File size: 5,606 Bytes
3c27aeb |
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 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 |
#!/usr/bin/env python3
"""
Basic BitTransformerLM Training Script
=====================================
A simple working training script that follows the ACTUAL BitTransformerLM
model implementation exactly as it exists in the codebase.
"""
import sys
import os
import logging
import torch
import torch.nn.functional as F
# Add paths for imports
sys.path.append('/data')
sys.path.append('/data/BitTransformerLM')
from bit_transformer import BitTransformerLM, text_to_bits
from BTLM_Extensions import configure_adafactor_optimizer
# Setup logging
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
logger = logging.getLogger(__name__)
def create_simple_dataset():
"""Create a simple bit dataset for testing."""
logger.info("Creating simple bit dataset...")
# Use some simple text examples
texts = [
"Hello world! This is a test.",
"BitTransformerLM processes bits natively.",
"Training on binary sequences is interesting.",
"Each character becomes 9 bits with parity.",
"The model learns bit patterns directly.",
]
# Convert to bits
bit_sequences = []
for text in texts:
bits = text_to_bits(text)
bit_sequences.append(bits)
# Pad to same length and create training data
max_len = min(64, max(len(bits) for bits in bit_sequences)) # Keep it small for testing
training_data = []
for bits in bit_sequences:
if len(bits) >= max_len:
# Take chunks of max_len
for i in range(0, len(bits) - max_len + 1, max_len // 2):
chunk = bits[i:i + max_len]
if len(chunk) == max_len:
training_data.append(chunk)
# Convert to tensor
data_tensor = torch.tensor(training_data, dtype=torch.long)
logger.info(f"Created dataset: {data_tensor.shape}")
return data_tensor
def create_model():
"""Create a small BitTransformerLM model for testing."""
logger.info("Creating BitTransformerLM model...")
# Small model configuration for basic testing
model = BitTransformerLM(
d_model=128,
nhead=8,
num_layers=2,
dim_feedforward=256,
max_seq_len=64,
lambda_K=0.1,
lambda_C=0.1,
lambda_S=0.1,
use_checkpoint=False, # Disable for simplicity
use_autocast=False, # Disable for simplicity
use_act=False # Disable for simplicity
)
total_params = sum(p.numel() for p in model.parameters())
logger.info(f"Model created: {total_params:,} parameters")
return model
def train_basic():
"""Basic training loop following the example_training_step pattern."""
logger.info("Starting basic BitTransformerLM training...")
# Create model and data
model = create_model()
data = create_simple_dataset()
# Calculate total steps
batch_size = 2
epochs = 5
total_steps = (len(data) // batch_size) * epochs
# Configure optimizer using Fixed LR Adafactor (breakthrough config)
logger.info("Configuring Fixed RL Adafactor optimizer...")
optimizer, scheduler = configure_adafactor_optimizer(
model,
lr=1e-3, # FIXED learning rate - key to breakthrough!
weight_decay=0.01,
total_steps=total_steps
)
logger.info("Starting training loop...")
# Training configuration
model.train()
for epoch in range(epochs):
epoch_losses = []
# Simple batching
for i in range(0, len(data), batch_size):
batch = data[i:i + batch_size]
if len(batch) < batch_size:
continue # Skip incomplete batches
# Zero gradients
optimizer.zero_grad()
# Forward pass - EXACTLY like example_training_step
logits, telemetry = model(batch)
# Loss calculation - EXACTLY like example_training_step
pred = logits[:, :-1, :].reshape(-1, 2)
target = batch[:, 1:].reshape(-1)
loss = F.cross_entropy(pred, target)
# Backward pass
loss.backward()
# Gradient clipping
torch.nn.utils.clip_grad_norm_(model.parameters(), 1.0)
# Optimizer step
optimizer.step()
if scheduler:
scheduler.step()
epoch_losses.append(loss.item())
# Log epoch results
avg_loss = sum(epoch_losses) / len(epoch_losses) if epoch_losses else float('inf')
logger.info(f"Epoch {epoch + 1}/{epochs}: Average Loss = {avg_loss:.6f}")
# Log telemetry if available
if telemetry:
for key, value in telemetry.items():
if torch.is_tensor(value):
logger.info(f" {key}: {value.mean().item():.4f}")
logger.info("Basic training completed successfully!")
return model
def main():
"""Main function."""
logger.info("🚀 Starting basic BitTransformerLM training test")
try:
trained_model = train_basic()
logger.info("✅ Basic training test PASSED!")
# Save the model
torch.save(trained_model.state_dict(), '/data/BitTransformerLM/basic_model.pt')
logger.info("Model saved to basic_model.pt")
except Exception as e:
logger.error(f"❌ Training failed: {e}")
raise
if __name__ == "__main__":
main() |