This commit is contained in:
2026-01-23 12:00:29 +08:00
parent 97e47be051
commit 0f74156460
5 changed files with 22 additions and 4 deletions

View File

@@ -68,6 +68,7 @@ python example/run_pipeline.py --device auto
- Continuous head can be bounded with `tanh` via `use_tanh_eps` in config. - Continuous head can be bounded with `tanh` via `use_tanh_eps` in config.
- Export now clamps continuous features to training min/max and preserves integer/decimal precision. - Export now clamps continuous features to training min/max and preserves integer/decimal precision.
- `<UNK>` tokens are replaced by the most frequent token for each discrete column at export. - `<UNK>` tokens are replaced by the most frequent token for each discrete column at export.
- Continuous prediction mode can be switched via `cont_pred` (`eps` or `x0`).
- The script only samples the first 5000 rows to stay fast. - The script only samples the first 5000 rows to stay fast.
- `prepare_data.py` runs without PyTorch, but `train.py` and `sample.py` require it. - `prepare_data.py` runs without PyTorch, but `train.py` and `sample.py` require it.
- `train.py` and `sample.py` auto-select GPU if available; otherwise they fall back to CPU. - `train.py` and `sample.py` auto-select GPU if available; otherwise they fall back to CPU.

View File

@@ -25,6 +25,7 @@
"cond_dim": 32, "cond_dim": 32,
"use_tanh_eps": true, "use_tanh_eps": true,
"eps_scale": 0.7, "eps_scale": 0.7,
"cont_pred": "x0",
"sample_batch_size": 8, "sample_batch_size": 8,
"sample_seq_len": 128 "sample_seq_len": 128
} }

View File

@@ -179,14 +179,20 @@ def main():
cond_id = torch.full((args.batch_size,), int(args.condition_id), device=device, dtype=torch.long) cond_id = torch.full((args.batch_size,), int(args.condition_id), device=device, dtype=torch.long)
cond = cond_id cond = cond_id
cont_pred = str(cfg.get("cont_pred", "eps")).lower()
for t in reversed(range(args.timesteps)): for t in reversed(range(args.timesteps)):
t_batch = torch.full((args.batch_size,), t, device=device, dtype=torch.long) t_batch = torch.full((args.batch_size,), t, device=device, dtype=torch.long)
eps_pred, logits = model(x_cont, x_disc, t_batch, cond) cont_pred_out, logits = model(x_cont, x_disc, t_batch, cond)
a_t = alphas[t] a_t = alphas[t]
a_bar_t = alphas_cumprod[t] a_bar_t = alphas_cumprod[t]
coef1 = 1.0 / torch.sqrt(a_t) coef1 = 1.0 / torch.sqrt(a_t)
coef2 = (1 - a_t) / torch.sqrt(1 - a_bar_t) coef2 = (1 - a_t) / torch.sqrt(1 - a_bar_t)
if cont_pred == "x0":
# eps = (x_t - sqrt(a_bar) * x0) / sqrt(1 - a_bar)
eps_pred = (x_cont - torch.sqrt(a_bar_t) * cont_pred_out) / torch.sqrt(1 - a_bar_t + 1e-8)
else:
eps_pred = cont_pred_out
mean_x = coef1 * (x_cont - coef2 * eps_pred) mean_x = coef1 * (x_cont - coef2 * eps_pred)
if t > 0: if t > 0:
noise = torch.randn_like(x_cont) noise = torch.randn_like(x_cont)

