-
Notifications
You must be signed in to change notification settings - Fork 1.3k
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
* ziya on CPU * add README for ziya * specify use_cache * add arc CPU * update prompt format * update link * add comments to emphasize use_cache * update pip cmd
- Loading branch information
Showing
6 changed files
with
323 additions
and
1 deletion.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
88 changes: 88 additions & 0 deletions
88
python/llm/example/CPU/HF-Transformers-AutoModels/Model/ziya/README.md
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,88 @@ | ||
# Ziya | ||
|
||
In this directory, you will find examples on how you could apply BigDL-LLM INT4 optimizations on Ziya models. For illustration purposes, we utilize the [IDEA-CCNL/Ziya-Coding-34B-v1.0](https://huggingface.co/IDEA-CCNL/Ziya-Coding-34B-v1.0) as a reference Ziya 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). | ||
> | ||
> BigDL-LLM optimizes the *Transformers* model in INT4 precision at runtime, and thus no explicit conversion is needed. | ||
## 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 Ziya 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 einops # additional package required for Ziya to conduct generation | ||
``` | ||
|
||
### 2. Run | ||
After setting up the Python environment, you could run the example by following steps. | ||
|
||
> **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 Ziya 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 'def quick_sort(arr):\n' | ||
``` | ||
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 'def quick_sort(arr):\n' | ||
``` | ||
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 Ziya model to be downloaded, or the path to the huggingface checkpoint folder. It is default to be `'IDEA-CCNL/Ziya-Coding-34B-v1.0'`. | ||
- `--prompt`: str, argument defining the prompt to be inferred (with integrated prompt format for chat). It is default to be `def quick_sort(arr):\n`. | ||
- `--n-predict`: int, argument defining the max number of tokens to predict. It is default to be `128`. | ||
|
||
#### 2.4 Sample Output | ||
#### [IDEA-CCNL/Ziya-Coding-34B-v1.0](https://huggingface.co/IDEA-CCNL/Ziya-Coding-34B-v1.0) | ||
```log | ||
Inference time: xxxx s | ||
-------------------- Prompt -------------------- | ||
<human>: | ||
def quick_sort(arr):\n | ||
<bot>: | ||
-------------------- Output -------------------- | ||
<s> <human>: | ||
def quick_sort(arr):\n | ||
<bot>: | ||
def partition(arr, low, high): | ||
i = (low-1) | ||
pivot = arr[high] | ||
for j in range(low, high): | ||
if arr[j] <= pivot: | ||
arr[i], arr[j] = arr[j], arr[i] | ||
i = i+1 | ||
arr[i], arr[high] = arr[high], arr[i] | ||
return i | ||
def quick_sort(arr, low, high): | ||
if low < high: | ||
pi = partition(arr, low, | ||
``` |
77 changes: 77 additions & 0 deletions
77
python/llm/example/CPU/HF-Transformers-AutoModels/Model/ziya/generate.py
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,77 @@ | ||
# | ||
# 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 | ||
import numpy as np | ||
|
||
from transformers import AutoTokenizer | ||
|
||
# you could tune the prompt based on your own model, | ||
# here the prompt tuning refers to https://huggingface.co/IDEA-CCNL/Ziya-Coding-34B-v1.0 | ||
ZIYA_PROMPT_FORMAT = "<human>: \n{prompt}\n<bot>: \n" | ||
|
||
if __name__ == '__main__': | ||
parser = argparse.ArgumentParser(description='Predict Tokens using `generate()` API for Ziya model') | ||
parser.add_argument('--repo-id-or-model-path', type=str, default="IDEA-CCNL/Ziya-Coding-34B-v1.0", | ||
help='The huggingface repo id for the Ziya model to be downloaded' | ||
', or the path to the huggingface checkpoint folder') | ||
parser.add_argument('--prompt', type=str, default="def quick_sort(arr):\n", | ||
help='Prompt to infer') | ||
parser.add_argument('--n-predict', type=int, default=128, | ||
help='Max tokens to predict') | ||
|
||
args = parser.parse_args() | ||
model_path = args.repo_id_or_model_path | ||
|
||
|
||
from bigdl.llm.transformers import AutoModelForCausalLM | ||
# enabling `use_cache=True` allows the model to utilize the previous | ||
# key/values attentions to speed up decoding; | ||
# to obtain optimal performance with BigDL-LLM INT4 optimizations, | ||
# it is important to set use_cache=True for Ziya models | ||
model = AutoModelForCausalLM.from_pretrained(model_path, | ||
load_in_4bit=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 = ZIYA_PROMPT_FORMAT.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, | ||
do_sample = True, | ||
top_p = 0.85, | ||
temperature = 0.8, | ||
repetition_penalty = 0.95, | ||
