Skip to content

Commit

Permalink
Add example for phi-3 (#10881)
Browse files Browse the repository at this point in the history
* Add example for phi-3

* add in readme and index

* fix

* fix

* fix

* fix indent

* fix
  • Loading branch information
JinBridger authored Apr 29, 2024
1 parent c936ba3 commit 1f876fd
Show file tree
Hide file tree
Showing 10 changed files with 703 additions and 0 deletions.
1 change: 1 addition & 0 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -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) |
Expand Down
7 changes: 7 additions & 0 deletions docs/readthedocs/source/index.rst
Original file line number Diff line number Diff line change
Expand Up @@ -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>
Expand Down
Original file line number Diff line number Diff line change
@@ -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
```
Original file line number Diff line number Diff line change
@@ -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)
67 changes: 67 additions & 0 deletions python/llm/example/CPU/PyTorch-Models/Model/phi-3/README.md
Original file line number Diff line number Diff line change
@@ -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
```
70 changes: 70 additions & 0 deletions python/llm/example/CPU/PyTorch-Models/Model/phi-3/generate.py
Original file line number Diff line number Diff line change
@@ -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)
Loading

0 comments on commit 1f876fd

Please sign in to comment.