|
import torch |
|
import torch.nn as nn |
|
from torch.nn import functional as F |
|
import tiktoken |
|
import gradio as gr |
|
import torch |
|
import torch.nn as nn |
|
from torch.nn import functional as F |
|
import tiktoken |
|
import gradio as gr |
|
import asyncio |
|
import gradio as gr |
|
import asyncio |
|
|
|
|
|
def post_process_text(text): |
|
|
|
text = text.capitalize() |
|
|
|
|
|
sentences = text.split('.') |
|
complete_sentences = sentences[:-1] if len(sentences) > 1 else sentences |
|
|
|
|
|
processed_text = '. '.join(complete_sentences) |
|
if not processed_text.endswith('.'): |
|
processed_text += '.' |
|
|
|
return processed_text |
|
|
|
class GPTConfig: |
|
def __init__(self): |
|
self.block_size = 1024 |
|
self.vocab_size = 50304 |
|
self.n_layer = 12 |
|
self.n_head = 12 |
|
self.n_embd = 768 |
|
|
|
class CausalSelfAttention(nn.Module): |
|
def __init__(self, config): |
|
super().__init__() |
|
assert config.n_embd % config.n_head == 0 |
|
self.c_attn = nn.Linear(config.n_embd, 3 * config.n_embd) |
|
self.c_proj = nn.Linear(config.n_embd, config.n_embd) |
|
self.n_head = config.n_head |
|
self.n_embd = config.n_embd |
|
self.register_buffer("bias", torch.tril(torch.ones(config.block_size, config.block_size)).view(1, 1, config.block_size, config.block_size)) |
|
|
|
def forward(self, x): |
|
B, T, C = x.size() |
|
q, k, v = self.c_attn(x).split(self.n_embd, dim=2) |
|
k = k.view(B, T, self.n_head, C // self.n_head).transpose(1, 2) |
|
q = q.view(B, T, self.n_head, C // self.n_head).transpose(1, 2) |
|
v = v.view(B, T, self.n_head, C // self.n_head).transpose(1, 2) |
|
y = F.scaled_dot_product_attention(q, k, v, attn_mask=None, dropout_p=0.0, is_causal=True) |
|
y = y.transpose(1, 2).contiguous().view(B, T, C) |
|
return self.c_proj(y) |
|
|
|
class MLP(nn.Module): |
|
def __init__(self, config): |
|
super().__init__() |
|
self.c_fc = nn.Linear(config.n_embd, 4 * config.n_embd) |
|
self.gelu = nn.GELU() |
|
self.c_proj = nn.Linear(4 * config.n_embd, config.n_embd) |
|
|
|
def forward(self, x): |
|
return self.c_proj(self.gelu(self.c_fc(x))) |
|
|
|
class Block(nn.Module): |
|
def __init__(self, config): |
|
super().__init__() |
|
self.ln_1 = nn.LayerNorm(config.n_embd) |
|
self.attn = CausalSelfAttention(config) |
|
self.ln_2 = nn.LayerNorm(config.n_embd) |
|
self.mlp = MLP(config) |
|
|
|
def forward(self, x): |
|
x = x + self.attn(self.ln_1(x)) |
|
x = x + self.mlp(self.ln_2(x)) |
|
return x |
|
|
|
class GPT(nn.Module): |
|
def __init__(self, config): |
|
super().__init__() |
|
self.config = config |
|
self.transformer = nn.ModuleDict(dict( |
|
wte = nn.Embedding(config.vocab_size, config.n_embd), |
|
wpe = nn.Embedding(config.block_size, config.n_embd), |
|
h = nn.ModuleList([Block(config) for _ in range(config.n_layer)]), |
|
ln_f = nn.LayerNorm(config.n_embd), |
|
)) |
|
self.lm_head = nn.Linear(config.n_embd, config.vocab_size, bias=False) |
|
self.transformer.wte.weight = self.lm_head.weight |
|
self.apply(self._init_weights) |
|
|
|
def _init_weights(self, module): |
|
if isinstance(module, nn.Linear): |
|
torch.nn.init.normal_(module.weight, mean=0.0, std=0.02) |
|
if module.bias is not None: |
|
