Make MineRL evaluation more generic

src/minecraft/getting_started_minerl.jl

@@ -17,9 +17,9 @@ minerl_grammar = @pcsgrammar begin
 end
 
 minerl_grammar_2 = @pcsgrammar begin
+    1:SEQ = [ACT]
     8:DIR = 0b0001 | 0b0010 | 0b0100 | 0b1000 | 0b0101 | 0b1001 | 0b0110 | 0b1010 # forward | back | left | right | forward-left | forward-right | back-left | back-right
-    1:sequence_actions = [action]
-    1:action = (TIMES, DIR)
+    1:ACT = (TIMES, Dict("move" => DIR, "sprint" => 1, "jump" => 1))
     6:TIMES = 5 | 10 | 25 | 50 | 75 | 100
 end
 
@@ -32,16 +32,18 @@ function evaluate_trace_minerl(prog, grammar, env, show_moves)
     sum_of_rewards = 0
     is_done = false
     obs = nothing
-    for saved_action ∈ sequence_of_actions
-        times, dir = saved_action
-
+    for (times, action) ∈ sequence_of_actions
         new_action = env.action_space.noop()
-        new_action["forward"] = dir & 1
-        new_action["back"] = dir >> 1 & 1
-        new_action["left"] = dir >> 2 & 1
-        new_action["right"] = dir >> 3
-        new_action["sprint"] = 1
-        new_action["jump"] = 1
+        for (key, val) in action
+            if key == "move"
+                new_action["forward"] = val & 1
+                new_action["back"] = val >> 1 & 1
+                new_action["left"] = val >> 2 & 1
+                new_action["right"] = val >> 3
+            else
+                new_action[key] = val
+            end
+        end
 
         for i in 1:times
             obs, reward, done, _ = env.step(new_action)
@@ -76,5 +78,5 @@ HerbSearch.evaluate_trace(prog::RuleNode, grammar::ContextSensitiveGrammar; show
 HerbSearch.calculate_rule_cost(rule_index::Int, grammar::ContextSensitiveGrammar) = HerbSearch.calculate_rule_cost_prob(rule_index, grammar)
 
 # resetEnv()
-iter = HerbSearch.GuidedTraceSearchIterator(minerl_grammar_2, :sequence_actions)
+iter = HerbSearch.GuidedTraceSearchIterator(minerl_grammar_2, :SEQ)
 program = @time probe(Vector{Trace}(), iter, 3000000, 6)

src/probe/update_grammar.jl

@@ -54,9 +54,9 @@ function update_grammar!(grammar::ContextSensitiveGrammar, PSols_with_eval_cache
         # TODO: think about better thing here
         fitness = min(best_reward / 100, 1)
 
-        p_current = 2 ^ (grammar.log_probabilities[rule_index])
+        p_current = 2^(grammar.log_probabilities[rule_index])
 
-        sum +=  p_current^(1 - fitness)
+        sum += p_current^(1 - fitness)
         log_prob = ((1 - fitness) * log(2, p_current))
         grammar.log_probabilities[rule_index] = log_prob
     end
@@ -66,7 +66,7 @@ function update_grammar!(grammar::ContextSensitiveGrammar, PSols_with_eval_cache
         total_sum += 2^(grammar.log_probabilities[rule_index])
     end
     expr = rulenode2expr(PSols_with_eval_cache[begin].program, grammar)
-    grammar.rules[9] = :([$expr; action])
+    grammar.rules[1] = :([$expr; ACT])
     @assert abs(total_sum - 1) <= 1e-4 "Total sum is $(total_sum) "
 end