adding naive jax and libraries and make adjustments

engine/sweeps/small_arch_compare.yaml

@@ -0,0 +1,86 @@
+method: random
+metric:
+  name: sweep/score
+  goal: maximize
+command:
+  - ${env}
+  - python
+  - -m
+  - engine.train
+parameters:
+  algo:
+    values: [ppo, a2c, dqn, qtable]
+  arch:
+    values: [tiny, small, medium]
+  activation:
+    values: [relu, tanh]
+  total_timesteps:
+    values: [8000, 12000, 20000]
+  seed:
+    values: [13, 42, 77]
+  n_products:
+    values: [6, 8, 10]
+  alpha:
+    distribution: uniform
+    min: 0.1
+    max: 0.5
+  lambda_coi:
+    distribution: uniform
+    min: 0.05
+    max: 0.4
+  robust_radius:
+    values: [0.0, 0.1, 0.2]
+  robust_points:
+    values: [3, 5]
+  info_value:
+    values: [0.75, 1.0, 1.5]
+  revenue_weight:
+    values: [0.005, 0.01, 0.02]
+  learning_rate:
+    distribution: log_uniform_values
+    min: 1.0e-5
+    max: 5.0e-4
+  gamma:
+    values: [0.98, 0.99]
+  buffer_size:
+    values: [10000, 30000, 50000]
+  batch_size:
+    values: [64, 128, 256]
+  tau:
+    values: [0.002, 0.005, 0.01]
+  train_freq:
+    values: [1, 4]
+  learning_starts:
+    values: [500, 1000, 2000]
+  n_steps:
+    values: [256, 512, 1024]
+  n_epochs:
+    values: [5, 10]
+  gae_lambda:
+    values: [0.9, 0.95]
+  clip_range:
+    values: [0.1, 0.2]
+  ent_coef:
+    values: [0.0, 0.005]
+  target_update_interval:
+    values: [500, 1000]
+  exploration_fraction:
+    values: [0.1, 0.2]
+  exploration_final_eps:
+    values: [0.02, 0.05]
+  action_levels:
+    values: [5, 7, 9]
+  action_scale_low:
+    values: [0.85, 0.9]
+  action_scale_high:
+    values: [1.1, 1.15]
+  q_lr:
+    values: [0.05, 0.1, 0.2]
+  q_bins:
+    values: [4, 6, 8]
+  eps_start:
+    value: 1.0
+  eps_end:
+    values: [0.02, 0.05]
+  eps_decay:
+    values: [0.999, 0.9995]