Spaces:
Sleeping
Sleeping
File size: 2,651 Bytes
823232d ed07f25 823232d ed07f25 823232d ed07f25 88de55d ed07f25 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 |
import torch, torch.nn as nn, torch.nn.functional as F
import gradio as gr, random
# 📍 Device
device = 'cpu'
# 📚 Q&A Pairs
qa_pairs = [
("What is AI?", "Artificial Intelligence simulates human cognition using machines."),
("What is ML?", "Machine Learning enables systems to learn from data."),
("What is deep learning?", "Deep learning uses neural networks with many layers to learn complex patterns."),
("What is supervised learning?", "Supervised learning trains models on labeled data to make predictions."),
("What is unsupervised learning?", "Unsupervised learning finds patterns in unlabeled data.")
]
# 🔤 Tokenizer (word-level)
vocab = sorted(set(" ".join(q + " " + a for q, a in qa_pairs).split() + ["<END>"]))
stoi = {w:i for i,w in enumerate(vocab)}
itos = {i:w for w,i in stoi.items()}
encode = lambda s: [stoi[w] for w in s.split() if w in stoi]
decode = lambda l: ' '.join([itos[i] for i in l])
# 🧱 Dataset
data = []
for q, a in qa_pairs:
full = encode("Q: " + q + " A: " + a + " <END>")
x = torch.tensor(full[:-1], dtype=torch.long)
y = torch.tensor(full[1:], dtype=torch.long)
data.append((x, y))
# 🧠 Model
class TinySLM(nn.Module):
def __init__(self, vocab_size, n_embed=64):
super().__init__()
self.embed = nn.Embedding(vocab_size, n_embed)
self.rnn = nn.GRU(n_embed, n_embed, batch_first=True)
self.lm_head = nn.Linear(n_embed, vocab_size)
def forward(self, idx):
x = self.embed(idx)
out, _ = self.rnn(x)
return self.lm_head(out)
model = TinySLM(len(vocab)).to(device)
optimizer = torch.optim.AdamW(model.parameters(), lr=1e-3)
# 🏋️ Train
for step in range(1000):
x, y = random.choice(data)
logits = model(x.unsqueeze(0))
loss = F.cross_entropy(logits.squeeze(0), y)
optimizer.zero_grad(); loss.backward(); optimizer.step()
# 🔮 Generate
def generate_answer(question, max_new=50):
model.eval()
idx = torch.tensor(encode("Q: " + question + " A:"), dtype=torch.long).unsqueeze(0)
answer_tokens = []
for _ in range(max_new):
logits = model(idx)
next_id = torch.argmax(logits[:, -1], dim=-1).item()
if itos[next_id] == "<END>": break
answer_tokens.append(next_id)
idx = torch.cat([idx, torch.tensor([[next_id]])], dim=1)
return decode(answer_tokens)
# 🎨 Gradio UI
def chat_interface(user_input):
return generate_answer(user_input)
demo = gr.Interface(
fn=chat_interface,
inputs="text",
outputs="text",
title="SLM Chatbot PYTORCH DEEPIKA",
description="Ask a question about AI or ML"
)
demo.launch()
|