luboslenco
535 lines
26 KiB
TypeScript
535 lines
26 KiB
TypeScript
|
|
type text_to_photo_node_t = {
|
|
base?: logic_node_t;
|
|
};
|
|
|
|
let text_to_photo_node_prompt: string = "";
|
|
let text_to_photo_node_image: gpu_texture_t = null;
|
|
let text_to_photo_node_tiling: bool = false;
|
|
let text_to_photo_node_text_encoder_blob: buffer_t;
|
|
let text_to_photo_node_unet_blob: buffer_t;
|
|
let text_to_photo_node_vae_decoder_blob: buffer_t;
|
|
|
|
function text_to_photo_node_create(raw: ui_node_t, args: f32_array_t): text_to_photo_node_t {
|
|
let n: text_to_photo_node_t = {};
|
|
n.base = logic_node_create(n);
|
|
n.base.get_as_image = text_to_photo_node_get_as_image;
|
|
n.base.get_cached_image = text_to_photo_node_get_cached_image;
|
|
return n;
|
|
}
|
|
|
|
function text_to_photo_node_get_as_image(self: text_to_photo_node_t, from: i32): gpu_texture_t {
|
|
text_to_photo_node_image = text_to_photo_node_stable_diffusion(text_to_photo_node_prompt);
|
|
return text_to_photo_node_image;
|
|
}
|
|
|
|
function text_to_photo_node_get_cached_image(self: text_to_photo_node_t): gpu_texture_t {
|
|
return text_to_photo_node_image;
|
|
}
|
|
|
|
function text_to_photo_node_button(node_id: i32) {
|
|
let node: ui_node_t = ui_get_node(ui_nodes_get_canvas(true).nodes, node_id);
|
|
|
|
text_to_photo_node_tiling = node.buttons[0].default_value[0] == 0 ? false : true;
|
|
text_to_photo_node_prompt = ui_text_area(ui_handle(__ID__), ui_align_t.LEFT, true, tr("prompt"), true);
|
|
node.buttons[1].height = string_split(text_to_photo_node_prompt, "\n").length;
|
|
}
|
|
|
|
function text_to_photo_node_stable_diffusion(prompt: string, inpaint_latents: f32_array_t = null, offset: i32 = 0, upscale: bool = true, mask: f32_array_t = null, latents_orig: f32_array_t = null): gpu_texture_t {
|
|
let _text_encoder_blob: buffer_t = data_get_blob("models/sd_text_encoder.quant.onnx");
|
|
let _unet_blob: buffer_t = data_get_blob("models/sd_unet.quant.onnx");
|
|
let _vae_decoder_blob: buffer_t = data_get_blob("models/sd_vae_decoder.quant.onnx");
|
|
text_to_photo_node_text_encoder_blob = _text_encoder_blob;
|
|
text_to_photo_node_unet_blob = _unet_blob;
|
|
text_to_photo_node_vae_decoder_blob = _vae_decoder_blob;
|
|
let enc: text_encoder_result_t = text_to_photo_node_text_encoder(prompt, inpaint_latents);
|
|
let latents: f32_array_t = text_to_photo_node_unet(enc.latents, enc.text_embeddings, mask, latents_orig, offset);
|
|
return text_to_photo_node_vae_decoder(latents, upscale);
|
|
}
|
|
|
|
type text_encoder_result_t = {
|
|
latents?: f32_array_t;
|
|
text_embeddings?: f32_array_t;
|
|
};
|
|
|
|
function text_to_photo_node_text_encoder(prompt: string, inpaint_latents: f32_array_t): text_encoder_result_t {
|
|
console_progress(tr("Processing") + " - " + tr("Text to Photo"));
|
|
|
|
let words: string[] = string_split(
|
|
trim_end(string_replace_all(string_replace_all(string_replace_all(prompt, "\n", " "), ",", " , "), " ", " ")), " "
|
|
);
|
|
|
|
if (text_to_photo_node_vocab == null) {
|
|
let vocab_buffer: buffer_t = data_get_blob("models/vocab.json");
|
|
let vocab_json: string = sys_buffer_to_string(vocab_buffer);
|
|
text_to_photo_node_vocab = json_parse_to_map(vocab_json);
|
|
}
|
|
|
|
for (let i: i32 = 0; i < words.length; ++i) {
|
|
let word: string = to_lower_case(words[i]) + "</w>";
|
|
let value_string: string = map_get(text_to_photo_node_vocab, word);
|
|
let value: i32 = parse_int(value_string);
|
|
text_to_photo_node_text_input_ids[i + 1] = value;
