Upload 3 files

2Gv37_AutoLR/emoairy.py

@@ -31,9 +31,9 @@ class EmoAiry(Optimizer):
     # 感情EMA更新(緊張と安静)
     def _update_ema(self, state, loss_val):
         ema = state.setdefault('ema', {})
-        ema['short']  = 0.3  * loss_val + 0.7  * ema.get('short', loss_val)
+        ema['short'] = 0.3 * loss_val + 0.7 * ema.get('short', loss_val)
         ema['medium'] = 0.05 * loss_val + 0.95 * ema.get('medium', loss_val)
-        ema['long']   = 0.01 * loss_val + 0.99 * ema.get('long', loss_val)
+        ema['long'] = 0.01 * loss_val + 0.99 * ema.get('long', loss_val)
         return ema
 
     # 感情スカラー値生成(EMA差分、滑らかな非線形スカラー、tanh(diff) は ±1.0 で有界性)
@@ -154,8 +154,8 @@ class EmoAiry(Optimizer):
                     update_term = grad / denom
 
                 # 最終的なパラメータ更新 (decoupled weight decayも適用)
-                # sign化で２次momentと１次ベクトルのバランス改善
-                p.add_(p, alpha=-group['weight_decay'] * emoPulse)
+                # sign化で２次momentと１次ベクトルのデータの質(粒度)を揃える
+                p.mul_(1.0 - group['weight_decay'] * emoPulse)
                 p.add_(update_term.sign_(), alpha=-emoPulse)
                 # --- End Gradient Update Logic ---
 

2Gv37_AutoLR/emocats.py

@@ -1,7 +1,6 @@
 import torch
 from torch.optim import Optimizer
 import math
-from typing import Callable
 
 """
 EmoCats v3.7.6 (260109) shadow-system v3.1 -moment v3.1 emoPulse v3.7
@@ -11,18 +10,14 @@ dNR係数により emoPulse に履歴を混ぜて安定させた(d / N 履歴
 Early scalar、Early Stop、効率化しつつ精度向上させ負荷も軽減する等の改修と微調整
 """
 
-# Helper function
-def exists(val):
-    return val is not None
-
 class EmoCats(Optimizer):
-    # クラス定義＆初期化 ベータ･互換性の追加
-    def __init__(self, params,
-                 lr=1.0,
-                 eps=1e-8,
-                 betas=(0.9, 0.995),
-                 weight_decay=0.01,
-                 use_shadow: bool = False):
+    # クラス定義＆初期化
+    def __init__(self, params, 
+                 lr=1.0, 
+                 eps=1e-8, 
+                 betas=(0.9, 0.995), 
+                 weight_decay=0.01, 
+                 use_shadow:bool=False): 
         defaults = dict(lr=lr, betas=betas, eps=eps, weight_decay=weight_decay)
         super().__init__(params, defaults)
         self._init_lr = lr
@@ -36,9 +31,9 @@ class EmoCats(Optimizer):
     # 感情EMA更新(緊張と安静)
     def _update_ema(self, state, loss_val):
         ema = state.setdefault('ema', {})
-        ema['short']  = 0.3  * loss_val + 0.7  * ema.get('short', loss_val)
+        ema['short'] = 0.3 * loss_val + 0.7 * ema.get('short', loss_val)
         ema['medium'] = 0.05 * loss_val + 0.95 * ema.get('medium', loss_val)
-        ema['long']   = 0.01 * loss_val + 0.99 * ema.get('long', loss_val)
+        ema['long'] = 0.01 * loss_val + 0.99 * ema.get('long', loss_val)
         return ema
 
     # 感情スカラー値生成(EMA差分、滑らかな非線形スカラー、tanh(diff) は ±1.0 で有界性)
@@ -75,11 +70,8 @@ class EmoCats(Optimizer):
 
     # 損失取得(損失値 loss_val を数値化、感情判定に使用、存在しないパラメータ(更新不要)はスキップ)
     @torch.no_grad()
-    def step(self, closure: Callable | None = None): # クロージャの型ヒントを追加
-        loss = None
-        if exists(closure): # 一貫性のためにexistsヘルパーを使う
-            with torch.enable_grad():
-                loss = closure()
+    def step(self, closure=None): 
+        loss = torch.enable_grad()(closure)() if closure is not None else None
         loss_val = loss.item() if loss is not None else 0.0
 
         # EMA更新・スカラー生成(EMA差分からスカラーを生成しスパイク比率等を決定)
@@ -112,10 +104,8 @@ class EmoCats(Optimizer):
         # --- End emoPulse (完全自動LR生成) ---
 
         for group in self.param_groups:
-            # 共通パラメータ抽出
-            _wd_actual, beta1, beta2 = group['weight_decay'], * group['betas']
-            # PGチェックにフィルタ
-            for p in filter(lambda p: exists(p.grad), group['params']):
+            beta1, beta2 = group['betas']
+            for p in (p for p in group['params'] if p.grad is not None):
 
                 grad = p.grad
                 state = self.state[p]
@@ -140,15 +130,14 @@ class EmoCats(Optimizer):
                     state['exp_avg'] = torch.zeros_like(p)
                 exp_avg = state['exp_avg']
 
-                # Stepweight decay : decoupled_wd
-                p.mul_(1 - emoPulse * _wd_actual)
-                beta1, beta2 = group['betas']
+                # decoupled weight decay
+                p.mul_(1.0 - group['weight_decay'] * emoPulse)
 
                 # 勾配ブレンド
                 blended_grad = grad.mul(1 - beta1).add(exp_avg, alpha=beta1)
 
                 # 最終的なパラメータ更新
-                p.add_(blended_grad.sign(), alpha = -emoPulse)
+                p.add_(blended_grad.sign_(), alpha = -emoPulse)
                 exp_avg.mul_(beta2).add_(grad, alpha = 1 - beta2)
                 # --- End Gradient Update Logic ---
 
@@ -171,5 +160,4 @@ class EmoCats(Optimizer):
  https://github.com/muooon/EmoSens
  Cats was developed with inspiration from Lion, Tiger, and emolynx,
  which we deeply respect for their lightweight and intelligent design.
- Cats also integrates EmoNAVI to enhance its capabilities.
 """

2Gv37_AutoLR/emosens.py

@@ -31,9 +31,9 @@ class EmoSens(Optimizer):
     # 感情EMA更新(緊張と安静)
     def _update_ema(self, state, loss_val):
         ema = state.setdefault('ema', {})
-        ema['short']  = 0.3  * loss_val + 0.7  * ema.get('short', loss_val)
+        ema['short'] = 0.3 * loss_val + 0.7 * ema.get('short', loss_val)
         ema['medium'] = 0.05 * loss_val + 0.95 * ema.get('medium', loss_val)
-        ema['long']   = 0.01 * loss_val + 0.99 * ema.get('long', loss_val)
+        ema['long'] = 0.01 * loss_val + 0.99 * ema.get('long', loss_val)
         return ema
 
     # 感情スカラー値生成(EMA差分、滑らかな非線形スカラー、tanh(diff) は ±1.0 で有界性)
@@ -136,7 +136,7 @@ class EmoSens(Optimizer):
                 denom = exp_avg_sq.sqrt().add_(group['eps'])
 
                 if group['weight_decay']:
-                    p.add_(p, alpha=-group['weight_decay'] * emoPulse)
+                    p.mul_(1.0 - group['weight_decay'] * emoPulse)
                 p.addcdiv_(exp_avg, denom, value=-emoPulse)
                 # --- End Gradient Update Logic ---