[Doc] Add Quick Start and Deployment
#1
by
RandomTao
- opened
README.md
CHANGED
@@ -52,6 +52,70 @@ For now, the standalone decoder is open-sourced and fully functional without hav
|
|
52 |
|
53 |
This model is static, trained on an offline dataset. Future versions may be released to enhance its performance on specialized tasks.
|
54 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
55 |
**License**
|
56 |
|
57 |
The TableGPT2-7B license permits both research and commercial use, with further details available in the [GitHub repository](https://github.com/tablegpt/tablegpt-agent).
|
|
|
52 |
|
53 |
This model is static, trained on an offline dataset. Future versions may be released to enhance its performance on specialized tasks.
|
54 |
|
55 |
+
**Quickstart**
|
56 |
+
|
57 |
+
Here provides a code snippet with apply_chat_template to show you how to load the tokenizer and model and how to generate contents.
|
58 |
+
```python
|
59 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
60 |
+
|
61 |
+
model_name = "tablegpt/TableGPT2-7B"
|
62 |
+
|
63 |
+
model = AutoModelForCausalLM.from_pretrained(
|
64 |
+
model_name,
|
65 |
+
torch_dtype="auto",
|
66 |
+
device_map="auto"
|
67 |
+
)
|
68 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
69 |
+
|
70 |
+
prompt = "Hey, who are you?"
|
71 |
+
messages = [
|
72 |
+
{"role": "system", "content": "You are a helpful assistant."},
|
73 |
+
{"role": "user", "content": prompt}
|
74 |
+
]
|
75 |
+
text = tokenizer.apply_chat_template(
|
76 |
+
messages,
|
77 |
+
tokenize=False,
|
78 |
+
add_generation_prompt=True
|
79 |
+
)
|
80 |
+
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
|
81 |
+
|
82 |
+
generated_ids = model.generate(
|
83 |
+
**model_inputs,
|
84 |
+
max_new_tokens=512
|
85 |
+
)
|
86 |
+
generated_ids = [
|
87 |
+
output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
|
88 |
+
]
|
89 |
+
|
90 |
+
response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
|
91 |
+
```
|
92 |
+
|
93 |
+
**Deployment**
|
94 |
+
|
95 |
+
For deployment, we recommend using vLLM.
|
96 |
+
* **Install vLLM**: You can install vLLM by running the following command.
|
97 |
+
```bash
|
98 |
+
pip install "vllm>=0.4.3"
|
99 |
+
```
|
100 |
+
* **Model Deployment**: Use vLLM to deploy your model. For example, you can use the command to set up a server similar to openAI:
|
101 |
+
```bash
|
102 |
+
python -m vllm.entrypoints.openai.api_server --served-model-name TableGPT2-7B --model path/to/weights
|
103 |
+
```
|
104 |
+
Then you can access the Chat API by:
|
105 |
+
|
106 |
+
```bash
|
107 |
+
curl http://localhost:8000/v1/chat/completions \
|
108 |
+
-H "Content-Type: application/json" \
|
109 |
+
-d '{
|
110 |
+
"model": "TableGPT2-7B",
|
111 |
+
"messages": [
|
112 |
+
{"role": "system", "content": "You are a helpful assistant."},
|
113 |
+
{"role": "user", "content": "Hey, who are you?"}
|
114 |
+
]
|
115 |
+
}'
|
116 |
+
|
117 |
+
```
|
118 |
+
|
119 |
**License**
|
120 |
|
121 |
The TableGPT2-7B license permits both research and commercial use, with further details available in the [GitHub repository](https://github.com/tablegpt/tablegpt-agent).
|