fix(SteamMachinist): fix inconsistent parameter name in MuZeroCollector and MuZeroPolicy (#463)

* fix: rename param x->data in policy forward call

* fix: fix uninitialized variable collected_step

---------

Co-authored-by: Iaroslav Roshchupkin <iroshchupkin@neoflex.ru>

Author: SteamMachinist

lzero/worker/muzero_collector.py

@@ -340,6 +340,7 @@ def collect(
 
         # --- Initializations ---
         collected_episode = 0
+        collected_step = 0
         env_nums = self._env_num
         retry_waiting_time = 0.05
 
@@ -411,7 +412,7 @@ def collect(
                 # Policy Forward Pass
                 # ==============================================================
                 policy_input = {
-                    'x': stack_obs_tensor,
+                    'data': stack_obs_tensor,
                     'action_mask': action_mask,
                     'temperature': temperature,
                     'to_play': to_play,
@@ -679,4 +680,4 @@ def _output_log(self, train_iter: int) -> None:
 
             if self.policy_config.use_wandb:
                 wandb_log_data = {tb_prefix_step + k: v for k, v in info.items()}
-                wandb.log(wandb_log_data, step=self._total_envstep_count)
+                wandb.log(wandb_log_data, step=self._total_envstep_count)