72 lines
3.0 KiB
Markdown
72 lines
3.0 KiB
Markdown
|
---
|
||
|
|
language:
|
||
|
|
- en
|
||
|
|
license: apache-2.0
|
||
|
|
tags:
|
||
|
|
- text-generation-inference
|
||
|
|
- transformers
|
||
|
|
- unsloth
|
||
|
|
- llama
|
||
|
|
- trl
|
||
|
|
- sft
|
||
|
|
base_model: unsloth/llama-3-8b-Instruct
|
||
|
|
datasets:
|
||
|
|
- tomasonjo/text2cypher-gpt4o-clean
|
||
|
|
---
|
||
|
|
|
||
|
|
# Uploaded model
|
||
|
|
|
||
|
|
- **Developed by:** tomasonjo
|
||
|
|
- **License:** apache-2.0
|
||
|
|
- **Finetuned from model :** unsloth/llama-3-8b-Instruct
|
||
|
|
|
||
|
|
This llama model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library.
|
||
|
|
|
||
|
|
**For more information visit [this link](https://github.com/neo4j-labs/text2cypher/tree/main/finetuning/unsloth-llama3#using-chat-prompt-template)**
|
||
|
|
|
||
|
|
|
||
|
|
## Example usage:
|
||
|
|
|
||
|
|
Install dependencies. Check [Unsloth documentation](https://github.com/unslothai/unsloth) for specific installation for other environments.
|
||
|
|
|
||
|
|
````python
|
||
|
|
%%capture
|
||
|
|
# Installs Unsloth, Xformers (Flash Attention) and all other packages!
|
||
|
|
!pip install "unsloth[colab-new] @ git+https://github.com/unslothai/unsloth.git"
|
||
|
|
!pip install --no-deps "xformers<0.0.26" trl peft accelerate bitsandbytes
|
||
|
|
````
|
||
|
|
|
||
|
|
Then you can load the model and use it as inference
|
||
|
|
|
||
|
|
```python
|
||
|
|
from unsloth.chat_templates import get_chat_template
|
||
|
|
|
||
|
|
tokenizer = get_chat_template(
|
||
|
|
tokenizer,
|
||
|
|
chat_template = "llama-3",
|
||
|
|
map_eos_token = True,
|
||
|
|
)
|
||
|
|
|
||
|
|
FastLanguageModel.for_inference(model) # Enable native 2x faster inference
|
||
|
|
|
||
|
|
schema = """Node properties: - **Question** - `favorites`: INTEGER Example: "0" - `answered`: BOOLEAN - `text`: STRING Example: "### This is: Bug ### Specifications OS: Win10" - `link`: STRING Example: "https://stackoverflow.com/questions/62224586/playg" - `createdAt`: DATE_TIME Min: 2020-06-05T16:57:19Z, Max: 2020-06-05T21:49:16Z - `title`: STRING Example: "Playground is not loading with apollo-server-lambd" - `id`: INTEGER Min: 62220505, Max: 62224586 - `upVotes`: INTEGER Example: "0" - `score`: INTEGER Example: "-1" - `downVotes`: INTEGER Example: "1" - **Tag** - `name`: STRING Example: "aws-lambda" - **User** - `image`: STRING Example: "https://lh3.googleusercontent.com/-NcFYSuXU0nk/AAA" - `link`: STRING Example: "https://stackoverflow.com/users/10251021/alexandre" - `id`: INTEGER Min: 751, Max: 13681006 - `reputation`: INTEGER Min: 1, Max: 420137 - `display_name`: STRING Example: "Alexandre Le" Relationship properties: The relationships: (:Question)-[:TAGGED]->(:Tag) (:User)-[:ASKED]->(:Question)"""
|
||
|
|
question = "Identify the top 5 questions with the most downVotes."
|
||
|
|
|
||
|
|
messages = [
|
||
|
|
{"role": "system", "content": "Given an input question, convert it to a Cypher query. No pre-amble."},
|
||
|
|
{"role": "user", "content": f"""Based on the Neo4j graph schema below, write a Cypher query that would answer the user's question:
|
||
|
|
{schema}
|
||
|
|
|
||
|
|
Question: {question}
|
||
|
|
Cypher query:"""}
|
||
|
|
]
|
||
|
|
inputs = tokenizer.apply_chat_template(
|
||
|
|
messages,
|
||
|
|
tokenize = True,
|
||
|
|
add_generation_prompt = True, # Must add for generation
|
||
|
|
return_tensors = "pt",
|
||
|
|
).to("cuda")
|
||
|
|
|
||
|
|
outputs = model.generate(input_ids = inputs, max_new_tokens = 128, use_cache = True)
|
||
|
|
tokenizer.batch_decode(outputs)
|
||
|
|
```
|