Add example for phi-3

README.md

@@ -177,6 +177,7 @@ Over 50 models have been optimized/verified on `ipex-llm`, including *LLaMA/LLaM
 | DeepSeek-MoE | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/deepseek-moe) |  |
 | Ziya-Coding-34B-v1.0 | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/ziya) | |
 | Phi-2 | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/phi-2) | [link](python/llm/example/GPU/HF-Transformers-AutoModels/Model/phi-2) |
+| Phi-3 | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/phi-3) | [link](python/llm/example/GPU/HF-Transformers-AutoModels/Model/phi-3) |
 | Yuan2 | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/yuan2) | [link](python/llm/example/GPU/HF-Transformers-AutoModels/Model/yuan2) |
 | Gemma | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/gemma) | [link](python/llm/example/GPU/HF-Transformers-AutoModels/Model/gemma) |
 | DeciLM-7B | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/deciLM-7b) | [link](python/llm/example/GPU/HF-Transformers-AutoModels/Model/deciLM-7b) |

docs/readthedocs/source/index.rst

@@ -538,6 +538,13 @@ Verified Models
          <td>
            <a href="https://github.com/intel-analytics/ipex-llm/tree/main/python/llm/example/GPU/HF-Transformers-AutoModels/Model/phi-2">link</a></td>
        </tr>
+       <tr>
+         <td>Phi-3</td>
+         <td>
+           <a href="https://github.com/intel-analytics/ipex-llm/tree/main/python/llm/example/CPU/HF-Transformers-AutoModels/Model/phi-3">link</a></td>
+         <td>
+           <a href="https://github.com/intel-analytics/ipex-llm/tree/main/python/llm/example/GPU/HF-Transformers-AutoModels/Model/phi-3">link</a></td>
+       </tr>
        <tr>
          <td>Yuan2</td>
          <td>

python/llm/example/CPU/HF-Transformers-AutoModels/Model/phi-3/README.md

@@ -0,0 +1,71 @@
+# phi-3
+
+In this directory, you will find examples on how you could apply IPEX-LLM INT4 optimizations on phi-3 models. For illustration purposes, we utilize the [microsoft/Phi-3-mini-4k-instruct](https://huggingface.co/microsoft/Phi-3-mini-4k-instruct) as a reference phi-3 model.
+
+> **Note**: If you want to download the Hugging Face *Transformers* model, please refer to [here](https://huggingface.co/docs/hub/models-downloading#using-git).
+>
+> IPEX-LLM optimizes the *Transformers* model in INT4 precision at runtime, and thus no explicit conversion is needed.
+
+## Requirements
+To run these examples with IPEX-LLM, we have some recommended requirements for your machine, please refer to [here](../README.md#recommended-requirements) for more information.
+
+## Example: Predict Tokens using `generate()` API
+In the example [generate.py](./generate.py), we show a basic use case for a phi-3 model to predict the next N tokens using `generate()` API, with IPEX-LLM INT4 optimizations.
+### 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 IPEX-LLM:
+```bash
+conda create -n llm python=3.11 # recommend to use Python 3.11
+conda activate llm
+
+pip install --pre --upgrade ipex-llm[all] # install the latest ipex-llm nightly build with 'all' option
+
+pip install transformers==4.37.0
+```
+
+### 2. Run
+After setting up the Python environment, you could run the example by following steps.
+
+> **Note**: When loading the model in 4-bit, IPEX-LLM converts linear layers in the model into INT4 format. In theory, a *X*B model saved in 16-bit will requires approximately 2*X* GB of memory for loading, and ~0.5*X* GB memory for further inference.
+>
+> Please select the appropriate size of the phi-3 model based on the capabilities of your machine.
+
+#### 2.1 Client
+On client Windows machines, it is recommended to run directly with full utilization of all cores:
+```powershell
+python ./generate.py --prompt 'What is AI?'
+```
+More information about arguments can be found in [Arguments Info](#23-arguments-info) section. The expected output can be found in [Sample Output](#24-sample-output) section.
+
+#### 2.2 Server
+For optimal performance on server, it is recommended to set several environment variables (refer to [here](../README.md#best-known-configuration-on-linux) for more information), and run the example with all the physical cores of a single socket.
+
+E.g. on Linux,
+```bash
+# set IPEX-LLM env variables
+source ipex-llm-init
+
+# e.g. for a server with 48 cores per socket
+export OMP_NUM_THREADS=48
+numactl -C 0-47 -m 0 python ./generate.py --prompt 'What is AI？'
+```
+More information about arguments can be found in [Arguments Info](#23-arguments-info) section. The expected output can be found in [Sample Output](#24-sample-output) section.
+
+#### 2.3 Arguments Info
+In the example, several arguments can be passed to satisfy your requirements:
+
+- `--repo-id-or-model-path`: str, argument defining the huggingface repo id for the phi-3 model to be downloaded, or the path to the huggingface checkpoint folder. It is default to be `'microsoft/Phi-3-mini-4k-instruct'`.
+- `--prompt`: str, argument defining the prompt to be inferred (with integrated prompt format for chat). It is default to be `What is AI?`.
+- `--n-predict`: int, argument defining the max number of tokens to predict. It is default to be `32`.
+
+#### 2.4 Sample Output
+#### [microsoft/Phi-3-mini-4k-instruct](https://huggingface.co/microsoft/Phi-3-mini-4k-instruct)
+```log
+-------------------- Prompt --------------------
+<|user|>
+What is AI?<|end|>
+<|assistant|>
+-------------------- Output --------------------
+<s><|user|> What is AI?<|end|><|assistant|> AI, or Artificial Intelligence, refers to the simulation of human intelligence in machines that are programmed to think and learn like humans. The goal
+```

