[LLM] Mixtral CPU examples (#9673)

* Mixtral CPU PyTorch and hugging face examples, based on #9661 and #9671

README.md

@@ -143,7 +143,7 @@ Over 20 models have been optimized/verified on `bigdl-llm`, including *LLaMA/LLa
 | ChatGLM2   | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/chatglm2)  | [link](python/llm/example/GPU/HF-Transformers-AutoModels/Model/chatglm2)   |
 | ChatGLM3   | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/chatglm3)  | [link](python/llm/example/GPU/HF-Transformers-AutoModels/Model/chatglm3)   |
 | Mistral    | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/mistral)   | [link](python/llm/example/GPU/HF-Transformers-AutoModels/Model/mistral)    |
-| Mixtral    |    | [link](python/llm/example/GPU/HF-Transformers-AutoModels/Model/mixtral)    |
+| Mixtral    | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/mixtral)   | [link](python/llm/example/GPU/HF-Transformers-AutoModels/Model/mixtral)    |
 | Falcon     | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/falcon)    | [link](python/llm/example/GPU/HF-Transformers-AutoModels/Model/falcon)     |
 | MPT        | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/mpt)       | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/mpt)        |
 | Dolly-v1   | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/dolly_v1)  | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/dolly_v1)   | 

python/llm/example/CPU/HF-Transformers-AutoModels/Model/mixtral/README.md

@@ -0,0 +1,45 @@
+# Mixtral
+In this directory, you will find examples on how you could apply BigDL-LLM INT4 optimizations on Mixtral models on [Intel CPUs](../README.md). For illustration purposes, we utilize the [mistralai/Mixtral-8x7B-Instruct-v0.1](https://huggingface.co/mistralai/Mixtral-8x7B-Instruct-v0.1) as a reference Mixtral model.
+
+## Requirements
+To run these examples with BigDL-LLM on Intel CPUs, we have some recommended requirements for your machine, please refer to [here](../README.md#recommended-requirements) for more information.
+
+**Important: Please make sure you have installed `transformers==4.36.0` to run the example.**
+
+## Example: Predict Tokens using `generate()` API
+In the example [generate.py](./generate.py), we show a basic use case for a Mixtral model to predict the next N tokens using `generate()` API, with BigDL-LLM INT4 optimizations on Intel CPUs.
+### 1. Install
+We suggest using conda to manage the Python environment. For more information about conda installation, please refer to [here](https://docs.conda.io/en/latest/miniconda.html#).
+
+After installing conda, create a Python environment for BigDL-LLM:
+```bash
+conda create -n llm python=3.9 # recommend to use Python 3.9
+conda activate llm
+
+# below command will install PyTorch CPU as default
+pip install torch==2.0.1 --index-url https://download.pytorch.org/whl/cpu
+pip install --pre --upgrade bigdl-llm[all]
+
+# Please make sure you are using a stable version of Transformers, 4.36.0 or newer.
+pip install transformers==4.36.0
+```
+
+### 2. Run
+
+```bash
+python ./generate.py --prompt 'What is AI?'
+```
+
+In the example, several arguments can be passed to satisfy your requirements:
+
+- `--repo-id-or-model-path REPO_ID_OR_MODEL_PATH`: argument defining the huggingface repo id for the Mixtral model (e.g. `mistralai/Mixtral-8x7B-Instruct-v0.1`) to be downloaded, or the path to the huggingface checkpoint folder. It is default to be `'mistralai/Mixtral-8x7B-Instruct-v0.1'`.
+- `--prompt PROMPT`: argument defining the prompt to be infered (with integrated prompt format for chat). It is default to be `'What is AI?'`.
+- `--n-predict N_PREDICT`: argument defining the max number of tokens to predict. It is default to be `32`.
+
+#### Sample Output
+#### [mistralai/Mixtral-8x7B-Instruct-v0.1](https://huggingface.co/mistralai/Mixtral-8x7B-Instruct-v0.1)
+```log
+Inference time: xxxx s 
+-------------------- Output --------------------
+[INST] What is AI? [/INST] AI, or Artificial Intelligence, refers to the development of computer systems that can perform tasks that would normally require human intelligence to accomplish. These tasks can include things
+```

