MiniAxion1-0.9M / model.py
AxionLab-official's picture
Update model.py
2ee3fce verified
"""
Nano Reasoning Model (NRM) - Main Architecture
ARCHITECTURE DESIGN PHILOSOPHY:
================================
This model maximizes reasoning ability per parameter through several key innovations:
1. SHARED LAYERS: The middle layers are shared (looped through multiple times).
This creates a form of "iterative refinement" - the model processes information
multiple passes, similar to how recurrent networks process sequences but applied
to depth instead. This is inspired by Universal Transformers and ALBERT.
WHY IT HELPS REASONING: Reasoning often requires iterative refinement of
intermediate representations. Shared layers let the model "think more" without
more parameters.
2. THINKING TOKENS: Special <THINK> and </THINK> tokens create a "scratchpad"
where the model can show intermediate reasoning steps. The model is trained to
use <STEP> tokens for each logical step.
WHY IT HELPS: Decomposing complex problems into steps is THE key capability
for reasoning. Even large models benefit from chain-of-thought prompting.
3. WEIGHT TYING: Input and output embeddings share the same weight matrix.
This halves the embedding parameter count and creates a natural link between
token understanding and token generation.
WHY IT HELPS CPU: Fewer parameters = faster forward/backward passes.
4. LOW-RANK PROJECTIONS: All attention and MLP projections use LoRA-style
factored matrices, cutting parameter count by ~8x in linear layers.
5. GROUPED QUERY ATTENTION: KV heads are shared across query heads,
reducing KV projection parameters and memory.
PARAMETER BUDGET (~10M):
Embedding: 2048 * 256 = 524K (shared with output head)
Per unique layer: ~200K
4 unique + 2 shared (run 2x) = 6 effective layers
Total: ~2.1M (layers) + 524K (embed) ≈ 2.6M unique params
Effective computation: ~3.1M param equivalent
"""
import torch
import torch.nn as nn
import torch.nn.functional as F
from typing import Optional, Dict
from components import TransformerBlock, RMSNorm
class NanoReasoningModel(nn.Module):
def __init__(self, config: dict):
super().__init__()
self.config = config
d_model = config['d_model']
n_heads = config['n_heads']
n_layers = config['n_layers']
n_shared = config.get('n_shared_layers', 2)
d_ff = config['d_ff']
vocab_size = config['vocab_size']
max_seq_len = config['max_seq_len']
dropout = config.get('dropout', 0.05)
rank = config.get('lora_rank', 16)
self.use_thinking = config.get('use_thinking_tokens', True)
self.n_thinking_steps = config.get('n_thinking_steps', 2)
n_kv_heads = config.get('n_kv_heads', n_heads // 2)
# Token embeddings (will be tied with output head)
self.token_embedding = nn.Embedding(vocab_size, d_model)
self.embedding_dropout = nn.Dropout(dropout)
# Entry layers (unique)
n_unique = n_layers - n_shared
self.entry_layers = nn.ModuleList([
TransformerBlock(d_model, n_heads, d_ff, rank, dropout, max_seq_len, n_kv_heads)
for _ in range(n_unique // 2)
])
# Shared layers (looped)
self.shared_layers = nn.ModuleList([
TransformerBlock(d_model, n_heads, d_ff, rank, dropout, max_seq_len, n_kv_heads)
for _ in range(n_shared)
])
# Exit layers (unique)
self.exit_layers = nn.ModuleList([
TransformerBlock(d_model, n_heads, d_ff, rank, dropout, max_seq_len, n_kv_heads)
for _ in range(n_unique - n_unique // 2)
])
# Final norm
self.final_norm = RMSNorm(d_model)
# Output head (tied with embeddings)
self.output_head = nn.Linear(d_model, vocab_size, bias=False)
if config.get('weight_tying', True):
self.output_head.weight = self.token_embedding.weight
# Thinking step gate: learned scalar for blending thinking iterations
if self.use_thinking:
self.think_gate = nn.Parameter(torch.tensor(0.5))
# Initialize weights
self.apply(self._init_weights)
# Count parameters
self._count_parameters()
def _init_weights(self, module: nn.Module):
"""Initialize weights with scaled initialization for stability."""