eos_token_id = tokenizer.eos_token_id, | ||
pad_token_id = tokenizer.pad_token_id, | ||
) | ||
end = time.time() | ||
output_str = tokenizer.batch_decode(output)[0] | ||
print(f'Inference time: {end-st} s') | ||
print('-'*20, 'Prompt', '-'*20) | ||
print(prompt) | ||
print('-'*20, 'Output', '-'*20) | ||
print(output_str) |
78 changes: 78 additions & 0 deletions
78
python/llm/example/CPU/PyTorch-Models/Model/ziya/README.md
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,78 @@ | ||
# Ziya | ||
In this directory, you will find examples on how you could use BigDL-LLM `optimize_model` API to accelerate Ziya models. For illustration purposes, we utilize the [IDEA-CCNL/Ziya-Coding-34B-v1.0](https://huggingface.co/IDEA-CCNL/Ziya-Coding-34B-v1.0) as a reference Ziya 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 Ziya model to predict the next N tokens using `generate()` API, with BigDL-LLM 'optimize_model' API. | ||
### 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 einops # additional package required for Ziya to conduct generation | ||
``` | ||
|
||
### 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 'def quick_sort(arr):\n' | ||
``` | ||
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 | ||
``` | ||
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 Ziya model to be downloaded, or the path to the huggingface checkpoint folder. It is default to be `'IDEA-CCNL/Ziya-Coding-34B-v1.0'`. | ||
- `--prompt`: str, argument defining the prompt to be inferred (with integrated prompt format for chat). It is default to be `def quick_sort(arr):\n`. | ||
- `--n-predict`: int, argument defining the max number of tokens to predict. It is default to be `128`. | ||
|
||
#### 2.4 Sample Output | ||
#### [IDEA-CCNL/Ziya-Coding-34B-v1.0](https://huggingface.co/IDEA-CCNL/Ziya-Coding-34B-v1.0) | ||
```log | ||
Inference time: xxxx s | ||
-------------------- Prompt -------------------- | ||
<human>: | ||
def quick_sort(arr):\n | ||
<bot>: | ||
-------------------- Output -------------------- | ||
<s> <human>: | ||
def quick_sort(arr):\n | ||
<bot>: | ||
def partition(arr, low, high): | ||
i = (low-1) | ||
pivot = arr[high] | ||
for j in range(low, high): | ||
if arr[j] <= pivot: | ||
arr[i], arr[j] = arr[j], arr[i] | ||
i = i+1 | ||
arr[i], arr[high] = arr[high], arr[i] | ||
return i | ||
def quick_sort(arr, low, high): | ||
if low < high: | ||
pi = partition(arr, low, | ||
``` |
78 changes: 78 additions & 0 deletions
78
python/llm/example/CPU/PyTorch-Models/Model/ziya/generate.py
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -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 | ||
import numpy as np | ||
|
||
from transformers import AutoTokenizer | ||
|
||
|
||
ZIYA_PROMPT_FORMAT = "<human>: \n{prompt}\n<bot>: \n" | ||
|
||
if __name__ == '__main__': | ||
parser = argparse.ArgumentParser(description='Predict Tokens using `generate()` API for Ziya model') | ||
parser.add_argument('--repo-id-or-model-path', type=str, default="IDEA-CCNL/Ziya-Coding-34B-v1.0", | ||
help='The huggingface repo id for the Ziya model to be downloaded' | ||
', or the path to the huggingface checkpoint folder') | ||
parser.add_argument('--prompt', type=str, default="def quick_sort(arr):\n", | ||
help='Prompt to infer') | ||
parser.add_argument('--n-predict', type=int, default=128, | ||
help='Max tokens to predict') | ||
|
||
args = parser.parse_args() | ||
model_path = args.repo_id_or_model_path | ||
|
||
|
||
from transformers import AutoModelForCausalLM | ||
from bigdl.llm import optimize_model | ||
# enabling `use_cache=True` allows the model to utilize the previous | ||
# key/values attentions to speed up decoding; | ||
# to obtain optimal performance with BigDL-LLM `optimization_model` API optimizations, | ||
# it is important to set use_cache=True for Ziya models | ||
model = AutoModelForCausalLM.from_pretrained(model_path, | ||
trust_remote_code=True, | ||
use_cache=True) | ||
model = optimize_model(model) | ||
|
||
# Load tokenizer | ||
tokenizer = AutoTokenizer.from_pretrained(model_path, | ||
trust_remote_code=True) | ||
|
||
# Generate predicted tokens | ||
with torch.inference_mode(): | ||
prompt = ZIYA_PROMPT_FORMAT.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, | ||
do_sample = True, | ||
top_p = 0.85, | ||
temperature = 0.8, | ||
repetition_penalty = 0.95, | ||
eos_token_id = tokenizer.eos_token_id, | ||
pad_token_id = tokenizer.pad_token_id, | ||
) | ||
end = time.time() | ||
output_str = tokenizer.batch_decode(output)[0] | ||
print(f'Inference time: {end-st} s') | ||
print('-'*20, 'Prompt', '-'*20) | ||
print(prompt) | ||
print('-'*20, 'Output', '-'*20) | ||
print(output_str) | ||
|