diff --git a/README.md b/README.md index d7e1876f..4f4765fb 100644 --- a/README.md +++ b/README.md @@ -193,6 +193,7 @@ Over 40 models have been optimized/verified on `bigdl-llm`, including *LLaMA/LLa | Phi-2 | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/phi-2) | [link](python/llm/example/GPU/HF-Transformers-AutoModels/Model/phi-2) | | 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) | ***For more details, please refer to the `bigdl-llm` [Document](https://test-bigdl-llm.readthedocs.io/en/main/doc/LLM/index.html), [Readme](python/llm), [Tutorial](https://github.com/intel-analytics/bigdl-llm-tutorial) and [API Doc](https://bigdl.readthedocs.io/en/latest/doc/PythonAPI/LLM/index.html).*** diff --git a/python/llm/README.md b/python/llm/README.md index 8cbaf5df..881c09f0 100644 --- a/python/llm/README.md +++ b/python/llm/README.md @@ -84,7 +84,7 @@ Over 20 models have been optimized/verified on `bigdl-llm`, including *LLaMA/LLa | Ziya-Coding-34B-v1.0 | [link](example/CPU/HF-Transformers-AutoModels/Model/ziya) | | | Phi-2 | [link](example/CPU/HF-Transformers-AutoModels/Model/phi-2) | [link](example/GPU/HF-Transformers-AutoModels/Model/phi-2) | | Yuan2 | [link](example/CPU/HF-Transformers-AutoModels/Model/yuan2) | [link](example/GPU/HF-Transformers-AutoModels/Model/yuan2) | - +| DeciLM-7B | [link](example/CPU/HF-Transformers-AutoModels/Model/deciLM-7b) | [link](example/GPU/HF-Transformers-AutoModels/Model/deciLM-7b) | ### Working with `bigdl-llm`
Table of Contents diff --git a/python/llm/example/CPU/HF-Transformers-AutoModels/Model/deciLM-7b/README.md b/python/llm/example/CPU/HF-Transformers-AutoModels/Model/deciLM-7b/README.md new file mode 100644 index 00000000..fbf91242 --- /dev/null +++ b/python/llm/example/CPU/HF-Transformers-AutoModels/Model/deciLM-7b/README.md @@ -0,0 +1,69 @@ +# DeciLM-7B +In this directory, you will find examples on how you could apply BigDL-LLM INT4 optimizations on DeciLM-7B models. For illustration purposes, we utilize the [Deci/DeciLM-7B-instruct](https://huggingface.co/Deci/DeciLM-7B-instruct) as a reference DeciLM-7B model. + +## 0. Requirements +To run these examples with BigDL-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 DeciLM-7B model to predict the next N tokens using `generate()` API, with BigDL-LLM INT4 optimizations. +### 1. Install +We suggest using conda to manage environment: +```bash +conda create -n llm python=3.9 +conda activate llm + +pip install --pre --upgrade bigdl-llm[all] # install the latest bigdl-llm nightly build with 'all' option +pip install transformers==4.35.2 # required by DeciLM-7B +``` + +### 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 DeciLM-7B model to be downloaded, or the path to the huggingface checkpoint folder. It is default to be `'Deci/DeciLM-7B-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, BigDL-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 DeciLM-7B 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 BigDL-LLM env variables +source bigdl-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 +#### [Deci/DeciLM-7B-instruct](https://huggingface.co/Deci/DeciLM-7B-instruct) +```log +Inference time: XXXX s +-------------------- Prompt -------------------- +### System: +You are an AI assistant that follows instruction extremely well. Help as much as you can. +### User: +What is AI? +### Assistant: +-------------------- Output -------------------- +### System: +You are an AI assistant that follows instruction extremely well. Help as much as you can. +### User: +What is AI? +### Assistant: + AI stands for Artificial Intelligence, which refers to the development of computer systems and software that can perform tasks that typically require human intelligence, such as recognizing patterns +``` \ No newline at end of file diff --git a/python/llm/example/CPU/HF-Transformers-AutoModels/Model/deciLM-7b/generate.py b/python/llm/example/CPU/HF-Transformers-AutoModels/Model/deciLM-7b/generate.py new file mode 100644 index 00000000..6f5ede1f --- /dev/null +++ b/python/llm/example/CPU/HF-Transformers-AutoModels/Model/deciLM-7b/generate.py @@ -0,0 +1,71 @@ +# +# 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 +from bigdl.llm.transformers import AutoModelForCausalLM + +# you could tune the prompt based on your own model, +# here the