python/llm/example/CPU/HF-Transformers-AutoModels/Model/mixtral/generate.py

@@ -0,0 +1,72 @@
+#
+# 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 torch
+import time
+import argparse
+
+from bigdl.llm.transformers import AutoModelForCausalLM
+from transformers import AutoTokenizer
+
+# you could tune the prompt based on your own model,
+# here the prompt tuning refers to https://huggingface.co/mistralai/Mixtral-8x7B-Instruct-v0.1#instruction-format
+MIXTRAL_PROMPT_FORMAT = """<s>[INST] {prompt} [/INST]"""
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser(description='Predict Tokens using `generate()` API for Mixtral model')
+    parser.add_argument('--repo-id-or-model-path', type=str, default="'mistralai/Mixtral-8x7B-Instruct-v0.1'",
+                        help='The huggingface repo id for the Mixtral (e.g. `mistralai/Mixtral-8x7B-Instruct-v0.1`) to be downloaded,'
+                             ', or the path to the huggingface checkpoint folder.')
+    parser.add_argument('--prompt', type=str, default="What is AI?",
+                        help='Prompt to infer')
+    parser.add_argument('--n-predict', type=int, default=32,
+                        help='Max tokens to predict')
+
+    args = parser.parse_args()
+    model_path = args.repo_id_or_model_path
+
+    # Load model in 4 bit,
+    # which convert the relevant layers in the model into INT4 format
+    model = AutoModelForCausalLM.from_pretrained(model_path,
+                                                 load_in_4bit=True,
+                                                 optimize_model=True,
+                                                 trust_remote_code=True,
+                                                 use_cache=True)
+
+    # Load tokenizer
+    tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
+    
+    # Generate predicted tokens
+    with torch.inference_mode():
+        prompt = MIXTRAL_PROMPT_FORMAT.format(prompt=args.prompt)
+        input_ids = tokenizer.encode(prompt, return_tensors="pt").to('cpu')
+        output = model.generate(input_ids,
+                                max_new_tokens=args.n_predict)
+
+        # start inference
+        st = time.time()
+        # if your selected model is capable of utilizing previous key/value attentions
+        # to enhance decoding speed, but has `"use_cache": false` in its model config,
+        # it is important to set `use_cache=True` explicitly in the `generate` function
+        # to obtain optimal performance with BigDL-LLM INT4 optimizations
+        output = model.generate(input_ids,
+                                max_new_tokens=args.n_predict)
+        end = time.time()
+        output = output.cpu()
+        output_str = tokenizer.decode(output[0], skip_special_tokens=True)
+        print(f'Inference time: {end-st} s')
+        print('-'*20, 'Output', '-'*20)
+        print(output_str)