|
|
}
|
|
for (let i: i32 = words.length; i < (text_to_photo_node_text_input_ids.length - 1); ++i) {
|
|
text_to_photo_node_text_input_ids[i + 1] = 49407; // <|endoftext|>
|
|
}
|
|
|
|
let i32a: i32_array_t = i32_array_create_from_array(text_to_photo_node_text_input_ids);
|
|
let tensors: buffer_t[] = [buffer_create_from_raw(i32a.buffer, i32a.length * 4)];
|
|
let input_shape: i32_array_t[] = [];
|
|
let input_shape0: i32[] = [1, 77];
|
|
array_push(input_shape, input_shape0);
|
|
let output_shape: i32[] = [1, 77, 768];
|
|
let text_embeddings_buf: buffer_t = iron_ml_inference(text_to_photo_node_text_encoder_blob, tensors, input_shape, output_shape, config_raw.gpu_inference);
|
|
let text_embeddings: f32_array_t = f32_array_create_from_buffer(text_embeddings_buf);
|
|
|
|
i32a = i32_array_create_from_array(text_to_photo_node_uncond_input_ids);
|
|
tensors = [buffer_create_from_raw(i32a.buffer, i32a.length * 4)];
|
|
let uncond_embeddings_buf: buffer_t = iron_ml_inference(text_to_photo_node_text_encoder_blob, tensors, input_shape, output_shape, config_raw.gpu_inference);
|
|
let uncond_embeddings: f32_array_t = f32_array_create_from_buffer(uncond_embeddings_buf);
|
|
|
|
let f32a: f32_array_t = f32_array_create(uncond_embeddings.length + text_embeddings.length);
|
|
for (let i: i32 = 0; i < uncond_embeddings.length; ++i) f32a[i] = uncond_embeddings[i];
|
|
for (let i: i32 = 0; i < text_embeddings.length; ++i) f32a[i + uncond_embeddings.length] = text_embeddings[i];
|
|
text_embeddings = f32a;
|
|
|
|
let width: i32 = 512;
|
|
let height: i32 = 512;
|
|
let latents: f32_array_t = f32_array_create(1 * 4 * math_floor(height / 8) * math_floor(width / 8));
|
|
if (inpaint_latents == null) {
|
|
for (let i: i32 = 0; i < latents.length; ++i) latents[i] = math_cos(2.0 * 3.14 * random_node_get_float()) * math_sqrt(-2.0 * math_log(random_node_get_float()));
|
|
}
|
|
else {
|
|
for (let i: i32 = 0; i < latents.length; ++i) latents[i] = inpaint_latents[i];
|
|
}
|
|
|
|
let res: text_encoder_result_t = {
|
|
latents: latents,
|
|
text_embeddings: text_embeddings
|
|
};
|
|
return res;
|
|
}
|
|
|
|
function text_to_photo_node_unet(latents: f32_array_t, text_embeddings: f32_array_t, mask: f32_array_t, latents_orig: f32_array_t, offset: i32): f32_array_t {
|
|
let latent_model_input: f32_array_t = f32_array_create(latents.length * 2);
|
|
let noise_pred_uncond: f32_array_t = f32_array_create(latents.length);
|
|
let noise_pred_text: f32_array_t = f32_array_create(latents.length);
|
|
|
|
let cur_latents: f32_array_t = null;
|
|
let num_train_timesteps: i32 = 1000;
|
|
let num_inference_steps: i32 = 50;
|
|
let ets: f32_array_t[] = [];
|
|
let counter: i32 = 0;
|
|
|
|
while (true) {
|
|
let a: i32 = counter;
|
|
let b: i32 = 50 - offset;
|
|
console_progress(tr("Processing") + " - " + tr("Text to Photo") + " (" + a + "/" + b + ")");
|
|
|
|
let timestep: i32 = text_to_photo_node_timesteps[counter + offset];
|
|
for (let i: i32 = 0; i < latents.length; ++i) latent_model_input[i] = latents[i];
|
|
for (let i: i32 = 0; i < latents.length; ++i) latent_model_input[i + latents.length] = latents[i];
|
|
|
|
let t32: i32_array_t = i32_array_create(2);
|
|
t32[0] = timestep;
|
|
let tensors: buffer_t[] = [
|
|
buffer_create_from_raw(latent_model_input.buffer, latent_model_input.length * 4),
|
|
buffer_create_from_raw(t32.buffer, t32.length * 4),
|
|
buffer_create_from_raw(text_embeddings.buffer, text_embeddings.length * 4),
|
|
];
|
|
let input_shape: i32_array_t[] = [];