prompt tuning refers to https://huggingface.co/Deci/DeciLM-7B-instruct#prompt-template +SYSTEM_PROMPT_TEMPLATE =""" +### System: +You are an AI assistant that follows instruction extremely well. Help as much as you can. +### User: +{prompt} +### Assistant: +""" + +if __name__ == '__main__': + parser = argparse.ArgumentParser(description='Predict Tokens using `generate()` API for DeciLM-7B model') + parser.add_argument('--repo-id-or-model-path', type=str, default="Deci/DeciLM-7B-instruct", + help='The huggingface repo id for the DeciLM-7B (e.g. `Deci/DeciLM-7B-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, + load_in_4bit=True, + trust_remote_code=True, + ) + + # Load tokenizer + tokenizer = AutoTokenizer.from_pretrained(model_path) + tokenizer.pad_token = tokenizer.eos_token + + # Generate predicted tokens + with torch.inference_mode(): + prompt = SYSTEM_PROMPT_TEMPLATE.format(prompt=args.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=True) + print(f'Inference time: {end-st} s') + print(prompt) + print('-'*20, 'Output', '-'*20) + print(output_str) + diff --git a/python/llm/example/CPU/PyTorch-Models/Model/deciLM-7b/README.md b/python/llm/example/CPU/PyTorch-Models/Model/deciLM-7b/README.md new file mode 100644 index 00000000..0731a9b9 --- /dev/null +++ b/python/llm/example/CPU/PyTorch-Models/Model/deciLM-7b/README.md @@ -0,0 +1,69 @@ +# DeciLM-7B +In this directory, you will find examples on how you could use BigDL-LLM `optimize_model` API to accelerate DeciLM-7B models. For illustration purposes, we utilize the [Deci/DeciLM-7B-instruct](https://huggingface.co/Deci/DeciLM-7B-instruct) as a reference DeciLM-7B model. + +## Requirements +To run these examples with BigDL-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 DeciLM-7B model to predict the next N tokens using `generate()` API, with BigDL-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 BigDL-LLM: +```bash +conda create -n llm python=3.9 # recommend to use Python 3.9 +conda activate llm + +pip install --pre --upgrade bigdl-llm[all] # install the latest bigdl-llm nightly build with 'all' option +pip install transformers==4.35.2 # required by DeciLM-7B +``` + +### 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 --repo-id-or-model-path REPO_ID_OR_MODEL_PATH --prompt PROMPT --n-predict N_PREDICT +``` +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 BigDL-LLM env variables +source bigdl-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 DeciLM-7B model to be downloaded, or the path to the huggingface checkpoint folder. It is default to be `'Deci/DeciLM-7B-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 +#### [Deci/DeciLM-7B-instruct](https://huggingface.co/Deci/DeciLM-7B-instruct) +```log +Inference time: XXXX s +-------------------- Prompt -------------------- +### System: +You are an AI assistant that follows instruction extremely well. Help as much as you can. +### User: +What is AI? +### Assistant: +-------------------- Output -------------------- +### System: +You are an AI assistant that follows instruction extremely well. Help as much as you can. +### User: +What is AI? +### Assistant: + AI stands for Artificial Intelligence, which refers to the development of computer systems and software that can perform tasks that typically require human intelligence, such as recognizing patterns +``` \ No newline at end of file diff --git a/python/llm/example/CPU/PyTorch-Models/Model/deciLM-7b/generate.py b/python/llm/example/CPU/PyTorch-Models/Model/deciLM-7b/generate.py new file mode 100644 index 00000000..2f0cdc72 --- /dev/null +++ b/python/llm/example/CPU/PyTorch-Models/Model/deciLM-7b/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 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/Deci/DeciLM-7B-instruct#prompt-template +SYSTEM_PROMPT_TEMPLATE =""" +### System: +You are an AI assistant that follows instruction extremely well. Help as much as you can. +### User: +{prompt} +### Assistant: +""" + +if __name__ == '__main__': + parser = argparse.ArgumentParser(description='Predict Tokens using `generate()` API for DeciLM-7B model') + parser.add_argument('--repo-id-or-model-path', type=str, default="Deci/DeciLM-7B-instruct", + help='The huggingface repo id for the DeciLM-7B (e.g. `Deci/DeciLM-7B-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, + ) + + # With only one line to enable BigDL-LLM optimization on model + model = optimize_model(model) + + # Load tokenizer + tokenizer = AutoTokenizer.from_pretrained(model_path) + tokenizer.pad_token = tokenizer.eos_token + + # Generate predicted tokens + with torch.inference_mode(): + prompt = SYSTEM_PROMPT_TEMPLATE.format(prompt=args.