python/llm/example/CPU/HF-Transformers-AutoModels/Model/phi-3/generate.py

@@ -0,0 +1,68 @@
+#
+# 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 ipex_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/microsoft/Phi-3-mini-4k-instruct#chat-format
+PHI3_PROMPT_FORMAT = "<|user|>\n{prompt}<|end|>\n<|assistant|>"
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser(description='Predict Tokens using `generate()` API for phi-3 model')
+    parser.add_argument('--repo-id-or-model-path', type=str, default="microsoft/Phi-3-mini-4k-instruct",
+                        help='The huggingface repo id for the phi-3 model 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 = PHI3_PROMPT_FORMAT.format(prompt=args.prompt)
+        input_ids = tokenizer.encode(prompt, return_tensors="pt")
+        st = time.time()
+
+        output = model.generate(input_ids,
+                                do_sample=False,
+                                max_new_tokens=args.n_predict)
+        end = time.time()
+        output_str = tokenizer.decode(output[0], skip_special_tokens=False)
+        print(f'Inference time: {end-st} s')
+        print('-'*20, 'Prompt', '-'*20)
+        print(prompt)
+        print('-'*20, 'Output', '-'*20)
+        print(output_str)

python/llm/example/CPU/PyTorch-Models/Model/phi-3/README.md

@@ -0,0 +1,67 @@
+# phi-3
+
+In this directory, you will find examples on how you could apply IPEX-LLM INT4 optimizations on phi-3 models. For illustration purposes, we utilize the [microsoft/Phi-3-mini-4k-instruct](https://huggingface.co/microsoft/Phi-3-mini-4k-instruct) as a reference phi-3 model.
+
+> **Note**: If you want to download the Hugging Face *Transformers* model, please refer to [here](https://huggingface.co/docs/hub/models-downloading#using-git).
+>
+> IPEX-LLM optimizes the *Transformers* model in INT4 precision at runtime, and thus no explicit conversion is needed.
+
+## Requirements
+To run these examples with IPEX-LLM, we have some recommended requirements for your machine, please refer to [here](../README.md#recommended-requirements) for more information.
+
+## Example: Predict Tokens using `generate()` API
+In the example [generate.py](./generate.py), we show a basic use case for a phi-3 model to predict the next N tokens using `generate()` API, with IPEX-LLM INT4 optimizations.
+### 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 IPEX-LLM:
+```bash
+conda create -n llm python=3.11 # recommend to use Python 3.11
+conda activate llm
+
+pip install --pre --upgrade ipex-llm[all] # install the latest ipex-llm nightly build with 'all' option
+
+pip install transformers==4.37.0
+```
+
+### 2. Run
+After setting up the Python environment, you could run the example by following steps.
+
+#### 2.1 Client
+On client Windows machines, it is recommended to run directly with full utilization of all cores:
+```powershell
+python ./generate.py --prompt 'What is AI?'
+```
+More information about arguments can be found in [Arguments Info](#23-arguments-info) section. The expected output can be found in [Sample Output](#24-sample-output) section.
+
+#### 2.2 Server