|
|
let input_shape0: i32[] = [2, 4, 64, 64];
|
|
let input_shape1: i32[] = [1];
|
|
let input_shape2: i32[] = [2, 77, 768];
|
|
array_push(input_shape, input_shape0);
|
|
array_push(input_shape, input_shape1);
|
|
array_push(input_shape, input_shape2);
|
|
let output_shape: i32[] = [2, 4, 64, 64];
|
|
let noise_pred_buf: buffer_t = iron_ml_inference(text_to_photo_node_unet_blob, tensors, input_shape, output_shape, config_raw.gpu_inference);
|
|
let noise_pred: f32_array_t = f32_array_create_from_buffer(noise_pred_buf);
|
|
|
|
for (let i: i32 = 0; i < noise_pred_uncond.length; ++i) noise_pred_uncond[i] = noise_pred[i];
|
|
for (let i: i32 = 0; i < noise_pred_text.length; ++i) noise_pred_text[i] = noise_pred[noise_pred_uncond.length + i];
|
|
|
|
let guidance_scale: f32 = 7.5;
|
|
noise_pred = f32_array_create(noise_pred_uncond.length);
|
|
for (let i: i32 = 0; i < noise_pred_uncond.length; ++i) {
|
|
noise_pred[i] = noise_pred_uncond[i] + guidance_scale * (noise_pred_text[i] - noise_pred_uncond[i]);
|
|
}
|
|
|
|
let prev_timestep: i32 = math_floor(math_max(timestep - math_floor(num_train_timesteps / num_inference_steps), 0));
|
|
|
|
if (counter != 1) {
|
|
array_push(ets, noise_pred);
|
|
}
|
|
else {
|
|
prev_timestep = timestep;
|
|
timestep = timestep + math_floor(num_train_timesteps / num_inference_steps);
|
|
}
|
|
|
|
if (ets.length == 1 && counter == 0) {
|
|
cur_latents = latents;
|
|
}
|
|
else if (ets.length == 1 && counter == 1) {
|
|
let _noise_pred: f32_array_t = f32_array_create(noise_pred.length);
|
|
for (let i: i32 = 0; i < noise_pred.length; ++i) {
|
|
_noise_pred[i] = (noise_pred[i] + ets[ets.length - 1][i]) / 2;
|
|
}
|
|
noise_pred = _noise_pred;
|
|
latents = cur_latents;
|
|
cur_latents = null;
|
|
}
|
|
else if (ets.length == 2) {
|
|
let _noise_pred: f32_array_t = f32_array_create(noise_pred.length);
|
|
for (let i: i32 = 0; i < noise_pred.length; ++i) {
|
|
_noise_pred[i] = (3 * ets[ets.length - 1][i] - ets[ets.length - 2][i]) / 2;
|
|
}
|
|
noise_pred = _noise_pred;
|
|
}
|
|
else if (ets.length == 3) {
|
|
let _noise_pred: f32_array_t = f32_array_create(noise_pred.length);
|
|
for (let i: i32 = 0; i < noise_pred.length; ++i) {
|
|
_noise_pred[i] = (23 * ets[ets.length - 1][i] - 16 * ets[ets.length - 2][i] + 5 * ets[ets.length - 3][i]) / 12;
|
|
}
|
|
noise_pred = _noise_pred;
|
|
}
|
|
else {
|
|
let _noise_pred: f32_array_t = f32_array_create(noise_pred.length);
|
|
for (let i: i32 = 0; i < noise_pred.length; ++i) {
|
|
_noise_pred[i] = (1 / 24) * (55 * ets[ets.length - 1][i] - 59 * ets[ets.length - 2][i] + 37 * ets[ets.length - 3][i] - 9 * ets[ets.length - 4][i]);
|
|
}
|
|
noise_pred = _noise_pred;
|
|
}
|
|
|
|
let alpha_prod_t: f32 = text_to_photo_node_alphas_cumprod[timestep + 1];
|
|
let alpha_prod_t_prev: f32 = text_to_photo_node_alphas_cumprod[prev_timestep + 1];
|
|
let beta_prod_t: f32 = 1 - alpha_prod_t;
|
|
let beta_prod_t_prev: f32 = 1 - alpha_prod_t_prev;
|
|
let latents_coeff: f32 = math_pow(alpha_prod_t_prev / alpha_prod_t, (0.5));
|
|
let noise_pred_denom_coeff: f32 = alpha_prod_t * math_pow(beta_prod_t_prev, (0.5)) + math_pow(alpha_prod_t * beta_prod_t * alpha_prod_t_prev, (0.5));
|
|
for (let i: i32 = 0; i < latents.length; ++i) {
|
|
latents[i] = (latents_coeff * latents[i] - (alpha_prod_t_prev - alpha_prod_t) * noise_pred[i] / noise_pred_denom_coeff);
|
|
}
|
|
counter += 1;
|
|
|
|
if (mask != null) {
|
|
let noise: f32_array_t = f32_array_create(latents.length);
|
|