python/llm/example/CPU/PyTorch-Models/Model/mixtral/README.md

@@ -0,0 +1,45 @@
+# Mixtral
+In this directory, you will find examples on how you could use BigDL-LLM `optimize_model` API to accelerate Mixtral models. For illustration purposes, we utilize the [mistralai/Mixtral-8x7B-Instruct-v0.1](https://huggingface.co/mistralai/Mixtral-8x7B-Instruct-v0.1) as a reference Mixtral model.
+
+## Requirements
+To run these examples with BigDL-LLM on Intel CPUs, we have some recommended requirements for your machine, please refer to [here](../README.md#recommended-requirements) for more information.
+
+**Important: Please make sure you have installed `transformers==4.36.0` to run the example.**
+
+## Example: Predict Tokens using `generate()` API
+In the example [generate.py](./generate.py), we show a basic use case for a Mixtral model to predict the next N tokens using `generate()` API, with BigDL-LLM INT4 optimizations on Intel CPUs.
+### 1. Install
+We suggest using conda to manage the Python environment. For more information about conda installation, please refer to [here](https://docs.conda.io/en/latest/miniconda.html#).
+
+After installing conda, create a Python environment for BigDL-LLM:
+```bash
+conda create -n llm python=3.9 # recommend to use Python 3.9
+conda activate llm
+
+# below command will install PyTorch CPU as default
+pip install torch==2.0.1 --index-url https://download.pytorch.org/whl/cpu
+pip install --pre --upgrade bigdl-llm[all]
+
+# Please make sure you are using a stable version of Transformers, 4.36.0 or newer.
+pip install transformers==4.36.0
+```
+
+### 2. Run
+
+```bash
+python ./generate.py --prompt 'What is AI?'
+```
+
+In the example, several arguments can be passed to satisfy your requirements:
+
+- `--repo-id-or-model-path REPO_ID_OR_MODEL_PATH`: argument defining the huggingface repo id for the Mixtral model (e.g. `mistralai/Mixtral-8x7B-Instruct-v0.1`) to be downloaded, or the path to the huggingface checkpoint folder. It is default to be `'mistralai/Mixtral-8x7B-Instruct-v0.1'`.
+- `--prompt PROMPT`: argument defining the prompt to be infered (with integrated prompt format for chat). It is default to be `'What is AI?'`.
+- `--n-predict N_PREDICT`: argument defining the max number of tokens to predict. It is default to be `32`.
+
+#### Sample Output
+#### [mistralai/Mixtral-8x7B-Instruct-v0.1](https://huggingface.co/mistralai/Mixtral-8x7B-Instruct-v0.1)
+```log
+Inference time: xxxx s 
+-------------------- Output --------------------
+[INST] What is AI? [/INST] AI, or Artificial Intelligence, refers to the development of computer systems that can perform tasks that would normally require human intelligence to accomplish. These tasks can include things
+```

python/llm/example/CPU/PyTorch-Models/Model/mixtral/generate.py

@@ -0,0 +1,70 @@
+#
+# 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 torch
+import time
+import argparse
+
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from bigdl.llm import optimize_model
+
+# you could tune the prompt based on your own model,
+# here the prompt tuning refers to https://huggingface.co/mistralai/Mixtral-8x7B-Instruct-v0.1#instruction-format
+MIXTRAL_PROMPT_FORMAT = """<s>[INST] {prompt} [/INST]"""
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser(description='Predict Tokens using `generate()` API for Mixtral model')
+    parser.add_argument('--repo-id-or-model-path', type=str, default="'mistralai/Mixtral-8x7B-Instruct-v0.1'",
+                        help='The huggingface repo id for the Mixtral (e.g. `mistralai/Mixtral-8x7B-Instruct-v0.1`) to be downloaded,'
+                             ', or the path to the huggingface checkpoint folder.')
+    parser.add_argument('--prompt', type=str, default="What is AI?",
+                        help='Prompt to infer')
+    parser.add_argument('--n-predict', type=int, default=32,
+                        help='Max tokens to predict')
+
+    args = parser.parse_args()
+    model_path = args.repo_id_or_model_path
+
+    # Load model
+    model = AutoModelForCausalLM.from_pretrained(model_path,
+                                                 trust_remote_code=True,
+                                                 torch_dtype='auto',
+                                                 low_cpu_mem_usage=True)
+
+    # With only one line to enable BigDL-LLM optimization on model
+    model = optimize_model(model)
+
+    # Load tokenizer
+    tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
+    
+    # Generate predicted tokens
+    with torch.inference_mode():
+        prompt = MIXTRAL_PROMPT_FORMAT.format(prompt=args.prompt)
+        input_ids = tokenizer.encode(prompt, return_tensors="pt").to('cpu')
+        # ipex model needs a warmup, then inference time can be accurate
+        output = model.generate(input_ids,
+                                max_new_tokens=args.n_predict)
+
+        # start inference
+        st = time.time()
+        output = model.generate(input_ids,
+                                max_new_tokens=args.n_predict)
+        end = time.time()
+        output = output.cpu()
+        output_str = tokenizer.decode(output[0], skip_special_tokens=True)
+        print(f'Inference time: {end-st} s')
+        print('-'*20, 'Output', '-'*20)
+        print(output_str)