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 _count_parameters(self):
"""Count and report parameters."""
total = sum(p.numel() for p in self.parameters())
trainable = sum(p.numel() for p in self.parameters() if p.requires_grad)
# Count unique parameters (shared layers counted once)
unique = sum(p.numel() for p in self.parameters())
self.total_params = total
self.trainable_params = trainable
print(f"\n{'='*50}")
print(f"NRM Model Configuration:")
print(f" d_model: {self.config['d_model']}")
print(f" n_heads: {self.config['n_heads']}")
print(f" n_layers: {self.config['n_layers']} "
f"({len(self.entry_layers)} entry + {len(self.shared_layers)} shared + {len(self.exit_layers)} exit)")
print(f" d_ff: {self.config['d_ff']}")
print(f" vocab_size: {self.config['vocab_size']}")
print(f" LoRA rank: {self.config.get('lora_rank', 16)}")
print(f" Thinking: {'enabled' if self.use_thinking else 'disabled'}")
print(f" Total parameters: {total:,}")
print(f" Trainable parameters: {trainable:,}")
print(f"{'='*50}\n")
def forward(self, input_ids: torch.Tensor,
attention_mask: Optional[torch.Tensor] = None,
labels: Optional[torch.Tensor] = None,
n_think_loops: int = 1) -> Dict[str, torch.Tensor]:
"""
Forward pass with optional thinking loops.
n_think_loops: How many times to loop through shared layers.
During reasoning, we increase this to give the model more "thinking time".
"""
B, T = input_ids.shape
# Embeddings
x = self.token_embedding(input_ids)
x = self.embedding_dropout(x)
# Padding mask
pad_mask = None
if attention_mask is not None:
pad_mask = (attention_mask == 0) # True where padded
# Entry layers
for layer in self.entry_layers:
x = layer(x, pad_mask)
# Shared layers with thinking loops
actual_loops = max(1, n_think_loops)
if self.use_thinking and actual_loops > 1:
# Store the "pre-think" state
x_original = x
for loop in range(actual_loops):
for layer in self.shared_layers:
x = layer(x, pad_mask)
if loop < actual_loops - 1:
# Blend with original (residual thinking)
gate = torch.sigmoid(self.think_gate)
x = gate * x + (1 - gate) * x_original
else:
for layer in self.shared_layers:
x = layer(x, pad_mask)
# Exit layers
for layer in self.exit_layers:
x = layer(x, pad_mask)
# Output
x = self.final_norm(x)
logits = self.output_head(x)
result = {"logits": logits}
if labels is not None:
# Shift for autoregressive loss
shift_logits = logits[:, :-1, :].contiguous()
shift_labels = labels[:, 1:].contiguous()
loss = F.cross_entropy(
shift_logits.view(-1, shift_logits.size(-1)),
shift_labels.view(-1),
ignore_index=0, # PAD token
label_smoothing=0.05 # Slight smoothing for better generalization
)
result["loss"] = loss
return result
@torch.no_grad()
def generate(self, input_ids: torch.Tensor, max_new_tokens: int = 100,
temperature: float = 0.7, top_k: int = 50, top_p: float = 0.9,
n_think_loops: int = 1, eos_token_id: int = 2) -> torch.Tensor:
"""
Autoregressive generation with temperature, top-k, and top-p sampling.
Uses nucleus (top-p) sampling for diverse but coherent generation.
"""
self.eval()
generated = input_ids.clone()
for _ in range(max_new_tokens):
# Truncate to max_seq_len
context = generated[:, -self.config['max_seq_len']:]
outputs = self.forward(context, n_think_loops=n_think_loops)
logits = outputs["logits"][:, -1, :] / max(temperature, 1e-5)
# Top-k filtering
if top_k > 0:
top_k_val = min(top_k, logits.size(-1))
indices_to_remove = logits < torch.topk(logits, top_k_val)[0][..., -1, None]
logits[indices_to_remove] = float('-inf')
# Top-p (nucleus) filtering
if top_p < 1.0:
sorted_logits, sorted_indices = torch.sort(logits, descending=True)
cumulative_probs = torch.cumsum(F.softmax(sorted_logits, dim=-1), dim=-1)
sorted_indices_to_remove = cumulative_probs > top_p
sorted_indices_to_remove[..., 1:] = sorted_indices_to_remove[..., :-1].clone()
sorted_indices_to_remove[..., 0] = 0
indices_to_remove = sorted_indices_to_remove.scatter(
1, sorted_indices, sorted_indices_to_remove)
logits[indices_to_remove] = float('-inf')
probs = F.softmax(logits, dim=-1)
next_token = torch.multinomial(probs, num_samples=1)
generated = torch.cat([generated, next_token], dim=1)
if next_token.item() == eos_token_id:
break
return generated
def save(self, path: str):
"""Save model state dict and config."""
import os, json
os.makedirs(path, exist_ok=True)
torch.save(self.state_dict(), os.path.join(path, "model.pt"))
with open(os.path.join(path, "config.json"), 'w') as f:
json.dump(self.config, f, indent=2)
print(f"Model saved to {path}")
@classmethod
def load(cls, path: str, device: str = 'cpu') -> 'NanoReasoningModel':
"""Load model from saved state."""
import os, json
with open(os.path.join(path, "config.json"), 'r') as f:
config = json.load(f)
model = cls(config)
model.load_state_dict(torch.load(os.path.join(path, "model.pt"),
map_location=device))
return model