for (let i: i32 = 0; i < noise.length; ++i) {
|
|
noise[i] = math_cos(2.0 * 3.14 * random_node_get_float()) * math_sqrt(-2.0 * math_log(random_node_get_float()));
|
|
}
|
|
let sqrt_alpha_prod: f32 = math_pow(text_to_photo_node_alphas_cumprod[timestep], 0.5);
|
|
let sqrt_one_minus_alpha_prod: f32 = math_pow(1.0 - text_to_photo_node_alphas_cumprod[timestep], 0.5);
|
|
|
|
let init_latents_proper: f32_array_t = f32_array_create(latents.length);
|
|
for (let i: i32 = 0; i < init_latents_proper.length; ++i) {
|
|
init_latents_proper[i] = sqrt_alpha_prod * latents_orig[i] + sqrt_one_minus_alpha_prod * noise[i];
|
|
}
|
|
|
|
for (let i: i32 = 0; i < latents.length; ++i) {
|
|
latents[i] = (init_latents_proper[i] * mask[i]) + (latents[i] * (1.0 - mask[i]));
|
|
}
|
|
}
|
|
|
|
if (counter == (51 - offset)) {
|
|
break;
|
|
}
|
|
}
|
|
|
|
return latents;
|
|
}
|
|
|
|
function text_to_photo_node_vae_decoder(latents: f32_array_t, upscale: bool): gpu_texture_t {
|
|
console_progress(tr("Processing") + " - " + tr("Text to Photo"));
|
|
|
|
for (let i: i32 = 0; i < latents.length; ++i) {
|
|
latents[i] = 1.0 / 0.18215 * latents[i];
|
|
}
|
|
|
|
let tensors: buffer_t[] = [buffer_create_from_raw(latents.buffer, latents.length * 4)];
|
|
let input_shape: i32_array_t[] = [];
|
|
let input_shape0: i32[] = [1, 4, 64, 64];
|
|
array_push(input_shape, input_shape0);
|
|
let output_shape: i32[] = [1, 3, 512, 512];
|
|
let pyimage_buf: buffer_t = iron_ml_inference(text_to_photo_node_vae_decoder_blob, tensors, input_shape, output_shape, config_raw.gpu_inference);
|
|
let pyimage: f32_array_t = f32_array_create_from_buffer(pyimage_buf);
|
|
|
|
for (let i: i32 = 0; i < pyimage.length; ++i) {
|
|
pyimage[i] = pyimage[i] / 2.0 + 0.5;
|
|
if (pyimage[i] < 0) pyimage[i] = 0;
|
|
else if (pyimage[i] > 1) pyimage[i] = 1;
|
|
}
|
|
|
|
let u8a: u8_array_t = u8_array_create(4 * 512 * 512);
|
|
for (let i: i32 = 0; i < (512 * 512); ++i) {
|
|
u8a[i * 4 ] = math_floor(pyimage[i ] * 255);
|
|
u8a[i * 4 + 1] = math_floor(pyimage[i + 512 * 512 ] * 255);
|
|
u8a[i * 4 + 2] = math_floor(pyimage[i + 512 * 512 * 2] * 255);
|
|
u8a[i * 4 + 3] = 255;
|
|
}
|
|
let image: gpu_texture_t = gpu_create_texture_from_bytes(u8a, 512, 512);
|
|
|
|
if (text_to_photo_node_tiling) {
|
|
tiling_node_prompt = text_to_photo_node_prompt;
|
|
let seed: i32 = random_node_get_seed();
|
|
return tiling_node_sd_tiling(image, seed);
|
|
}
|
|
else {
|
|
if (upscale) {
|
|
upscale_node_load_blob();
|
|
while (image.width < config_get_texture_res_x()) {
|
|
let last_image: gpu_texture_t = image;
|
|
image = upscale_node_esrgan(image);
|
|
iron_delete_texture(last_image);
|
|
}
|
|
return image;
|
|
}
|
|
else {
|
|
return image;
|
|
}
|
|
}
|
|
}
|
|
|
|
let text_to_photo_node_def: ui_node_t = {
|
|
id: 0,
|
|
name: _tr("Text to Photo"),
|
|
type: "text_to_photo_node",
|
|
x: 0,
|
|
y: 0,
|
|
color: 0xff4982a0,
|
|
inputs: [],
|
|
outputs: [
|
|
{
|
|
id: 0,
|
|
node_id: 0,
|
|
name: _tr("Color"),
|
|
type: "RGBA",
|
|
color: 0xffc7c729,
|
|
default_value: f32_array_create_xyzw(0.0, 0.0, 0.0, 1.0),
|
|
min: 0.0,
|
|
max: 1.0,
|
|
precision: 100,
|
|
display: 0
|
|
}
|
|
],
|
|
buttons: [
|
|
{
|
|
name: _tr("tiling"),
|
|
type: "BOOL",
|
|
output: 0,
|
|
default_value: f32_array_create_x(0),
|
|
data: null,
|
|
min: 0.0,
|
|
max: 1.0,
|
|
precision: 100,
|
|
height: 0
|
|
},
|
|
{
|
|
name: "text_to_photo_node_button",
|
|
type: "CUSTOM",
|
|
output: -1,
|
|
default_value: f32_array_create_x(0),
|
|
data: null,
|
|
min: 0.0,
|
