LLM: add replit and starcoder to gpu pytorch model example (#9154)

python/llm/example/GPU/PyTorch-Models/Model/README.md

@@ -9,6 +9,10 @@ You can use `optimize_model` API to accelerate general PyTorch models on Intel G
 | ChatGLM2       | [link](chatglm2)                                         |
 | Baichuan       | [link](baichuan)                                         |
 | Baichuan2      | [link](baichuan2)                                        |
+| Replit         | [link](replit)                                           |
+| StarCoder      | [link](starcoder)                                        |
+| Dolly v1       | [link](dolly-v1)                                         |
+| Dolly v2       | [link](dolly-v2)                                         |
 
 ## Verified Hardware Platforms
 

python/llm/example/GPU/PyTorch-Models/Model/replit/README.md

@@ -0,0 +1,60 @@
+# Replit
+In this directory, you will find examples on how you could use BigDL-LLM `optimize_model` API to accelerate Replit models. For illustration purposes, we utilize the [replit/replit-code-v1-3b](https://huggingface.co/replit/replit-code-v1-3b) as reference Replit models.
+
+## 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 Replit model to predict the next N tokens using `generate()` API, with BigDL-LLM INT4 optimizations on Intel GPUs.
+### 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
+
+# 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
+```
+
+### 2. Configures OneAPI environment variables
+```bash
+source /opt/intel/oneapi/setvars.sh
+```
+
+### 3. Run
+
+For optimal performance on Arc, it is recommended to set several environment variables.
+
+```bash
+export USE_XETLA=OFF
+export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1
+```
+
+```bash
+python ./generate.py --prompt 'def print_hello_world():'
+```
+
+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 Replit model to be downloaded, or the path to the huggingface checkpoint folder. It is default to be `'replit/replit-code-v1-3b'`.
+- `--prompt PROMPT`: argument defining the prompt to be infered (with integrated prompt format for chat). It is default to be `def print_hello_world():'`.
+- `--n-predict N_PREDICT`: argument defining the max number of tokens to predict. It is default to be `32`.
+
+#### 2.3 Sample Output
+#### [replit/replit-code-v1-3b](https://huggingface.co/replit/replit-code-v1-3b)
+```log
+Inference time: xxxx s
+-------------------- Output --------------------
+def print_hello_world():
+    print("Hello")
+    print("World")
+
+print_hello_world()
+
+
+def print_hello_world():
+    print
+```

python/llm/example/GPU/PyTorch-Models/Model/replit/generate.py

@@ -0,0 +1,73 @@
+#
+# 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 intel_extension_for_pytorch as ipex
+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,
+REPLIT_PROMPT_FORMAT = "{prompt}"
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser(description='Predict Tokens using `generate()` API for Replit model')
+    parser.add_argument('--repo-id-or-model-path', type=str, default="replit/replit-code-v1-3b",
+                        help='The huggingface repo id for the Replit model to be downloaded'
+                             ', or the path to the huggingface checkpoint folder')
+    parser.add_argument('--prompt', type=str, default="def print_hello_world():",
+                        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)
+
+    # With only one line to enable BigDL-LLM optimization on model
+    model = optimize_model(model)
+
+    model = model.to('xpu')
+
+    # Load tokenizer
+    tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
+    
+    # Generate predicted tokens
+    with torch.inference_mode():
+        prompt = REPLIT_PROMPT_FORMAT.format(prompt=args.prompt)
+        input_ids = tokenizer.encode(prompt, return_tensors="pt").to('xpu')
+        # ipex model needs a warmup, then inference time can be accurate
+        output = model.generate(input_ids,
+                                max_new_tokens=args.n_predict)
+
+        # start inference
+        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)

python/llm/example/GPU/PyTorch-Models/Model/starcoder/README.md

@@ -0,0 +1,60 @@
+# StarCoder
+In this directory, you will find examples on how you could use BigDL-LLM `optimize_model` API to accelerate StarCoder models. For illustration purposes, we utilize the [bigcode/starcoder](https://huggingface.co/bigcode/starcoder) as reference StarCoder models.
+
+## 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 StarCoder model to predict the next N tokens using `generate()` API, with BigDL-LLM INT4 optimizations on Intel GPUs.
+### 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
+
+# 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
+```
+
+### 2. Configures OneAPI environment variables
+```bash
+source /opt/intel/oneapi/setvars.sh
+```
+
+### 3. Run
+
+For optimal performance on Arc, it is recommended to set several environment variables.
+
+```bash
+export USE_XETLA=OFF
+export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1
+```
+
+```bash
+python ./generate.py --prompt 'def print_hello_world():'
+```
+
+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 StarCoder model to be downloaded, or the path to the huggingface checkpoint folder. It is default to be `'bigcode/starcoder'`.
+- `--prompt PROMPT`: argument defining the prompt to be infered (with integrated prompt format for chat). It is default to be `def print_hello_world():'`.
+- `--n-predict N_PREDICT`: argument defining the max number of tokens to predict. It is default to be `32`.
+
+#### 2.3 Sample Output
+#### [bigcode/starcoder](https://huggingface.co/bigcode/starcoder)
+```log
+Inference time: xxxx s
+-------------------- Output --------------------
+def print_hello_world():
+    print("Hello World!")
+
+
+def print_hello_name(name):
+    print(f"Hello {name}!")
+
+
+def print_
+```

python/llm/example/GPU/PyTorch-Models/Model/starcoder/generate.py

@@ -0,0 +1,73 @@
+#
+# 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 intel_extension_for_pytorch as ipex
+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,
+STARCODER_PROMPT_FORMAT = "{prompt}"
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser(description='Predict Tokens using `generate()` API for StarCoder model')
+    parser.add_argument('--repo-id-or-model-path', type=str, default="bigcode/starcoder",
+                        help='The huggingface repo id for the StarCoder model to be downloaded'
+                             ', or the path to the huggingface checkpoint folder')
+    parser.add_argument('--prompt', type=str, default="def print_hello_world():",
+                        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)
+
+    # With only one line to enable BigDL-LLM optimization on model
+    model = optimize_model(model)
+
+    model = model.to('xpu')
+
+    # Load tokenizer
+    tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
+    
+    # Generate predicted tokens
+    with torch.inference_mode():
+        prompt = STARCODER_PROMPT_FORMAT.format(prompt=args.prompt)
+        input_ids = tokenizer.encode(prompt, return_tensors="pt").to('xpu')
+        # ipex model needs a warmup, then inference time can be accurate
+        output = model.generate(input_ids,
+                                max_new_tokens=args.n_predict)
+
+        # start inference
+        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)