stupdi/ASLTranslator
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

grok lock in pt 3

Browse Source
This commit is contained in:
Stupdi Go
2026-01-10 23:26:33 -06:00
parent ea0cb9bd87
commit 1922898517
1 changed files with 68 additions and 62 deletions
Download Patch File Download Diff File Expand all files Collapse all files

130
training.py
View File

@@ -14,6 +14,7 @@ from sklearn.preprocessing import LabelEncoder, StandardScaler
from multiprocessing import Pool, cpu_count from multiprocessing import Pool, cpu_count
from functools import partial from functools import partial
from tqdm import tqdm from tqdm import tqdm
from collections import Counter
def load_kaggle_asl_data(base_path): def load_kaggle_asl_data(base_path):
@@ -26,7 +27,6 @@ def load_kaggle_asl_data(base_path):
def extract_hand_landmarks_from_parquet(path): def extract_hand_landmarks_from_parquet(path):
try: try:
df = pd.read_parquet(path) df = pd.read_parquet(path)
# Take either left or right hand - prefer the one with more landmarks
left = df[df["type"] == "left_hand"] left = df[df["type"] == "left_hand"]
right = df[df["type"] == "right_hand"] right = df[df["type"] == "right_hand"]
@@ -53,8 +53,8 @@ def extract_hand_landmarks_from_parquet(path):
]) ])
landmarks_seq.append(lm_list) landmarks_seq.append(lm_list)
return np.array(landmarks_seq, dtype=np.float32) # (T, 21, 3) return np.array(landmarks_seq, dtype=np.float32)
except Exception: except:
return None return None
@@ -63,20 +63,20 @@ def get_features_sequence(landmarks_seq, max_frames=100):
return None return None
# Center on wrist # Center on wrist
landmarks_seq = landmarks_seq - landmarks_seq[:, 0:1, :] landmarks_seq -= landmarks_seq[:, 0:1, :]
# Better scale: distance between index finger tip and middle finger tip # Scale using index → middle finger tip distance (more stable than single point)
scale = np.linalg.norm(landmarks_seq[:, 8] - landmarks_seq[:, 12], axis=1, keepdims=True) scale = np.linalg.norm(landmarks_seq[:, 8] - landmarks_seq[:, 12], axis=1, keepdims=True)
scale = np.maximum(scale, 1e-6) scale = np.maximum(scale, 1e-6)
landmarks_seq = landmarks_seq / scale landmarks_seq /= scale
# Flatten to (T, 63) # Flatten
seq = landmarks_seq.reshape(landmarks_seq.shape[0], -1) seq = landmarks_seq.reshape(landmarks_seq.shape[0], -1)
# Pad or truncate # Pad / truncate
if len(seq) < max_frames: if len(seq) < max_frames:
pad = np.zeros((max_frames - len(seq), seq.shape[1]), dtype=np.float32) pad = np.zeros((max_frames - len(seq), seq.shape[1]), dtype=np.float32)
seq = np.concatenate([seq, pad], axis=0) seq = np.concatenate([seq, pad])
else: else:
seq = seq[:max_frames] seq = seq[:max_frames]
@@ -84,21 +84,18 @@ def get_features_sequence(landmarks_seq, max_frames=100):
def process_row(row, base_path, max_frames=100): def process_row(row, base_path, max_frames=100):
path = os.path.join(base_path, row['path']) path = os.path.join(base_path, row["path"])
if not os.path.exists(path): if not os.path.exists(path):
return None, None return None, None
try: try:
lm_seq = extract_hand_landmarks_from_parquet(path) lm = extract_hand_landmarks_from_parquet(path)
if lm_seq is None: if lm is None:
return None, None return None, None
feat = get_features_sequence(lm, max_frames)
feat_seq = get_features_sequence(lm_seq, max_frames) if feat is None:
if feat_seq is None:
return None, None return None, None
return feat, row["sign"]
return feat_seq, row['sign'] except:
except Exception:
return None, None return None, None
@@ -123,7 +120,7 @@ class TransformerASL(nn.Module):
self.norm_in = nn.LayerNorm(d_model) self.norm_in = nn.LayerNorm(d_model)
self.pos = PositionalEncoding(d_model) self.pos = PositionalEncoding(d_model)
encoder_layer = nn.TransformerEncoderLayer( enc_layer = nn.TransformerEncoderLayer(
d_model=d_model, d_model=d_model,
nhead=nhead, nhead=nhead,
dim_feedforward=d_model * 4, dim_feedforward=d_model * 4,
@@ -132,7 +129,7 @@ class TransformerASL(nn.Module):
batch_first=True, batch_first=True,