|
max: 1.0,
|
|
precision: 100,
|
|
height: 1
|
|
}
|
|
],
|
|
width: 0
|
|
};
|
|
|
|
let text_to_photo_node_text_input_ids: i32[] = [49406, 49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407,
|
|
49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407,
|
|
49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407,
|
|
49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407,
|
|
49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407,
|
|
49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407,
|
|
49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407,
|
|
49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407,
|
|
49407, 49407, 49407, 49407, 49407];
|
|
|
|
let text_to_photo_node_uncond_input_ids: i32[] = [49406, 49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407,
|
|
49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407,
|
|
49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407,
|
|
49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407,
|
|
49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407,
|
|
49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407,
|
|
49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407,
|
|
49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407, 49407,
|
|
49407, 49407, 49407, 49407, 49407];
|
|
|
|
let text_to_photo_node_alphas_cumprod: f32[] = [0.99915, 0.998296, 0.9974381, 0.99657613, 0.99571025, 0.9948404,
|
|
0.9939665, 0.99308866, 0.9922069, 0.9913211, 0.9904313, 0.98953754,
|
|
0.9886398, 0.9877381, 0.9868324, 0.9859227, 0.985009, 0.98409134,
|
|
0.9831697, 0.982244, 0.98131436, 0.9803807, 0.97944313, 0.97850156,
|
|
0.977556, 0.9766064, 0.9756529, 0.9746954, 0.9737339, 0.9727684,
|
|
0.97179896, 0.97082555, 0.96984816, 0.96886677, 0.9678814, 0.96689206,
|
|
0.9658988, 0.96490157, 0.9639003, 0.96289515, 0.961886, 0.9608729,
|
|
0.9598558, 0.9588347, 0.9578097, 0.95678073, 0.95574784, 0.95471096,
|
|
0.95367014, 0.9526254, 0.95157677, 0.9505242, 0.9494677, 0.9484073,
|
|
0.94734293, 0.94627464, 0.9452024, 0.9441263, 0.9430463, 0.94196236,
|
|
0.9408745, 0.9397828, 0.9386872, 0.93758774, 0.93648434, 0.93537706,
|
|
0.9342659, 0.9331509, 0.93203205, 0.93090934, 0.9297828, 0.92865235,
|
|
0.92751807, 0.92638, 0.9252381, 0.9240923, 0.9229427, 0.92178935,
|
|
0.9206321, 0.9194711, 0.9183063, 0.9171377, 0.9159653, 0.91478914,
|
|
0.9136092, 0.9124255, 0.9112381, 0.9100469, 0.908852, 0.90765333,
|
|
0.90645087, 0.9052447, 0.90403485, 0.9028213, 0.90160406, 0.9003831,
|
|
0.8991585, 0.89793015, 0.8966982, 0.8954625, 0.8942232, 0.8929803,
|
|
0.8917337, 0.8904835, 0.88922966, 0.88797224, 0.8867112, 0.88544655,
|
|
0.88417834, 0.88290656, 0.8816312, 0.8803523, 0.87906986, 0.8777839,
|
|
0.87649435, 0.87520134, 0.8739048, 0.8726048, 0.8713013, 0.8699943,
|
|
0.8686838, 0.86736983, 0.86605245, 0.8647316, 0.8634073, 0.8620797,
|
|
0.86074865, 0.85941416, 0.85807633, 0.8567351, 0.85539055, 0.8540426,
|
|
0.8526913, 0.8513367, 0.8499788, 0.8486176, 0.84725314, 0.84588534,
|
|
0.8445143, 0.84314, 0.8417625, 0.84038174, 0.8389977, 0.83761054,
|
|
0.83622015, 0.8348266, 0.8334298, 0.83202994, 0.8306269, 0.8292208,
|
|
0.8278115, 0.8263991, 0.82498366, 0.8235651, 0.82214355, 0.8207189,
|
|
0.8192912, 0.8178604, 0.8164267, 0.81499, 0.8135503, 0.8121076,
|
|
0.810662, 0.8092134, 0.8077619, 0.80630755, 0.8048503, 0.80339015,
|
|
0.80192715, 0.8004613, 0.7989926, 0.79752105, 0.79604673, 0.7945696,
|
|
0.79308975, 0.79160714, 0.79012173, 0.78863364, 0.7871428, 0.7856493,
