Upload README.md with huggingface_hub

README.md

@@ -1,15 +1,44 @@
-# Mistral-7B SQL Model (Saved)
+---
+library_name: transformers
+pipeline_tag: text-generation
+tags:
+- mistral
+- text-to-sql
+- sql
+language:
+- en
+license: apache-2.0
+base_model: mistralai/Mistral-7B-v0.1
+---
 
-This folder contains the tokenizer and either full model weights or LoRA adapters.
+# Mistral-7B SQL (fine-tuned)
 
-## Loading
+Fine-tuned Mistral-7B for Text-to-SQL on `b-mc2/sql-create-context`.
 
-```python
-from transformers import AutoModelForCausalLM, AutoTokenizer
-from peft import PeftModel
+## Usage (Transformers)
+
+from transformers import AutoTokenizer, AutoModelForCausalLM
 import torch
 
-model_path = '/content/drive/MyDrive/mistral7b_sql_model'
-tokenizer = AutoTokenizer.from_pretrained(model_path)
-model = AutoModelForCausalLM.from_pretrained(model_path, torch_dtype=torch.bfloat16, device_map='auto')
-``` 
+model_id = "kirankotha/mistral7b-sql-model"
+tok = AutoTokenizer.from_pretrained(model_id)
+mdl = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto", torch_dtype=torch.float16)
+
+prompt = (
+    "You are a text-to-SQL model.
+"
+    "### Input:
+"
+    "Which product has the highest price?
+"
+    "### Context:
+"
+    "CREATE TABLE products (id INTEGER, name TEXT, price REAL)
+"
+    "### Response:
+"
+)
+
+ids = tok(prompt, return_tensors="pt").to(mdl.device)
+out = mdl.generate(**ids, max_new_tokens=100, do_sample=False, pad_token_id=tok.pad_token_id)
+print(tok.decode(out[0], skip_special_tokens=True))