add tools into previously built images (intel-analytics#9317)

* modify Dockerfile

* manually build

* modify Dockerfile

* add chat.py into inference-xpu

* add benchmark into inference-cpu

* manually build

* add benchmark into inference-cpu

* add benchmark into inference-cpu

* add benchmark into inference-cpu

* add chat.py into inference-xpu

* add chat.py into inference-xpu

* change ADD to COPY in dockerfile

* fix dependency issue

* temporarily remove run-spr in llm-cpu

* temporarily remove run-spr in llm-cpu

docker/llm/inference/xpu/docker/Dockerfile

@@ -4,6 +4,9 @@ ARG http_proxy
 ARG https_proxy
 
 ENV TZ=Asia/Shanghai
+ENV PYTHONUNBUFFERED=1
+
+COPY chat.py /llm/chat.py
 
 # Disable pip's cache behavior
 ARG PIP_NO_CACHE_DIR=false
@@ -33,4 +36,6 @@ RUN curl -fsSL https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-P
     pip install --pre --upgrade bigdl-llm[xpu] -f https://developer.intel.com/ipex-whl-stable-xpu && \
     # Install opencl-related repos
     apt-get update && \
-    apt-get install -y intel-opencl-icd intel-level-zero-gpu level-zero level-zero-dev
+    apt-get install -y intel-opencl-icd intel-level-zero-gpu level-zero level-zero-dev && \
+    # Install related libary of chat.py
+    pip install --upgrade colorama

docker/llm/inference/xpu/docker/chat.py

@@ -0,0 +1,102 @@
+#
+# Copyright 2016 The BigDL Authors.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+import intel_extension_for_pytorch as ipex
+import torch
+import argparse
+import sys
+# todo: support more model class
+from transformers import AutoModel, AutoModelForCausalLM, AutoTokenizer, AutoConfig
+from transformers import TextIteratorStreamer
+from transformers.tools.agents import StopSequenceCriteria
+from transformers.generation.stopping_criteria import StoppingCriteriaList
+from colorama import Fore
+from bigdl.llm import optimize_model
+SYSTEM_PROMPT = "A chat between a curious human <human> and an artificial intelligence assistant <bot>.\
+The assistant gives helpful, detailed, and polite answers to the human's questions."
+HUMAN_ID = "<human>"
+BOT_ID = "<bot>"
+# chat_history formated in [(iput_str, output_str)]
+def format_prompt(input_str,
+                  chat_history):
+    prompt = [f"{SYSTEM_PROMPT}\n"]
+    for history_input_str, history_output_str in chat_history:
+        prompt.append(f"{HUMAN_ID} {history_input_str}\n{BOT_ID} {history_output_str}\n")
+    prompt.append(f"{HUMAN_ID} {input_str}\n{BOT_ID} ")
+    return "".join(prompt)
+def stream_chat(model,
+                tokenizer,
+                stopping_criteria,
+                input_str,
+                chat_history):
+    prompt = format_prompt(input_str, chat_history)
+    # print(prompt)
+    input_ids = tokenizer([prompt], return_tensors="pt").to('xpu')
+    streamer = TextIteratorStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
+    generate_kwargs = dict(input_ids, streamer=streamer, max_new_tokens=512, stopping_criteria=stopping_criteria)
+    from threading import Thread
+    # to ensure non-blocking access to the generated text, generation process should be ran in a separate thread
+    thread = Thread(target=model.generate, kwargs=generate_kwargs)
+    thread.start()
+    output_str = []
+    print(Fore.BLUE+"BigDL-LLM: "+Fore.RESET, end="")
+    for partial_output_str in streamer:
+        output_str.append(partial_output_str)
+        # remove the last HUMAN_ID if exists
+        print(partial_output_str.replace(f"{HUMAN_ID}", ""), end="")
+    chat_history.append((input_str, "".join(output_str).replace(f"{HUMAN_ID}", "").rstrip()))
+def auto_select_model(model_name):
+    try:
+        try:
+            model = AutoModelForCausalLM.from_pretrained(model_path,
+                                                        low_cpu_mem_usage=True,
+                                                        torch_dtype="auto",
+                                                        trust_remote_code=True,
+                                                        use_cache=True)
+        except:
+            model = AutoModel.from_pretrained(model_path,
+                                             low_cpu_mem_usage=True,
+                                             torch_dtype="auto",
+                                             trust_remote_code=True,
+                                             use_cache=True)
+    except:
+        print("Sorry, the model you entered is not supported in installer.")
+        sys.exit()
+
+    return model
+
+if __name__ == "__main__":
+  parser = argparse.ArgumentParser()
+  parser.add_argument("--model-path", type=str, help="path to an llm")
+  args = parser.parse_args()
+  model_path = args.model_path
+
+  model = auto_select_model(model_path)
+  model = optimize_model(model)
+  model = model.to('xpu')
+  tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
+  stopping_criteria = StoppingCriteriaList([StopSequenceCriteria(HUMAN_ID, tokenizer)])
+  chat_history = []
+  while True:
+      with torch.inference_mode():
+          user_input = input(Fore.GREEN+"\nHuman: "+Fore.RESET)
+          if user_input == "stop": # let's stop the conversation when user input "stop"
+              break
+          stream_chat(model=model,
+                      tokenizer=tokenizer,
+                      stopping_criteria=stopping_criteria,
+                      input_str=user_input,
+                      chat_history=chat_history)
+

docker/llm/serving/cpu/docker/Dockerfile

@@ -10,8 +10,7 @@ ARG PIP_NO_CACHE_DIR=false
 COPY ./entrypoint.sh /opt/entrypoint.sh
 ADD  https://github.com/krallin/tini/releases/download/${TINI_VERSION}/tini /sbin/tini
 # Install Serving Dependencies
-RUN mkdir /llm && \
-    cd /llm && \
+RUN cd /llm && \
     git clone https://github.com/analytics-zoo/FastChat.git && \
     cd FastChat && \
     git checkout dev-2023-09-22 && \

docker/llm/serving/xpu/docker/Dockerfile

@@ -7,8 +7,7 @@ ARG https_proxy
 ARG PIP_NO_CACHE_DIR=false
 
 # Install Serving Dependencies
-RUN mkdir /llm && \
-    cd /llm && \
+RUN cd /llm && \
     git clone https://github.com/analytics-zoo/FastChat.git && \
     cd FastChat && \
     git checkout dev-2023-09-22 && \