Update README.md

README.md

@@ -1,5 +1,9 @@
 ---
 base_model: unsloth/phi-3-mini-4k-instruct-bnb-4bit
+datasets:
+- b-mc2/sql-create-context
+- Clinton/Text-to-sql-v1
+- knowrohit07/know_sql
 language:
 - en
 license: apache-2.0
@@ -7,16 +11,65 @@ tags:
 - text-generation-inference
 - transformers
 - unsloth
-- mistral
+- phi-3
 - trl
 ---
 
-# Uploaded model
+This is a **phi-3-mini-4k-instruct-bnb-4bit** model, fine-tuned on **[b-mc2/sql-create-context](https://huggingface.co/datasets/b-mc2/sql-create-context)**, **[Clinton/Text-to-sql-v1](https://huggingface.co/datasets/Clinton/Text-to-sql-v1)** and **[knowrohit07/know_sql](https://huggingface.co/datasets/knowrohit07/know_sql)** 
+dataset. 
 
-- **Developed by:** dmedhi
-- **License:** apache-2.0
-- **Finetuned from model :** unsloth/phi-3-mini-4k-instruct-bnb-4bit
+## Model Usage
 
-This mistral model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library.
+Use the `unsloth` library to laod and run the model.
+
+Install `unsloth` and other dependencies.
+
+```python
+# 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 "trl<0.9.0" peft accelerate bitsandbytes torch
+```
+
+Use `FastLanguageModel` to download and laod the model from hf hub.
+
+```python
+from unsloth import FastLanguageModel
+model, tokenizer = FastLanguageModel.from_pretrained(
+ model_name = "dmedhi/Phi-3-mini-text2SQL-4k-instruct",
+ max_seq_length = 2048
+ dtype = None
+ load_in_4bit = True
+)
+FastLanguageModel.for_inference(model)
+
+prompt = """Below is a question that describes a SQL function, paired with a table Context that provides SQL table context. Write an answer that fullfils the user query.
+
+### Question:
+{}
+
+### Context:
+{}
+
+### Answer:
+{}"""
+
+inputs = tokenizer(
+[
+ prompt.format(
+ "What is the latest year that has ferrari 166 fl as the winning constructor?",
+ """CREATE TABLE table_name_7 (
+ year INTEGER,
+ winning_constructor VARCHAR
+ )""",
+ ""
+ )
+], return_tensors = "pt").to("cuda")
+
+outputs = model.generate(**inputs, max_new_tokens = 64, use_cache = True)
+tokenizer.batch_decode(outputs)
+
+```
+```bash
+# ["<s> Below is a question that describes a SQL function, paired with a table Context that provides SQL table context. Write an answer that fullfils the user query.\n\n### Question:\nWhat is the latest year that has ferrari 166 fl as the winning constructor?\n\n### Context:\nCREATE TABLE table_name_7 (\n year INTEGER,\n winning_constructor VARCHAR\n)\n\n### Answer:\nTo find the latest year that Ferrari 166 FL was the winning constructor, you can use the following SQL query:\n\n```sql\nSELECT MAX(year)\nFROM table_name_7\nWHERE winning_constructor = 'Ferrari 166 FL';\n```\n"]
+```
 
-[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)