View File

@@ -44,6 +44,7 @@ def main():
batch_size = int(cfg.get("sample_batch_size", cfg.get("batch_size", BATCH_SIZE))) batch_size = int(cfg.get("sample_batch_size", cfg.get("batch_size", BATCH_SIZE)))
clip_k = float(cfg.get("clip_k", CLIP_K)) clip_k = float(cfg.get("clip_k", CLIP_K))
use_condition = bool(cfg.get("use_condition")) and cfg.get("condition_type") == "file_id" use_condition = bool(cfg.get("use_condition")) and cfg.get("condition_type") == "file_id"
cont_pred = str(cfg.get("cont_pred", "eps")).lower()
cond_dim = int(cfg.get("cond_dim", 32)) cond_dim = int(cfg.get("cond_dim", 32))
use_tanh_eps = bool(cfg.get("use_tanh_eps", False)) use_tanh_eps = bool(cfg.get("use_tanh_eps", False))
eps_scale = float(cfg.get("eps_scale", 1.0)) eps_scale = float(cfg.get("eps_scale", 1.0))
@@ -96,13 +97,17 @@ def main():
for t in reversed(range(timesteps)): for t in reversed(range(timesteps)):
t_batch = torch.full((batch_size,), t, device=DEVICE, dtype=torch.long) t_batch = torch.full((batch_size,), t, device=DEVICE, dtype=torch.long)
eps_pred, logits = model(x_cont, x_disc, t_batch, cond) cont_pred_out, logits = model(x_cont, x_disc, t_batch, cond)
# Continuous reverse step (DDPM): x_{t-1} mean # Continuous reverse step (DDPM): x_{t-1} mean
a_t = alphas[t] a_t = alphas[t]
a_bar_t = alphas_cumprod[t] a_bar_t = alphas_cumprod[t]
coef1 = 1.0 / torch.sqrt(a_t) coef1 = 1.0 / torch.sqrt(a_t)
coef2 = (1 - a_t) / torch.sqrt(1 - a_bar_t) coef2 = (1 - a_t) / torch.sqrt(1 - a_bar_t)
if cont_pred == "x0":
eps_pred = (x_cont - torch.sqrt(a_bar_t) * cont_pred_out) / torch.sqrt(1 - a_bar_t + 1e-8)
else:
eps_pred = cont_pred_out
mean = coef1 * (x_cont - coef2 * eps_pred) mean = coef1 * (x_cont - coef2 * eps_pred)
if t > 0: if t > 0:
noise = torch.randn_like(x_cont) noise = torch.randn_like(x_cont)

View File

@@ -51,6 +51,7 @@ DEFAULTS = {
"cond_dim": 32, "cond_dim": 32,
"use_tanh_eps": True, "use_tanh_eps": True,
"eps_scale": 1.0, "eps_scale": 1.0,
"cont_pred": "eps",
} }
@@ -161,6 +162,7 @@ def main():
device = resolve_device(str(config["device"])) device = resolve_device(str(config["device"]))
print("device", device) print("device", device)
cont_pred = str(config.get("cont_pred", "eps")).lower()
model = HybridDiffusionModel( model = HybridDiffusionModel(
cont_dim=len(cont_cols), cont_dim=len(cont_cols),
disc_vocab_sizes=vocab_sizes, disc_vocab_sizes=vocab_sizes,
@@ -217,9 +219,12 @@ def main():
mask_tokens = torch.tensor(vocab_sizes, device=device) mask_tokens = torch.tensor(vocab_sizes, device=device)
x_disc_t, mask = q_sample_discrete(x_disc, t, mask_tokens, int(config["timesteps"])) x_disc_t, mask = q_sample_discrete(x_disc, t, mask_tokens, int(config["timesteps"]))
eps_pred, logits = model(x_cont_t, x_disc_t, t, cond) cont_pred_out, logits = model(x_cont_t, x_disc_t, t, cond)
loss_cont = F.mse_loss(eps_pred, noise) if cont_pred == "x0":
loss_cont = F.mse_loss(cont_pred_out, x_cont)
else:
loss_cont = F.mse_loss(cont_pred_out, noise)
loss_disc = 0.0 loss_disc = 0.0
loss_disc_count = 0 loss_disc_count = 0
for i, logit in enumerate(logits): for i, logit in enumerate(logits):