torch.nn.init.zeros_(module.bias) |
|
elif isinstance(module, nn.Embedding): |
|
torch.nn.init.normal_(module.weight, mean=0.0, std=0.02) |
|
|
|
def forward(self, idx, targets=None): |
|
device = idx.device |
|
b, t = idx.size() |
|
assert t <= self.config.block_size, f"Cannot forward sequence of length {t}, block size is only {self.config.block_size}" |
|
pos = torch.arange(0, t, dtype=torch.long, device=device).unsqueeze(0) |
|
|
|
tok_emb = self.transformer.wte(idx) |
|
pos_emb = self.transformer.wpe(pos) |
|
x = tok_emb + pos_emb |
|
for block in self.transformer.h: |
|
x = block(x) |
|
x = self.transformer.ln_f(x) |
|
logits = self.lm_head(x) |
|
|
|
loss = None |
|
if targets is not None: |
|
loss = F.cross_entropy(logits.view(-1, logits.size(-1)), targets.view(-1), ignore_index=-1) |
|
|
|
return logits, loss |
|
|
|
|
|
def load_model(model_path): |
|
config = GPTConfig() |
|
model = GPT(config) |
|
|
|
checkpoint = torch.load(model_path, map_location=torch.device('cpu')) |
|
|
|
print("Checkpoint keys:", checkpoint.keys()) |
|
|
|
if 'model_state_dict' in checkpoint: |
|
model.load_state_dict(checkpoint['model_state_dict']) |
|
else: |
|
model.load_state_dict(checkpoint) |
|
|
|
model.eval() |
|
return model |
|
|
|
|
|
model = load_model('gpt_model.pth') |
|
enc = tiktoken.get_encoding('gpt2') |
|
|
|
|
|
import torch |
|
import torch.nn as nn |
|
from torch.nn import functional as F |
|
import tiktoken |
|
import gradio as gr |
|
|
|
|
|
|
|
|
|
async def generate_text(prompt, max_length=432, temperature=0.8, top_k=40): |
|
input_ids = torch.tensor(enc.encode(prompt)).unsqueeze(0) |
|
generated = [] |
|
|
|
with torch.no_grad(): |
|
for _ in range(max_length): |
|
outputs, _ = model(input_ids) |
|
next_token_logits = outputs[:, -1, :] |
|
next_token_logits = next_token_logits / temperature |
|
top_k_logits, top_k_indices = torch.topk(next_token_logits, top_k, dim=-1) |
|
next_token_probs = F.softmax(top_k_logits, dim=-1) |
|
next_token_index = torch.multinomial(next_token_probs, num_samples=1) |
|
next_token = top_k_indices.gather(-1, next_token_index) |
|
|
|
input_ids = torch.cat([input_ids, next_token], dim=-1) |
|
generated.append(next_token.item()) |
|
|
|
next_token_str = enc.decode([next_token.item()]) |
|
yield next_token_str |
|
|
|
if next_token.item() == enc.encode('\n')[0] and len(generated) > 100: |
|
break |
|
|
|
await asyncio.sleep(0.02) |
|
|
|
if len(generated) == max_length: |
|
yield "... (output truncated due to length)" |
|
|
|
async def gradio_generate(prompt, max_length, temperature, top_k): |
|
output = "" |
|
async for token in generate_text(prompt, max_length, temperature, top_k): |
|
output += token |
|
yield output |
|
|
|
|
|
import gradio as gr |
|
import asyncio |
|
|
|
|
|
|
|
css = """ |
|
<style> |
|
body { |
|
background-color: #0f1624; |
|
color: #e0e0e0; |
|
font-family: 'Courier New', monospace; |
|
background-image: |
|
radial-gradient(white, rgba(255,255,255,.2) 2px, transparent 40px), |
|
radial-gradient(white, rgba(255,255,255,.15) 1px, transparent 30px), |
|
radial-gradient(white, rgba(255,255,255,.1) 2px, transparent 40px), |
|
radial-gradient(rgba(255,255,255,.4), rgba(255,255,255,.1) 2px, transparent 30px); |
