename(), "a") as fout:
+    #     data = {
+    #         "tstamp": round(time.time(), 4),
+    #         "type": vote_type,
+    #         "model": model_selector,
+    #         "state": state.dict(),
+    #         "ip": request.client.host,
+    #     }
+    #     fout.write(json.dumps(data) + "\n")
 
 
 def regenerate(state, image_process_mode, request: gr.Request):
-    logger.info(f"regenerate. ip: {request.client.host}")
+    # logger.info(f"regenerate. ip: {request.client.host}")
     state.messages[-1][-1] = None
     prev_human_msg = state.messages[-2]
     if type(prev_human_msg[1]) in (tuple, list):
@@ -106,13 +107,13 @@ def regenerate(state, image_process_mode, request: gr.Request):
 
 
 def clear_history(request: gr.Request):
-    logger.info(f"clear_history. ip: {request.client.host}")
+    # logger.info(f"clear_history. ip: {request.client.host}")
     state = default_conversation.copy()
     return (state, state.to_gradio_chatbot(), "", None) + (disable_btn,) * 5
 
 
 def add_text(state, text, image, image_process_mode, request: gr.Request):
-    logger.info(f"add_text. ip: {request.client.host}. len: {len(text)}")
+    # logger.info(f"add_text. ip: {request.client.host}. len: {len(text)}")
     if len(text) <= 0 and image is None:
         state.skip_next = True
         return (state, state.to_gradio_chatbot(), "", None) + (no_change_btn,) * 5
@@ -138,7 +139,7 @@ def add_text(state, text, image, image_process_mode, request: gr.Request):
 
 
 def http_bot(state, model_selector, interaction_mode, temperature, max_new_tokens, request: gr.Request):
-    logger.info(f"http_bot. ip: {request.client.host}")
+    # logger.info(f"http_bot. ip: {request.client.host}")
     start_tstamp = time.time()
     model_name = model_selector
 
@@ -159,7 +160,7 @@ def http_bot(state, model_selector, interaction_mode, temperature, max_new_token
     controller_url = args.controller_url
     ret = requests.post(controller_url + "/get_worker_address", json={"model": model_name})
     worker_addr = ret.json()["address"]
-    logger.info(f"model_name: {model_name}, worker_addr: {worker_addr}")
+    # logger.info(f"model_name: {model_name}, worker_addr: {worker_addr}")
 
     # No available worker
     if worker_addr == "":
@@ -188,7 +189,7 @@ def http_bot(state, model_selector, interaction_mode, temperature, max_new_token
         "max_new_tokens": int(max_new_tokens),
         "images": f"List of {len(state.get_images())} images: {all_image_hash}",
     }
-    logger.info(f"==== request ====\n{pload}")
+    # logger.info(f"==== request ====\n{pload}")
 
     pload["images"] = state.get_images()
 
@@ -228,20 +229,20 @@ def http_bot(state, model_selector, interaction_mode, temperature, max_new_token
     yield (state, state.to_gradio_chatbot()) + (enable_btn,) * 5
 
     finish_tstamp = time.time()
-    logger.info(f"{output}")
-
-    with open(get_conv_log_filename(), "a") as fout:
-        data = {
-            "tstamp": round(finish_tstamp, 4),
-            "type": "chat",
-            "model": model_name,
-            "start": round(start_tstamp, 4),
-            "finish": round(finish_tstamp, 4),
-            "state": state.dict(),
-            "images": all_image_hash,
-            "ip": request.client.host,
-        }
-        fout.write(json.dumps(data) + "\n")
+    # logger.info(f"{output}")
+
+    # with open(get_conv_log_filename(), "a") as fout:
+    #     data = {
+    #         "tstamp": round(finish_tstamp, 4),
+    #         "type": "chat",
+    #         "model": model_name,
+    #         "start": round(start_tstamp, 4),
+    #         "finish": round(finish_tstamp, 4),
+    #         "state": state.dict(),
+    #         "images": all_image_hash,
+    #         "ip": request.client.host,
+    #     }
+    #     fout.write(json.dumps(data) + "\n")
 
 
 title_markdown = """
@@ -425,11 +426,11 @@ if __name__ == "__main__":
     parser.add_argument("--moderate", action="store_true")
     parser.add_argument("--embed", action="store_true")
     args = parser.parse_args()
-    logger.info(f"args: {args}")
+    # logger.info(f"args: {args}")
 
     models = get_model_list()
 
-    logger.info(args)
+    # logger.info(args)
     demo = build_demo(args.embed)
     demo.queue(concurrency_count=args.concurrency_count, api_open=False).launch(
         server_name=args.host, server_port=args.port, share=args.share

startup.sh

@@ -1,3 +1,5 @@
 #!/bin/sh
 python -m serve.controller --host 0.0.0.0 --port 10000 &
-python -m serve.gradio_web_server --controller http://localhost:10000 --model-list-mode reload --share
+python -m serve.gradio_web_server --controller http://localhost:10000 --model-list-mode reload --share &
+python -m interactive_demo --port 40000 --model_id prism-dinosiglip+7b &
+python -m interactive_demo --port 40001 --model_family llava-v15 --model_id llava-v1.5-7b --model_dir liuhaotian/llava-v1.5-7b &