norm_first=True norm_first=True
) )
self.encoder = nn.TransformerEncoder(encoder_layer, num_layers=num_layers) self.encoder = nn.TransformerEncoder(enc_layer, num_layers=num_layers)
self.head = nn.Sequential( self.head = nn.Sequential(
nn.LayerNorm(d_model), nn.LayerNorm(d_model),
@@ -155,7 +152,7 @@ def create_padding_mask(lengths, max_len):
def main(): def main():
# =============================== # ===============================
# DEVICE SETUP # DEVICE
# =============================== # ===============================
device = torch.device("cuda" if torch.cuda.is_available() else "cpu") device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print(f"Using device: {device}") print(f"Using device: {device}")
@@ -163,13 +160,14 @@ def main():
print("GPU:", torch.cuda.get_device_name(0)) print("GPU:", torch.cuda.get_device_name(0))
# =============================== # ===============================
# PATHS & PARAMETERS # CONFIG
# =============================== # ===============================
base_path = "asl_kaggle" # ← CHANGE THIS TO YOUR ACTUAL FOLDER base_path = "asl_kaggle" # ← CHANGE THIS TO YOUR ACTUAL PATH
max_frames = 100 max_frames = 100
MIN_SAMPLES_PER_CLASS = 6 # ← important! prevents stratified split crash
# =============================== # ===============================
# DATA PROCESSING # DATA LOADING & PROCESSING
# =============================== # ===============================
print("Loading metadata...") print("Loading metadata...")
train_df, sign_to_idx = load_kaggle_asl_data(base_path) train_df, sign_to_idx = load_kaggle_asl_data(base_path)
@@ -184,25 +182,25 @@ def main():
rows rows
), ),
total=len(rows), total=len(rows),
desc="Processing" desc="Extracting landmarks"
)) ))
X, y = [], [] X_list, y_list = [], []
for feat, sign in results: for feat, sign in results:
if feat is not None: if feat is not None:
X.append(feat) X_list.append(feat)
y.append(sign) y_list.append(sign)
if not X: if not X_list:
print("No valid sequences found!") print("No valid sequences found. Check parquet files / paths.")
return return
X = np.stack(X) X = np.stack(X_list)
print(f"Loaded {len(X)} valid samples | shape: {X.shape}") print(f"Loaded {len(X)} valid sequences | shape: {X.shape}")
# Global normalization - very important! # Global normalization (very important for stability)
print("Before global norm → mean:", X.mean(), "std:", X.std()) print("Before global norm → mean:", X.mean(), "std:", X.std())
X = np.clip(X, -5.0, 5.0) # prevent crazy outliers X = np.clip(X, -5.0, 5.0)
mean = X.mean(axis=(0, 1), keepdims=True) mean = X.mean(axis=(0, 1), keepdims=True)
std = X.std(axis=(0, 1), keepdims=True) + 1e-8 std = X.std(axis=(0, 1), keepdims=True) + 1e-8
X = (X - mean) / std X = (X - mean) / std
@@ -212,15 +210,30 @@ def main():
# LABELS # LABELS
# =============================== # ===============================
le = LabelEncoder() le = LabelEncoder()
y = le.fit_transform(y_list)
# Remove classes with too few samples (prevents stratify error)
counts = Counter(y)
valid_classes = [cls for cls, cnt in counts.items() if cnt >= MIN_SAMPLES_PER_CLASS]
mask = np.isin(y, valid_classes)
X = X[mask]
y = y[mask]
# Re-encode labels consecutively (0,1,2,... no gaps)
le = LabelEncoder()
y = le.fit_transform(y) y = le.fit_transform(y)
num_classes = len(le.classes_)
print(f"Number of classes: {num_classes}") print(f"After filtering: {len(X)} samples remain | {len(le.classes_)} classes")
# =============================== # ===============================
# SPLIT # SPLIT
# =============================== # ===============================
X_train, X_test, y_train, y_test = train_test_split( X_train, X_test, y_train, y_test = train_test_split(
X, y, test_size=0.15, stratify=y, random_state=42 X, y,
test_size=0.15,
stratify=y, # should be safe now
random_state=42
) )
# =============================== # ===============================
@@ -258,7 +271,7 @@ def main():
