Initial llama3 example (#10799)

* Add initial hf huggingface GPU example

* Small fix

* Add llama3 gpu pytorch model example

* Add llama 3 hf transformers CPU example

* Add llama 3 pytorch model CPU example

* Fixes

* Small fix

* Small fixes

* Small fix

* Small fix

* Add links

* update repo id

* change prompt tuning url

* remove system header if there is no system prompt

---------

Co-authored-by: Yuwen Hu <yuwen.hu@intel.com>
Co-authored-by: Yuwen Hu <54161268+Oscilloscope98@users.noreply.github.com>

README.md

@@ -128,6 +128,7 @@ Over 50 models have been optimized/verified on `ipex-llm`, including *LLaMA/LLaM
 |------------|----------------------------------------------------------------|-----------------------------------------------------------------|
 | LLaMA *(such as Vicuna, Guanaco, Koala, Baize, WizardLM, etc.)* | [link1](python/llm/example/CPU/Native-Models), [link2](python/llm/example/CPU/HF-Transformers-AutoModels/Model/vicuna) |[link](python/llm/example/GPU/HF-Transformers-AutoModels/Model/vicuna)|
 | LLaMA 2    | [link1](python/llm/example/CPU/Native-Models), [link2](python/llm/example/CPU/HF-Transformers-AutoModels/Model/llama2) | [link](python/llm/example/GPU/HF-Transformers-AutoModels/Model/llama2)  |
+| LLaMA 3    | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/llama3) | [link](python/llm/example/GPU/HF-Transformers-AutoModels/Model/llama3)  |
 | ChatGLM    | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/chatglm)   |    | 
 | 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)   |

docs/readthedocs/source/index.rst

@@ -219,6 +219,13 @@ Verified Models
            <a href="https://github.com/intel-analytics/ipex-llm/tree/main/python/llm/example/GPU/HF-Transformers-AutoModels/Model/llama2">link</a>
            <a href="https://github.com/intel-analytics/ipex-llm/tree/main/python/llm/example/GPU/HF-Transformers-AutoModels/Model/llama2">link</a></td>
        </tr>
+       <tr>
+         <td>LLaMA 3</td>
+         <td>
+           <a href="https://github.com/intel-analytics/ipex-llm/tree/main/python/llm/example/CPU/HF-Transformers-AutoModels/Model/llama3">link</a></td>
+         <td>
+           <a href="https://github.com/intel-analytics/ipex-llm/tree/main/python/llm/example/GPU/HF-Transformers-AutoModels/Model/llama3">link</a></td>
+       </tr>
        <tr>
          <td>ChatGLM</td>
          <td>

python/llm/example/CPU/HF-Transformers-AutoModels/Model/llama3/README.md

@@ -0,0 +1,68 @@
+# Llama3
+In this directory, you will find examples on how you could apply IPEX-LLM INT4 optimizations on Llama3 models. For illustration purposes, we utilize the [meta-llama/Meta-Llama-3-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct) as a reference Llama3 model.
+
+## 0. 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 Llama3 model to predict the next N tokens using `generate()` API, with IPEX-LLM INT4 optimizations.
+### 1. Install
+We suggest using conda to manage environment:
+```bash
+conda create -n llm python=3.11
+conda activate llm
+
+pip install --pre --upgrade ipex-llm[all] # install ipex-llm with 'all' option
+
+# transformers>=4.33.0 is required for Llama3 with IPEX-LLM optimizations
+pip install transformers==4.37.0 
+```
+
+### 2. Run
+```
+python ./generate.py --repo-id-or-model-path REPO_ID_OR_MODEL_PATH --prompt PROMPT --n-predict N_PREDICT
+```
+
+Arguments info:
+- `--repo-id-or-model-path REPO_ID_OR_MODEL_PATH`: argument defining the huggingface repo id for the Llama3 model (e.g. `meta-llama/Meta-Llama-3-8B-Instruct`) to be downloaded, or the path to the huggingface checkpoint folder. It is default to be `'meta-llama/Meta-Llama-3-8B-Instruct'`.
+- `--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`.
+
+> **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 Llama3 model based on the capabilities of your machine.
+
+#### 2.1 Client
+On client Windows machine, it is recommended to run directly with full utilization of all cores:
+```powershell
+python ./generate.py 
+```
+
+#### 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
+```
+
+#### 2.3 Sample Output
+#### [meta-llama/Meta-Llama-3-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct)
+```log
+Inference time: xxxx s
+-------------------- Prompt --------------------
+<|begin_of_text|><|start_header_id|>user<|end_header_id|>
+What is AI?<|eot_id|><|start_header_id|>assistant<|end_header_id|>
+
+-------------------- Output (skip_special_tokens=False) --------------------
+<|begin_of_text|><|start_header_id|>user<|end_header_id|>
+What is AI?<|eot_id|><|start_header_id|>assistant<|end_header_id|>
+A question that gets to the heart of the 21st century!
+
+Artificial Intelligence (AI) refers to the development of computer systems that can perform tasks that
+```