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=True) + print(f'Inference time: {end-st} s') + print('-'*20, 'Output', '-'*20) + print(output_str) + diff --git a/python/llm/example/GPU/HF-Transformers-AutoModels/Model/deciLM-7b/README.md b/python/llm/example/GPU/HF-Transformers-AutoModels/Model/deciLM-7b/README.md new file mode 100644 index 00000000..337398c0 --- /dev/null +++ b/python/llm/example/GPU/HF-Transformers-AutoModels/Model/deciLM-7b/README.md @@ -0,0 +1,130 @@ +# DeciLM-7B +In this directory, you will find examples on how you could apply BigDL-LLM INT4 optimizations on DeciLM-7B models on [Intel GPUs](../README.md). For illustration purposes, we utilize the [Deci/DeciLM-7B-instruct](https://huggingface.co/Deci/DeciLM-7B-instruct) as a reference DeciLM-7B model. + +## Requirements +To run these examples with BigDL-LLM on Intel GPUs, 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 DeciLM-7B model to predict the next N tokens using `generate()` API, with BigDL-LLM INT4 optimizations on Intel GPUs. +### 1. Install +#### 1.1 Installation on Linux +We suggest using conda to manage 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 +conda activate llm +# below command will install intel_extension_for_pytorch==2.0.110+xpu as default +# you can install specific ipex/torch version for your need +pip install --pre --upgrade bigdl-llm[xpu] -f https://developer.intel.com/ipex-whl-stable-xpu +pip install transformers==4.35.2 # required by DeciLM-7B +``` +#### 1.2 Installation on Windows +We suggest using conda to manage environment: +```bash +conda create -n llm python=3.9 libuv +conda activate llm +# below command will install intel_extension_for_pytorch==2.1.10+xpu as default +pip install --pre --upgrade bigdl-llm[xpu] -f https://developer.intel.com/ipex-whl-stable-xpu +pip install transformers==4.35.2 # required by DeciLM-7B +``` + +### 2. Configures OneAPI environment variables +#### 2.1 Configurations for Linux +```bash +source /opt/intel/oneapi/setvars.sh +``` +#### 2.2 Configurations for Windows +```cmd +call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat" +``` +> Note: Please make sure you are using **CMD** (**Anaconda Prompt** if using conda) to run the command as PowerShell is not supported. +### 3. Runtime Configurations +For optimal performance, it is recommended to set several environment variables. Please check out the suggestions based on your device. +#### 3.1 Configurations for Linux +
+ +For Intel Arc™ A-Series Graphics and Intel Data Center GPU Flex Series + +```bash +export USE_XETLA=OFF +export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1 +``` + +
+ +
+ +For Intel Data Center GPU Max Series + +```bash +export LD_PRELOAD=${LD_PRELOAD}:${CONDA_PREFIX}/lib/libtcmalloc.so +export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1 +export ENABLE_SDP_FUSION=1 +``` +> Note: Please note that `libtcmalloc.so` can be installed by `conda install -c conda-forge -y gperftools=2.10`. +
+ +#### 3.2 Configurations for Windows +
+ +For Intel iGPU + +```cmd +set SYCL_CACHE_PERSISTENT=1 +set BIGDL_LLM_XMX_DISABLED=1 +``` + +
+ +
+ +For Intel Arc™ A300-Series or Pro A60 + +```cmd +set SYCL_CACHE_PERSISTENT=1 +``` + +
+ +
+ +For other Intel dGPU Series + +There is no need to set further environment variables. + +
+ +> Note: For the first time that each model runs on Intel iGPU/Intel Arc™ A300-Series or Pro A60, it may take several minutes to compile. + +### 4. Running examples + +``` +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 DeciLM-7B model (e.g `Deci/DeciLM-7B-instruct`) to be downloaded, or the path to the huggingface checkpoint folder. It is default to be `'Deci/DeciLM-7B-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`. + +#### Sample Output +#### [Deci/DeciLM-7B-instruct](https://huggingface.co/Deci/DeciLM-7B-instruct) + +```log +Inference time: XXXX s +-------------------- Prompt -------------------- +### System: +You are an AI