# =============================== # ===============================
model = TransformerASL( model = TransformerASL(
input_dim=63, input_dim=63,
num_classes=num_classes, num_classes=len(le.classes_),
d_model=192, d_model=192,
nhead=6, nhead=6,
num_layers=4 num_layers=4
@@ -274,18 +287,15 @@ def main():
scheduler = optim.lr_scheduler.CosineAnnealingWarmRestarts(optimizer, T_0=10) scheduler = optim.lr_scheduler.CosineAnnealingWarmRestarts(optimizer, T_0=10)
# =============================== # ===============================
# TRAIN / EVAL FUNCTIONS # TRAIN / EVAL
# =============================== # ===============================
def train_epoch(): def train_epoch():
model.train() model.train()
total_loss = 0 total_loss = 0
correct = 0 correct = total = 0
total = 0
for x, y in tqdm(train_loader, desc="Training"): for x, y in tqdm(train_loader, desc="Train"):
x, y = x.to(device), y.to(device) x, y = x.to(device), y.to(device)
# Rough length estimation
lengths = (x.abs().sum(dim=2) > 1e-5).sum(dim=1) lengths = (x.abs().sum(dim=2) > 1e-5).sum(dim=1)
mask = create_padding_mask(lengths, x.size(1)) mask = create_padding_mask(lengths, x.size(1))
@@ -297,13 +307,10 @@ def main():
grad_norm = torch.nn.utils.clip_grad_norm_(model.parameters(), max_norm=0.8) grad_norm = torch.nn.utils.clip_grad_norm_(model.parameters(), max_norm=0.8)
if torch.isnan(loss) or grad_norm > 20:
print(f"Warning - large grad or NaN! norm = {grad_norm:.2f}")
optimizer.step() optimizer.step()
total_loss += loss.item() total_loss += loss.item()
correct += (logits.argmax(dim=-1) == y).sum().item() correct += (logits.argmax(-1) == y).sum().item()
total += y.size(0) total += y.size(0)
return total_loss / len(train_loader), correct / total * 100 return total_loss / len(train_loader), correct / total * 100
@@ -311,31 +318,30 @@ def main():
@torch.no_grad() @torch.no_grad()
def evaluate(): def evaluate():
model.eval() model.eval()
correct = 0 correct = total = 0
total = 0
for x, y in test_loader: for x, y in test_loader:
x, y = x.to(device), y.to(device) x, y = x.to(device), y.to(device)
lengths = (x.abs().sum(dim=2) > 1e-5).sum(dim=1) lengths = (x.abs().sum(dim=2) > 1e-5).sum(dim=1)
mask = create_padding_mask(lengths, x.size(1)) mask = create_padding_mask(lengths, x.size(1))
logits = model(x, key_padding_mask=mask) logits = model(x, key_padding_mask=mask)
correct += (logits.argmax(dim=-1) == y).sum().item() correct += (logits.argmax(-1) == y).sum().item()
total += y.size(0) total += y.size(0)
return correct / total * 100 if total > 0 else 0 return correct / total * 100 if total > 0 else 0.0
# =============================== # ===============================
# TRAINING LOOP # TRAINING LOOP
# =============================== # ===============================
best_acc = 0 best_acc = 0.0
patience = 15 patience = 15
wait = 0 wait = 0
epochs = 60 epochs = 70
for epoch in range(epochs): for epoch in range(epochs):
loss, train_acc = train_epoch() loss, train_acc = train_epoch()
test_acc = evaluate() test_acc = evaluate()
print(f"Epoch {epoch + 1:2d}/{epochs} | Loss: {loss:.4f} | Train: {train_acc:.2f}% | Test: {test_acc:.2f}%") print(f"[{epoch + 1:2d}/{epochs}] loss: {loss:.4f} | train: {train_acc:.2f}% | test: {test_acc:.2f}%")
scheduler.step() scheduler.step()
@@ -345,18 +351,18 @@ def main():
torch.save({ torch.save({
'model': model.state_dict(), 'model': model.state_dict(),
'optimizer': optimizer.state_dict(), 'optimizer': optimizer.state_dict(),
'label_encoder': le.classes_, 'label_encoder_classes': le.classes_,
'epoch': epoch, 'acc': best_acc,
'acc': best_acc 'epoch': epoch
}, "best_asl_transformer.pth") }, "best_asl_transformer.pth")
print(" → New best model saved") print(" → New best saved")
else: else:
wait += 1 wait += 1
if wait >= patience: if wait >= patience:
print("Early stopping triggered") print("Early stopping")
break break
print(f"\nTraining finished. Best test accuracy: {best_acc:.2f}%") print(f"\nBest test accuracy reached: {best_acc:.2f}%")
if __name__ == '__main__': if __name__ == '__main__':