+For optimal performance on server, it is recommended to set several environment variables (refer to [here](../README.md#best-known-configuration-on-linux) for more information), and run the example with all the physical cores of a single socket.
+
+E.g. on Linux,
+```bash
+# set IPEX-LLM env variables
+source ipex-llm-init
+
+# e.g. for a server with 48 cores per socket
+export OMP_NUM_THREADS=48
+numactl -C 0-47 -m 0 python ./generate.py --prompt 'What is AI？'
+```
+More information about arguments can be found in [Arguments Info](#23-arguments-info) section. The expected output can be found in [Sample Output](#24-sample-output) section.
+
+#### 2.3 Arguments Info
+In the example, several arguments can be passed to satisfy your requirements:
+
+- `--repo-id-or-model-path`: str, argument defining the huggingface repo id for the phi-3 model to be downloaded, or the path to the huggingface checkpoint folder. It is default to be `'microsoft/Phi-3-mini-4k-instruct'`.
+- `--prompt`: str, argument defining the prompt to be inferred (with integrated prompt format for chat). It is default to be `What is AI?`.
+- `--n-predict`: int, argument defining the max number of tokens to predict. It is default to be `32`.
+
+#### 2.4 Sample Output
+#### [microsoft/Phi-3-mini-4k-instruct](https://huggingface.co/microsoft/Phi-3-mini-4k-instruct)
+```log
+-------------------- Prompt --------------------
+<|user|>
+What is AI?<|end|>
+<|assistant|>
+-------------------- Output --------------------
+<s><|user|> What is AI?<|end|><|assistant|> AI, or Artificial Intelligence, refers to the simulation of human intelligence in machines that are programmed to think and learn like humans. The goal
+```

python/llm/example/CPU/PyTorch-Models/Model/phi-3/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 AutoTokenizer, AutoModelForCausalLM
+from ipex_llm import optimize_model
+
+# you could tune the prompt based on your own model,
+# here the prompt tuning refers to https://huggingface.co/microsoft/Phi-3-mini-4k-instruct#chat-format
+PHI3_PROMPT_FORMAT = "<|user|>\n{prompt}<|end|>\n<|assistant|>"
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser(description='Predict Tokens using `generate()` API for phi-3 model')
+    parser.add_argument('--repo-id-or-model-path', type=str, default="microsoft/Phi-3-mini-4k-instruct",
+                        help='The huggingface repo id for the phi-3 model 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,
+                                                 use_cache=True)
+
+    # With only one line to enable IPEX-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 = PHI3_PROMPT_FORMAT.format(prompt=args.prompt)
+        input_ids = tokenizer.encode(prompt, return_tensors="pt")
+        st = time.time()
+
+        output = model.generate(input_ids,
+                                do_sample=False,
+                                max_new_tokens=args.n_predict)
+        end = time.time()
+        output_str = tokenizer.decode(output[0], skip_special_tokens=False)
+        print(f'Inference time: {end-st} s')
+        print('-'*20, 'Prompt', '-'*20)
+        print(prompt)
+        print('-'*20, 'Output', '-'*20)
+        print(output_str)