assistant that follows instruction extremely well. Help as much as you can. +### User: +What is AI? +### Assistant: +-------------------- Output -------------------- +### System: +You are an AI assistant that follows instruction extremely well. Help as much as you can. +### User: +What is AI? +### Assistant: + AI stands for Artificial Intelligence, which refers to the development of computer systems and software that can perform tasks that typically require human intelligence, such as recognizing patterns +``` diff --git a/python/llm/example/GPU/HF-Transformers-AutoModels/Model/deciLM-7b/generate.py b/python/llm/example/GPU/HF-Transformers-AutoModels/Model/deciLM-7b/generate.py new file mode 100644 index 00000000..b9a63832 --- /dev/null +++ b/python/llm/example/GPU/HF-Transformers-AutoModels/Model/deciLM-7b/generate.py @@ -0,0 +1,78 @@ +# +# 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 +from bigdl.llm.transformers import AutoModelForCausalLM + +# you could tune the prompt based on your own model, +# here the prompt tuning refers to https://huggingface.co/Deci/DeciLM-7B-instruct#prompt-template +SYSTEM_PROMPT_TEMPLATE =""" +### System: +You are an AI assistant that follows instruction extremely well. Help as much as you can. +### User: +{prompt} +### Assistant: +""" + +if __name__ == '__main__': + parser = argparse.ArgumentParser(description='Predict Tokens using `generate()` API for DeciLM-7B model') + parser.add_argument('--repo-id-or-model-path', type=str, default="Deci/DeciLM-7B-instruct", + help='The huggingface repo id for the DeciLM-7B (e.g. `Deci/DeciLM-7B-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 + # When running LLMs on Intel iGPUs for Windows users, we recommend setting `cpu_embedding=True` in the from_pretrained function. + # This will allow the memory-intensive embedding layer to utilize the CPU instead of iGPU. + model = AutoModelForCausalLM.from_pretrained( + model_path, + load_in_4bit=True, + trust_remote_code=True, + cpu_embedding=True + ) + + # With only one line to enable BigDL-LLM optimization on model + model = model.to('xpu') + # Load tokenizer + tokenizer = AutoTokenizer.from_pretrained(model_path) + tokenizer.pad_token = tokenizer.eos_token + + # Generate predicted tokens + with torch.inference_mode(): + prompt = SYSTEM_PROMPT_TEMPLATE.format(prompt=args.prompt) + input_ids = tokenizer.encode(prompt, return_tensors="pt").to('xpu') + st = time.time() + output = model.generate(input_ids, + max_new_tokens=args.n_predict) + torch.xpu.synchronize() + 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) + diff --git a/python/llm/example/GPU/PyTorch-Models/Model/deciLM-7b/README.md b/python/llm/example/GPU/PyTorch-Models/Model/deciLM-7b/README.md new file mode 100644 index 00000000..4c3ff88e --- /dev/null +++ b/python/llm/example/GPU/PyTorch-Models/Model/deciLM-7b/README.md @@ -0,0 +1,132 @@ +# DeciLM-7B +In this directory, you will find examples on how you could use BigDL-LLM `optimize_model` API to accelerate DeciLM-7B models. For illustration purposes, we utilize the [Deci/DeciLM-7B-instruct](https://huggingface.co/Deci/DeciLM-7B-instruct) as a reference DeciLM-7B model. + +## Requirements +To run these examples with BigDL-LLM on Intel GPUs, 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 DeciLM-7B model to predict the next N tokens using `generate()` API, with BigDL-LLM INT4 optimizations on Intel GPUs. +### 1. Install +#### 1.1 Installation on Linux +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 intel_extension_for_pytorch==2.0.110+xpu as default +# you can install specific ipex/torch version for your need +pip install --pre --upgrade bigdl-llm[xpu] -f https://developer.intel.com/ipex-whl-stable-xpu +pip install transformers==4.35.2 # required by DeciLM-7B +``` + +#### 1.2 Installation on Windows +We suggest using conda to manage environment: +```bash +conda create -n llm python=3.9 libuv +conda activate llm +# below command will install intel_extension_for_pytorch==2.1.10+xpu as default +pip install --pre --upgrade bigdl-llm[xpu] -f https://developer.intel.com/ipex-whl-stable-xpu +``` + +### 2. Configures OneAPI environment variables +#### 2.1 Configurations for Linux +```bash +source /opt/intel/oneapi/setvars.sh +``` +#### 2.2 Configurations for Windows +```cmd +call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat" +``` +> Note: Please make sure you are using **CMD** (**Anaconda Prompt** if using conda) to run the command as PowerShell is not supported. +### 3. Runtime Configurations +For optimal performance, it is recommended to set several environment variables. Please check out the suggestions based on your device. +#### 3.1 Configurations for Linux +