|
|
0.78415316, 0.78265435, 0.78115284, 0.7796487, 0.778142, 0.77663267,
|
|
0.77512074, 0.77360624, 0.7720892, 0.7705696, 0.7690475, 0.7675229,
|
|
0.7659958, 0.7644662, 0.76293427, 0.76139987, 0.759863, 0.7583238,
|
|
0.7567822, 0.7552382, 0.75369185, 0.7521432, 0.7505923, 0.74903905,
|
|
0.74748355, 0.7459258, 0.74436575, 0.7428035, 0.7412391, 0.7396724,
|
|
0.7381036, 0.7365327, 0.73495966, 0.7333845, 0.73180723, 0.7302279,
|
|
0.7286465, 0.72706306, 0.7254776, 0.7238901, 0.72230065, 0.72070926,
|
|
0.7191159, 0.71752065, 0.7159235, 0.7143244, 0.7127235, 0.7111207,
|
|
0.7095161, 0.7079097, 0.7063015, 0.7046916, 0.7030799, 0.70146644,
|
|
0.6998513, 0.69823444, 0.69661593, 0.69499576, 0.693374, 0.6917506,
|
|
0.6901256, 0.68849903, 0.6868709, 0.6852412, 0.6836101, 0.6819774,
|
|
0.6803433, 0.6787077, 0.67707074, 0.6754323, 0.67379254, 0.6721514,
|
|
0.67050886, 0.668865, 0.6672199, 0.6655735, 0.6639258, 0.6622769,
|
|
0.6606268, 0.6589755, 0.657323, 0.65566933, 0.6540145, 0.6523586,
|
|
0.6507016, 0.6490435, 0.64738435, 0.6457242, 0.644063, 0.6424008,
|
|
0.64073765, 0.63907355, 0.6374085, 0.63574255, 0.63407576, 0.6324081,
|
|
0.63073957, 0.6290703, 0.6274002, 0.6257293, 0.6240577, 0.6223853,
|
|
0.6207122, 0.61903846, 0.61736405, 0.615689, 0.6140133, 0.61233705,
|
|
0.6106602, 0.6089828, 0.6073049, 0.6056264, 0.60394746, 0.60226804,
|
|
0.6005882, 0.59890795, 0.5972273, 0.59554625, 0.59386486, 0.5921831,
|
|
0.59050107, 0.5888187, 0.5871361, 0.5854532, 0.5837701, 0.5820868,
|
|
0.5804033, 0.57871974, 0.57703596, 0.5753521, 0.57366806, 0.571984,
|
|
0.5702999, 0.5686158, 0.56693166, 0.56524754, 0.56356347, 0.56187946,
|
|
0.56019557, 0.55851173, 0.5568281, 0.55514455, 0.5534612, 0.551778,
|
|
0.5500951, 0.5484124, 0.5467299, 0.54504776, 0.5433659, 0.5416844,
|
|
0.5400032, 0.5383224, 0.536642, 0.534962, 0.53328246, 0.53160334,
|
|
0.52992475, 0.52824664, 0.52656907, 0.52489203, 0.5232156, 0.5215397,
|
|
0.51986444, 0.5181898, 0.51651585, 0.51484257, 0.51317, 0.51149815,
|
|
0.509827, 0.50815666, 0.5064871, 0.5048183, 0.5031504, 0.5014833,
|
|
0.49981716, 0.49815187, 0.4964875, 0.49482408, 0.49316162, 0.49150014,
|
|
0.48983967, 0.48818022, 0.48652178, 0.48486444, 0.48320818, 0.48155302,
|
|
0.479899, 0.47824612, 0.47659442, 0.4749439, 0.47329462, 0.47164655,
|
|
0.46999976, 0.46835423, 0.46671, 0.4650671, 0.46342552, 0.46178532,
|
|
0.46014652, 0.45850912, 0.45687312, 0.45523855, 0.45360544, 0.45197386,
|
|
0.45034376, 0.44871515, 0.44708812, 0.44546264, 0.44383875, 0.44221646,
|
|
0.4405958, 0.4389768, 0.43735942, 0.43574375, 0.43412977, 0.43251753,
|
|
0.430907, 0.42929825, 0.42769128, 0.42608613, 0.4244828, 0.4228813,
|
|
0.42128167, 0.4196839, 0.41808805, 0.4164941, 0.4149021, 0.41331202,
|
|
0.41172394, 0.41013786, 0.40855378, 0.40697172, 0.40539172, 0.40381378,
|
|
0.40223792, 0.40066415, 0.39909253, 0.39752305, 0.3959557, 0.39439055,
|
|
0.3928276, 0.39126685, 0.38970834, 0.38815206, 0.38659805, 0.38504633,
|
|
0.3834969, 0.3819498, 0.38040504, 0.37886262, 0.37732255, 0.37578487,
|
|
0.3742496, 0.37271675, 0.37118635, 0.36965838, 0.3681329, 0.3666099,
|
|
0.3650894, 0.3635714, 0.36205596, 0.36054307, 0.35903275, 0.35752502,
|
|
0.35601988, 0.35451737, 0.35301748, 0.35152024, 0.35002568, 0.34853378,
|
|
0.3470446, 0.34555808, 0.3440743, 0.34259328, 0.341115, 0.3396395,
|
|
0.3381668, 0.3366969, 0.33522978, 0.3337655, 0.3323041, 0.3308455,
|
|
0.3293898, 0.327937, 0.3264871, 0.3250401, 0.32359603, 0.3221549,
|
|
