rnow 0.2.4__py3-none-any.whl → 0.3.9__py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (22) hide show
  1. rnow/cli/commands.py +226 -84
  2. rnow/cli/test.py +536 -441
  3. rnow/core/__init__.py +4 -1
  4. rnow/core/reward.py +34 -3
  5. rnow/core/tool.py +29 -7
  6. rnow/models.py +88 -6
  7. rnow/templates/deepseek-aha/config.yml +1 -1
  8. rnow/templates/mcp-tavily/config.yml +1 -1
  9. rnow/templates/rl-single/config.yml +7 -7
  10. rnow/templates/rl-single/train.jsonl +0 -908
  11. rnow/templates/rl-tools/config.yml +1 -1
  12. rnow/templates/tutorial-reward/config.yml +7 -7
  13. rnow/templates/tutorial-reward/train.jsonl +0 -908
  14. rnow/templates/tutorial-tool/config.yml +1 -1
  15. {rnow-0.2.4.dist-info → rnow-0.3.9.dist-info}/METADATA +23 -9
  16. {rnow-0.2.4.dist-info → rnow-0.3.9.dist-info}/RECORD +22 -22
  17. /rnow/templates/rl-tools/{env.py → tools.py} +0 -0
  18. /rnow/templates/tutorial-tool/{env.py → tools.py} +0 -0
  19. {rnow-0.2.4.dist-info → rnow-0.3.9.dist-info}/WHEEL +0 -0
  20. {rnow-0.2.4.dist-info → rnow-0.3.9.dist-info}/entry_points.txt +0 -0
  21. {rnow-0.2.4.dist-info → rnow-0.3.9.dist-info}/licenses/LICENSE +0 -0
  22. {rnow-0.2.4.dist-info → rnow-0.3.9.dist-info}/top_level.txt +0 -0
rnow/templates/rl-tools/config.yml CHANGED
@@ -9,7 +9,7 @@ data:
9
9
  batch_size: 32
10
10
  group_size: 16
11
11
  model:
12
- path: Qwen/Qwen3-8B
12
+ path: Qwen/Qwen3-30B-A3B-Instruct-2507
13
13
  qlora_rank: 32
14
14
  name: "SimpleQA Agent"
15
15
  description: "Multi-turn RL model trained on SimpleQA factual questions with tool use"
rnow/templates/tutorial-reward/config.yml CHANGED
@@ -1,13 +1,13 @@
1
1
  project_id: ""
2
- project_name: "OpenMathReasoning"
2
+ project_name: "rl-project"
3
3
  dataset_id: ""
4
- dataset_name: "math-problems"
4
+ dataset_name: "train"
5
5
  dataset_type: rl
6
6
  organization_id: ""
7
7
  data:
8
8
  train_file: train.jsonl
9
- batch_size: 32
10
- group_size: 16
9
+ batch_size: 16
10
+ group_size: 8
11
11
  model:
12
12
  path: Qwen/Qwen3-8B
13
13
  qlora_rank: 32
@@ -19,9 +19,9 @@ algorithm:
19
19
  kl_penalty_coef: 0.01
20
20
  rollout:
21
21
  max_turns: 1
22
- max_tokens: 16384
22
+ max_tokens: 4096
23
23
  termination_policy: last_tool
24
24
  trainer:
25
- num_epochs: 4
25
+ num_epochs: 6
26
26
  learning_rate: 0.0001
27
- save_step: 333
27
+ save_step: 8