Update README.md
Browse files
README.md
CHANGED
|
@@ -29,42 +29,57 @@ pass
|
|
| 29 |
## Inference
|
| 30 |
```python
|
| 31 |
|
| 32 |
-
max_seq_length = 2048 # Choose any! We auto support RoPE Scaling internally!
|
| 33 |
-
dtype = None # None for auto detection. Float16 for Tesla T4, V100, Bfloat16 for Ampere+
|
| 34 |
-
load_in_4bit = True # Use 4bit quantization to reduce memory usage. Can be False.
|
| 35 |
import torch
|
| 36 |
-
from peft import AutoPeftModelForCausalLM
|
| 37 |
from transformers import AutoTokenizer
|
| 38 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 39 |
model = AutoPeftModelForCausalLM.from_pretrained(
|
| 40 |
-
|
| 41 |
-
|
| 42 |
-
|
| 43 |
-
|
| 44 |
-
)
|
| 45 |
tokenizer = AutoTokenizer.from_pretrained(model_id)
|
| 46 |
-
|
| 47 |
-
|
|
|
|
| 48 |
|
| 49 |
### Context:
|
| 50 |
-
{}
|
| 51 |
|
| 52 |
### Question:
|
| 53 |
-
{}
|
| 54 |
|
| 55 |
### Answer:
|
| 56 |
-
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
|
| 62 |
-
|
| 63 |
-
)
|
| 64 |
-
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 68 |
|
| 69 |
|
| 70 |
|
|
|
|
| 29 |
## Inference
|
| 30 |
```python
|
| 31 |
|
|
|
|
|
|
|
|
|
|
| 32 |
import torch
|
|
|
|
| 33 |
from transformers import AutoTokenizer
|
| 34 |
+
from peft import AutoPeftModelForCausalLM
|
| 35 |
+
import gradio as gr
|
| 36 |
+
|
| 37 |
+
# Load the model and tokenizer
|
| 38 |
+
model_id = "DisgustingOzil/MIstral_Pak_Law"
|
| 39 |
+
dtype = torch.float16 # Adjust as necessary
|
| 40 |
+
load_in_4bit = True
|
| 41 |
+
|
| 42 |
model = AutoPeftModelForCausalLM.from_pretrained(
|
| 43 |
+
model_id,
|
| 44 |
+
load_in_4bit=load_in_4bit,
|
| 45 |
+
torch_dtype=dtype,
|
| 46 |
+
).to("cuda")
|
|
|
|
| 47 |
tokenizer = AutoTokenizer.from_pretrained(model_id)
|
| 48 |
+
|
| 49 |
+
def generate_response(context, question):
|
| 50 |
+
alpaca_prompt = f"""Given the context and a specific question, generate a comprehensive and detailed response that accurately addresses the query.
|
| 51 |
|
| 52 |
### Context:
|
| 53 |
+
{context}
|
| 54 |
|
| 55 |
### Question:
|
| 56 |
+
{question}
|
| 57 |
|
| 58 |
### Answer:
|
| 59 |
+
"""
|
| 60 |
+
|
| 61 |
+
inputs = tokenizer(
|
| 62 |
+
[alpaca_prompt], return_tensors="pt", padding=True, truncation=True, max_length=2048
|
| 63 |
+
).to("cuda")
|
| 64 |
+
|
| 65 |
+
outputs = model.generate(**inputs, max_new_tokens=64, use_cache=True)
|
| 66 |
+
response = tokenizer.batch_decode(outputs, skip_special_tokens=True)
|
| 67 |
+
summary_start_index = response[0].find("### Answer:")
|
| 68 |
+
summary_text = response[0][summary_start_index:].replace("### Answer:", "").strip()
|
| 69 |
+
return summary_text
|
| 70 |
+
# return response[0]
|
| 71 |
+
|
| 72 |
+
# Define the Gradio interface
|
| 73 |
+
iface = gr.Interface(
|
| 74 |
+
fn=generate_response,
|
| 75 |
+
inputs=[gr.Textbox(label="Context"), gr.Textbox(label="Question")],
|
| 76 |
+
outputs=gr.Textbox(label="Answer"),
|
| 77 |
+
title="Abandoned Properties Act 1975 Query",
|
| 78 |
+
description="Enter the context and a specific question to generate a response based on the Abandoned Properties (Taking Over and Management) Act 1975.",
|
| 79 |
+
)
|
| 80 |
+
|
| 81 |
+
# Run the app
|
| 82 |
+
iface.launch(debug=True)
|
| 83 |
|
| 84 |
|
| 85 |
|