0.32071674, 0.31928152, 0.3178493, 0.3164201, 0.3149939, 0.3135707,
|
|
0.31215054, 0.31073344, 0.3093194, 0.30790845, 0.30650055, 0.30509576,
|
|
0.3036941, 0.30229557, 0.30090016, 0.2995079, 0.29811877, 0.29673284,
|
|
0.2953501, 0.29397056, 0.29259422, 0.2912211, 0.28985122, 0.28848457,
|
|
0.28712118, 0.28576106, 0.28440422, 0.28305066, 0.2817004, 0.28035346,
|
|
0.27900982, 0.27766952, 0.27633256, 0.27499893, 0.27366868, 0.2723418,
|
|
0.27101827, 0.26969814, 0.26838142, 0.26706812, 0.26575825, 0.26445177,
|
|
0.26314875, 0.26184916, 0.26055303, 0.25926036, 0.25797117, 0.25668547,
|
|
0.25540325, 0.25412452, 0.2528493, 0.25157762, 0.25030944, 0.24904479,
|
|
0.24778369, 0.24652614, 0.24527213, 0.2440217, 0.24277483, 0.24153154,
|
|
0.24029182, 0.2390557, 0.23782317, 0.23659426, 0.23536895, 0.23414725,
|
|
0.23292919, 0.23171476, 0.23050396, 0.2292968, 0.2280933, 0.22689344,
|
|
0.22569725, 0.22450472, 0.22331588, 0.22213072, 0.22094923, 0.21977143,
|
|
0.21859734, 0.21742693, 0.21626024, 0.21509725, 0.21393798, 0.21278243,
|
|
0.2116306, 0.2104825, 0.20933813, 0.20819749, 0.2070606, 0.20592746,
|
|
0.20479806, 0.20367241, 0.20255052, 0.20143238, 0.200318, 0.19920738,
|
|
0.19810054, 0.19699748, 0.19589819, 0.19480269, 0.19371095, 0.192623,
|
|
0.19153884, 0.19045846, 0.18938187, 0.18830907, 0.18724008, 0.18617487,
|
|
0.18511346, 0.18405585, 0.18300205, 0.18195206, 0.18090586, 0.17986348,
|
|
0.1788249, 0.17779014, 0.17675918, 0.17573205, 0.17470871, 0.17368919,
|
|
0.17267348, 0.17166159, 0.1706535, 0.16964924, 0.1686488, 0.16765216,
|
|
0.16665934, 0.16567034, 0.16468513, 0.16370374, 0.16272618, 0.16175242,
|
|
0.16078247, 0.15981634, 0.15885401, 0.1578955, 0.1569408, 0.15598992,
|
|
0.15504283, 0.15409954, 0.15316005, 0.15222436, 0.15129249, 0.15036441,
|
|
0.14944012, 0.14851964, 0.14760293, 0.14669003, 0.1457809, 0.14487557,
|
|
0.14397402, 0.14307626, 0.14218228, 0.14129207, 0.14040563, 0.13952295,
|
|
0.13864405, 0.13776892, 0.13689755, 0.13602993, 0.13516606, 0.13430595,
|
|
0.13344958, 0.13259697, 0.1317481, 0.13090296, 0.13006155, 0.12922388,
|
|
0.12838994, 0.1275597, 0.1267332, 0.1259104, 0.12509131, 0.12427593,
|
|
0.12346424, 0.12265625, 0.12185195, 0.12105133, 0.1202544, 0.11946114,
|
|
0.11867155, 0.11788563, 0.11710336, 0.11632475, 0.11554979, 0.11477847,
|
|
0.1140108, 0.11324675, 0.11248633, 0.11172953, 0.11097635, 0.11022678,
|
|
0.10948081, 0.10873844, 0.10799967, 0.10726449, 0.10653288, 0.10580485,
|
|
0.10508038, 0.10435947, 0.10364211, 0.1029283, 0.10221803, 0.1015113,
|
|
0.10080809, 0.10010841, 0.09941223, 0.09871957, 0.0980304, 0.09734473,
|
|
0.09666254, 0.09598383, 0.09530859, 0.09463682, 0.0939685, 0.09330362,
|
|
0.09264219, 0.09198419, 0.09132962, 0.09067846, 0.09003071, 0.08938637,
|
|
0.08874542, 0.08810785, 0.08747366, 0.08684284, 0.08621538, 0.08559129,
|
|
0.08497053, 0.0843531, 0.08373901, 0.08312824, 0.08252078, 0.08191663,
|
|
0.08131576, 0.08071819, 0.08012389, 0.07953286, 0.07894509, 0.07836057,
|
|
0.07777929, 0.07720125, 0.07662643, 0.07605482, 0.07548642, 0.07492122,
|
|
0.07435921, 0.07380038, 0.07324471, 0.07269221, 0.07214285, 0.07159664,
|
|
0.07105356, 0.07051361, 0.06997676, 0.06944302, 0.06891238, 0.06838482,
|
|
0.06786034, 0.06733891, 0.06682055, 0.06630524, 0.06579296, 0.06528371,
|
|
0.06477747, 0.06427424, 0.06377401, 0.06327677, 0.0627825, 0.06229121,
|
|
0.06180287, 0.06131747, 0.06083502, 0.06035549, 0.05987888, 0.05940517,
|
|
0.05893436, 0.05846644, 0.05800139, 0.05753921, 0.05707989, 0.05662341,