|
background-size: 550px 550px, 350px 350px, 250px 250px, 150px 150px; |
|
background-position: 0 0, 40px 60px, 130px 270px, 70px 100px; |
|
animation: backgroundScroll 60s linear infinite; |
|
} |
|
@keyframes backgroundScroll { |
|
0% { background-position: 0 0, 40px 60px, 130px 270px, 70px 100px; } |
|
100% { background-position: 550px 550px, 590px 610px, 680px 820px, 620px 650px; } |
|
} |
|
.container { max-width: 800px; margin: 0 auto; padding: 20px; } |
|
.header { |
|
text-align: center; |
|
margin-bottom: 30px; |
|
font-family: 'Copperplate', fantasy; |
|
color: #ffd700; |
|
text-shadow: 0 0 10px #ffd700, 0 0 20px #ffd700, 0 0 30px #ffd700; |
|
} |
|
.chat-box { |
|
background-color: rgba(42, 42, 42, 0.7); |
|
border-radius: 15px; |
|
padding: 20px; |
|
margin-bottom: 20px; |
|
box-shadow: 0 0 20px rgba(255, 215, 0, 0.3); |
|
} |
|
.user-input { |
|
background-color: rgba(58, 58, 58, 0.8); |
|
border: 2px solid #ffd700; |
|
color: #ffffff; |
|
padding: 10px; |
|
border-radius: 5px; |
|
width: 100%; |
|
transition: all 0.3s ease; |
|
} |
|
.user-input:focus { |
|
box-shadow: 0 0 15px #ffd700; |
|
} |
|
.generate-btn { |
|
background-color: #ffd700; |
|
color: #0f1624; |
|
border: none; |
|
padding: 10px 20px; |
|
border-radius: 5px; |
|
cursor: pointer; |
|
font-weight: bold; |
|
transition: all 0.3s ease; |
|
} |
|
.generate-btn:hover { |
|
background-color: #ffec8b; |
|
transform: scale(1.05); |
|
} |
|
.output-box { |
|
background-color: rgba(42, 42, 42, 0.7); |
|
border-radius: 15px; |
|
padding: 20px; |
|
margin-top: 20px; |
|
min-height: 100px; |
|
border: 1px solid #ffd700; |
|
white-space: pre-wrap; |
|
font-family: 'Georgia', serif; |
|
line-height: 1.6; |
|
box-shadow: inset 0 0 10px rgba(255, 215, 0, 0.3); |
|
} |
|
.gr-slider { |
|
--slider-color: #ffd700; |
|
} |
|
.gr-box { |
|
border-color: #ffd700; |
|
background-color: rgba(42, 42, 42, 0.7); |
|
} |
|
</style> |
|
""" |
|
|
|
with gr.Blocks(css=css) as demo: |
|
gr.HTML("<div class='header'><h1>π Enchanted Tales Generator π</h1></div>") |
|
|
|
with gr.Row(): |
|
with gr.Column(scale=3): |
|
prompt = gr.Textbox( |
|
placeholder="Begin your magical journey here (e.g., 'In a realm beyond the mists of time...')", |
|
label="Story Incantation", |
|
elem_classes="user-input" |
|
) |
|
with gr.Column(scale=1): |
|
generate_btn = gr.Button("Weave the Tale", elem_classes="generate-btn") |
|
|
|
with gr.Row(): |
|
max_length = gr.Slider(minimum=50, maximum=500, value=432, step=1, label="Scroll Length") |
|
temperature = gr.Slider(minimum=0.1, maximum=1.0, value=0.8, step=0.1, label="Magical Intensity") |
|
top_k = gr.Slider(minimum=1, maximum=100, value=40, step=1, label="Arcane Diversity") |
|
|
|
output = gr.Markdown(elem_classes="output-box") |
|
|
|
generate_btn.click( |
|
gradio_generate, |
|
inputs=[prompt, max_length, temperature, top_k], |
|
outputs=output |
|
) |
|
|
|
gr.HTML(""" |
|
<div style="text-align: center; margin-top: 20px; font-style: italic; color: #ffd700;"> |
|
"In the realm of imagination, every word is a spell, every sentence a charm." |
|
</div> |
|
""") |
|
|
|
if __name__ == "__main__": |
|
demo.launch() |