+ +For Intel Arc™ A-Series Graphics and Intel Data Center GPU Flex Series + +```bash +export USE_XETLA=OFF +export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1 +``` + +
+ +
+ +For Intel Data Center GPU Max Series + +```bash +export LD_PRELOAD=${LD_PRELOAD}:${CONDA_PREFIX}/lib/libtcmalloc.so +export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1 +export ENABLE_SDP_FUSION=1 +``` +> Note: Please note that `libtcmalloc.so` can be installed by `conda install -c conda-forge -y gperftools=2.10`. +
+ +#### 3.2 Configurations for Windows +
+ +For Intel iGPU + +```cmd +set SYCL_CACHE_PERSISTENT=1 +set BIGDL_LLM_XMX_DISABLED=1 +``` + +
+ +
+ +For Intel Arc™ A300-Series or Pro A60 + +```cmd +set SYCL_CACHE_PERSISTENT=1 +``` + +
+ +
+ +For other Intel dGPU Series + +There is no need to set further environment variables. + +
+ +> Note: For the first time that each model runs on Intel iGPU/Intel Arc™ A300-Series or Pro A60, it may take several minutes to compile. + +### 4. Running examples + + +```bash +python ./generate.py --repo-id-or-model-path REPO_ID_OR_MODEL_PATH --prompt PROMPT --n-predict N_PREDICT +``` + +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 DeciLM-7B model (e.g `Deci/DeciLM-7B-instruct`) to be downloaded, or the path to the huggingface checkpoint folder. It is default to be `'Deci/DeciLM-7B-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`. + +#### Sample Output +#### [Deci/DeciLM-7B-instruct](https://huggingface.co/Deci/DeciLM-7B-instruct) +```log +Inference time: XXXX s +-------------------- Prompt -------------------- +### System: +You are an AI assistant that follows instruction extremely well. Help as much as you can. +### User: +What is AI? +### Assistant: +-------------------- Output -------------------- +### System: +You are an AI assistant that follows instruction extremely well. Help as much as you can. +### User: +What is AI? +### Assistant: + AI stands for Artificial Intelligence, which refers to the development of computer systems and software that can perform tasks that typically require human intelligence, such as recognizing patterns +``` diff --git a/python/llm/example/GPU/PyTorch-Models/Model/deciLM-7b/generate.py b/python/llm/example/GPU/PyTorch-Models/Model/deciLM-7b/generate.py new file mode 100644 index 00000000..d2f32ce7 --- /dev/null +++ b/python/llm/example/GPU/PyTorch-Models/Model/deciLM-7b/generate.py @@ -0,0 +1,79 @@ +# +# 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/Deci/DeciLM-7B-instruct#prompt-template +SYSTEM_PROMPT_TEMPLATE =""" +### System: +You are an AI assistant that follows instruction extremely well. Help as much as you can. +### User: +{prompt} +### Assistant: +""" + +if __name__ == '__main__': + parser = argparse.ArgumentParser(description='Predict Tokens using `generate()` API for DeciLM-7B model') + parser.add_argument('--repo-id-or-model-path', type=str, default="Deci/DeciLM-7B-instruct", + help='The huggingface repo id for the DeciLM-7B (e.g. `Deci/DeciLM-7B-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, + ) + + # With only one line to enable BigDL-LLM optimization on model + # When running LLMs on Intel iGPUs for Windows users, we recommend setting `cpu_embedding=True` in the from_pretrained function. + # This will allow the memory-intensive embedding layer to utilize the CPU instead of iGPU. + model = optimize_model( + model, + cpu_embedding=True + ) + model = model.to('xpu') + # Load tokenizer + tokenizer = AutoTokenizer.from_pretrained(model_path) + tokenizer.pad_token = tokenizer.eos_token + + # Generate predicted tokens + with torch.inference_mode(): + prompt = SYSTEM_PROMPT_TEMPLATE.format(prompt=args.prompt) + input_ids = tokenizer.encode(prompt, return_tensors="pt").to('xpu') + st = time.time() + output = model.generate(input_ids, + max_new_tokens=args.n_predict) + torch.xpu.synchronize() + 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) +