|
|
0.05616977, 0.05571895, 0.05527094, 0.05482575, 0.05438334, 0.05394372,
|
|
0.05350687, 0.05307278, 0.05264145, 0.05221286, 0.05178699, 0.05136385,
|
|
0.05094342, 0.05052568, 0.05011064, 0.04969827, 0.04928857, 0.04888153,
|
|
0.04847714, 0.04807537, 0.04767624, 0.04727972, 0.0468858, 0.04649448,
|
|
0.04610574, 0.04571956, 0.04533596, 0.0449549, 0.04457638, 0.04420039,
|
|
0.04382691, 0.04345594, 0.04308747, 0.04272148, 0.04235797, 0.04199693,
|
|
0.04163833, 0.04128218, 0.04092846, 0.04057716, 0.04022827, 0.03988178,
|
|
0.03953768, 0.03919596, 0.0388566, 0.0385196, 0.03818495, 0.03785263,
|
|
0.03752263, 0.03719494, 0.03686956, 0.03654647, 0.03622566, 0.03590712,
|
|
0.03559083, 0.0352768, 0.034965, 0.03465543, 0.03434808, 0.03404293,
|
|
0.03373997, 0.0334392, 0.0331406, 0.03284416, 0.03254988, 0.03225773,
|
|
0.03196772, 0.03167982, 0.03139404, 0.03111035, 0.03082875, 0.03054923,
|
|
0.03027177, 0.02999637, 0.02972301, 0.02945168, 0.02918238, 0.0289151,
|
|
0.02864981, 0.02838652, 0.02812521, 0.02786587, 0.02760849, 0.02735306,
|
|
0.02709957, 0.02684801, 0.02659837, 0.02635064, 0.0261048, 0.02586086,
|
|
0.02561878, 0.02537858, 0.02514023, 0.02490373, 0.02466906, 0.02443622,
|
|
0.0242052, 0.02397598, 0.02374856, 0.02352292, 0.02329905, 0.02307695,
|
|
0.02285661, 0.02263801, 0.02242114, 0.022206, 0.02199257, 0.02178084,
|
|
0.02157081, 0.02136246, 0.02115579, 0.02095079, 0.02074743, 0.02054573,
|
|
0.02034565, 0.0201472, 0.01995037, 0.01975514, 0.01956151, 0.01936947,
|
|
0.019179, 0.0189901, 0.01880275, 0.01861695, 0.01843269, 0.01824996,
|
|
0.01806875, 0.01788905, 0.01771084, 0.01753413, 0.0173589, 0.01718514,
|
|
0.01701284, 0.016842, 0.0166726, 0.01650463, 0.0163381, 0.01617297,
|
|
0.01600925, 0.01584694, 0.01568601, 0.01552646, 0.01536828, 0.01521146,
|
|
0.015056, 0.01490187, 0.01474909, 0.01459763, 0.01444749, 0.01429865,
|
|
0.01415112, 0.01400487, 0.01385991, 0.01371622, 0.0135738, 0.01343263,
|
|
0.01329271, 0.01315403, 0.01301658, 0.01288035, 0.01274534, 0.01261153,
|
|
0.01247892, 0.01234749, 0.01221725, 0.01208818, 0.01196027, 0.01183351,
|
|
0.01170791, 0.01158344, 0.0114601, 0.01133789, 0.01121679, 0.0110968,
|
|
0.01097791, 0.01086011, 0.01074339, 0.01062774, 0.01051317, 0.01039965,
|
|
0.01028718, 0.01017576, 0.01006538, 0.00995602, 0.00984768, 0.00974036,
|
|
0.00963405, 0.00952873, 0.0094244, 0.00932106, 0.00921869, 0.00911729,
|
|
0.00901685, 0.00891737, 0.00881884, 0.00872124, 0.00862457, 0.00852883,
|
|
0.00843401, 0.0083401, 0.0082471, 0.00815499, 0.00806377, 0.00797343,
|
|
0.00788397, 0.00779538, 0.00770765, 0.00762078, 0.00753476, 0.00744958,
|
|
0.00736523, 0.00728171, 0.00719902, 0.00711714, 0.00703607, 0.0069558,
|
|
0.00687633, 0.00679765, 0.00671975, 0.00664263, 0.00656627, 0.00649069,
|
|
0.00641586, 0.00634178, 0.00626845, 0.00619586, 0.006124, 0.00605286,
|
|
0.00598245, 0.00591276, 0.00584377, 0.00577549, 0.0057079, 0.00564101,
|
|
0.0055748, 0.00550927, 0.00544442, 0.00538023, 0.00531671, 0.00525384,
|
|
0.00519163, 0.00513006, 0.00506913, 0.00500883, 0.00494917, 0.00489013,
|
|
0.0048317, 0.00477389, 0.00471669, 0.00466009];
|
|
|
|
let text_to_photo_node_timesteps: i32[] = [981, 961, 961, 941, 921, 901, 881, 861, 841, 821, 801, 781, 761, 741, 721, 701, 681, 661,
|
|
641, 621, 601, 581, 561, 541, 521, 501, 481, 461, 441, 421, 401, 381, 361, 341, 321, 301,
|
|
281, 261, 241, 221, 201, 181, 161, 141, 121, 101, 81, 61, 41, 21, 1];
|
|
|
|
let text_to_photo_node_vocab: map_t<string, string> = null;
|