python/llm/example/CPU/HF-Transformers-AutoModels/Model/llama3/generate.py

@@ -0,0 +1,81 @@
+#
+# 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://llama.meta.com/docs/model-cards-and-prompt-formats/meta-llama-3
+DEFAULT_SYSTEM_PROMPT = """\
+"""
+
+def get_prompt(user_input: str, chat_history: list[tuple[str, str]],
+               system_prompt: str) -> str:
+    prompt_texts = [f'<|begin_of_text|>']
+
+    if system_prompt != '':
+        prompt_texts.append(f'<|start_header_id|>system<|end_header_id|>\n{system_prompt}<|eot_id|>')
+
+    for history_input, history_response in chat_history:
+        prompt_texts.append(f'<|start_header_id|>user<|end_header_id|>\n{history_input.strip()}<|eot_id|>')
+        prompt_texts.append(f'<|start_header_id|>assistant<|end_header_id|>\n{history_response.strip()}<|eot_id|>')
+
+    prompt_texts.append(f'<|start_header_id|>user<|end_header_id|>\n{user_input.strip()}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n')
+    return ''.join(prompt_texts)
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser(description='Predict Tokens using `generate()` API for Llama3 model')
+    parser.add_argument('--repo-id-or-model-path', type=str, default="meta-llama/Meta-Llama-3-8B-Instruct",
+                        help='The huggingface repo id for the Llama3 (e.g. `meta-llama/Meta-Llama-3-8B-Instruct`) 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 = get_prompt(args.prompt, [], system_prompt=DEFAULT_SYSTEM_PROMPT)
+        input_ids = tokenizer.encode(prompt, return_tensors="pt")
+        st = time.time()
+        output = model.generate(input_ids,
+                                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 (skip_special_tokens=False)', '-'*20)
+        print(output_str)
+

python/llm/example/CPU/PyTorch-Models/Model/llama3/README.md

@@ -0,0 +1,68 @@
+# Llama3
+In this directory, you will find examples on how you could use IPEX-LLM `optimize_model` API to accelerate Llama3 models. For illustration purposes, we utilize the [meta-llama/Meta-Llama-3-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct) as a reference Llama3 model.
+
+## 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 Llama3 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
+
+# transformers>=4.33.0 is required for Llama3 with IPEX-LLM optimizations
+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 REPO_ID_OR_MODEL_PATH`: argument defining the huggingface repo id for the Llama3 model (e.g. `meta-llama/Meta-Llama-3-8B-Instruct`) to be downloaded, or the path to the huggingface checkpoint folder. It is default to be `'meta-llama/Meta-Llama-3-8B-Instruct'`.
+- `--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`.
+
+#### 2.3 Sample Output
+#### [meta-llama/Meta-Llama-3-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct)
+```log
+Inference time: xxxx s
+-------------------- Prompt --------------------
+<|begin_of_text|><|start_header_id|>user<|end_header_id|>
+What is AI?<|eot_id|><|start_header_id|>assistant<|end_header_id|>
+
+-------------------- Output (skip_special_tokens=False) --------------------
+<|begin_of_text|><|start_header_id|>user<|end_header_id|>
+What is AI?<|eot_id|><|start_header_id|>assistant<|end_header_id|>
+A question that gets to the heart of the 21st century!
+
+Artificial Intelligence (AI) refers to the development of computer systems that can perform tasks that
+```

python/llm/example/CPU/PyTorch-Models/Model/llama3/generate.py

@@ -0,0 +1,82 @@
+#
+# 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 ipex_llm import optimize_model
+
+# you could tune the prompt based on your own model,
+# here the prompt tuning refers to https://llama.meta.com/docs/model-cards-and-prompt-formats/meta-llama-3
+DEFAULT_SYSTEM_PROMPT = """\
+"""
+
+def get_prompt(user_input: str, chat_history: list[tuple[str, str]],
+               system_prompt: str) -> str:
+    prompt_texts = [f'<|begin_of_text|>']
+
+    if system_prompt != '':
+        prompt_texts.append(f'<|start_header_id|>system<|end_header_id|>\n{system_prompt}<|eot_id|>')
+
+    for history_input, history_response in chat_history:
+        prompt_texts.append(f'<|start_header_id|>user<|end_header_id|>\n{history_input.strip()}<|eot_id|>')
+        prompt_texts.append(f'<|start_header_id|>assistant<|end_header_id|>\n{history_response.strip()}<|eot_id|>')
+
+    prompt_texts.append(f'<|start_header_id|>user<|end_header_id|>\n{user_input.strip()}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n')
+    return ''.join(prompt_texts)
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser(description='Predict Tokens using `generate()` API for Llama3 model')
+    parser.add_argument('--repo-id-or-model-path', type=str, default="meta-llama/Meta-Llama-3-8B-Instruct",
+                        help='The huggingface repo id for the Llama3 (e.g. `meta-llama/Meta-Llama-3-8B-Instruct`) 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 = get_prompt(args.prompt, [], system_prompt=DEFAULT_SYSTEM_PROMPT)
+        input_ids = tokenizer.encode(prompt, return_tensors="pt")
+        st = time.time()
+        output = model.generate(input_ids,
+                                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 (skip_special_tokens=False)', '-'*20)
+        print(output_str)