Upload latest retained checkpoint envstep_300000.pth.tar

metadata/total_config.py

@@ -171,9 +171,9 @@ exp_config = {
         'use_augmentation': False,
         'augmentation': ['shift', 'intensity'],
         'ignore_done': False,
-        'update_per_collect': 10,
+        'update_per_collect': 50,
         'replay_ratio': 0.25,
-        'batch_size': 128,
+        'batch_size': 256,
         'optim_type': 'Adam',
         'learning_rate': 0.003,
         'target_update_freq': 100,
@@ -183,7 +183,7 @@ exp_config = {
         'grad_clip_value': 0.5,
         'n_episode': 4,
         'num_segments': 8,
-        'num_simulations': 20,
+        'num_simulations': 50,
         'discount_factor': 1,
         'td_steps': 81,
         'num_unroll_steps': 5,
@@ -219,10 +219,10 @@ exp_config = {
         'type': 'efficientzero',
         'import_names': ['lzero.policy.efficientzero'],
         'model_path':
-        '/workspace/combinatorial_reasoning_post_training/models/simplified5_constant_lr_fixed_20260519/round-01/three_six_nine/attempt-01_260519_093513/ckpt/envstep_90000.pth.tar',
+        '/workspace/combinatorial_reasoning_post_training/models/simplified5_two_machine_continuation_20260520/shard-0/three_six_nine/attempt-01_260520_133229/ckpt/envstep_10000.pth.tar',
         'reanalyze_ratio': 0.0,
-        'eval_freq': 100001,
-        'replay_buffer_size': 50000,
+        'eval_freq': 300001,
+        'replay_buffer_size': 100000,
         'best_ckpt_strategy': 'raw',
         'best_ckpt_ema_alpha': 0.3,
         'best_ckpt_min_episodes': 20,
@@ -241,6 +241,6 @@ exp_config = {
         'device': 'cuda'
     },
     'exp_name':
-    'models/three_six_nine_dual_eval_continuation_20260519/round-01/three_six_nine/attempt-01_260519_131901',
+    '/workspace/combinatorial_reasoning_post_training/models/simplified5_two_machine_continuation_20260520/shard-0/three_six_nine/attempt-03_260520_145623',
     'seed': 0
 }