Skip to content

Commit 4c66dde

Browse files
john-b-yangjohn-b-yang
committed
Merge branch 'john/ladder-run' of github.com:CodeClash-ai/CodeClash into john/ladder-run
2 parents f9b7eb6 + f43eeb1 commit 4c66dde

18 files changed

Lines changed: 916 additions & 21 deletions

‎configs/ablations/ladder/battlesnake_llama_smoke.yaml‎

Lines changed: 7 additions & 21 deletions
Original file line numberDiff line numberDiff line change
@@ -1,9 +1,7 @@
1-
# Cheap smoke to validate the custom-endpoint wiring before a full ladder run:
2-
# one Claude player (native Anthropic API via AnthropicModel) climbs a 2-rung BattleSnake ladder.
1+
# Cheap smoke to validate the llama-endpoint wiring before a full ladder run:
2+
# Opus 4.8 (native Anthropic API via AnthropicModel) climbs a 2-rung BattleSnake ladder.
33
# 1 round, 10 sims — just enough to confirm the model authenticates, caches, and edits code.
4-
#
5-
# uv pip install -e '.[llama]'
6-
# # set LLAMA_API_KEY, LLAMA_BASE_URL, LLAMA_MODEL in .env
4+
# Requires LLAMA_API_KEY in .env.
75
# uv run codeclash ladder run configs/ablations/ladder/battlesnake_llama_smoke.yaml -c
86
tournament:
97
rounds: 1
@@ -16,27 +14,15 @@ game:
1614
browser: false
1715
player:
1816
agent: mini
19-
name: claude-gateway
17+
name: opus-4-8
2018
branch_init: human/moxuz/pinky-snek
2119
config:
2220
agent: !include mini/default.yaml
23-
model:
24-
model_class: codeclash.agents.mini_anthropic_model.AnthropicModel
25-
model_name: claude-gateway # placeholder; real id comes from LLAMA_MODEL
26-
model_name_env: LLAMA_MODEL
27-
api_key_env: LLAMA_API_KEY
28-
base_url_env: LLAMA_BASE_URL
29-
max_tokens: 4096
30-
model_kwargs:
31-
temperature: 0.2
32-
cost:
33-
input: 5.0e-06
34-
output: 2.5e-05
35-
cache_creation_input: 6.25e-06
36-
cache_read_input: 5.0e-07
21+
model: !include mini/models/llama_opus_4_8.yaml
3722
push: True
3823
prompts:
39-
game_description: "BattleSnake ladder smoke over a custom Anthropic-compatible endpoint."
24+
game_description: "BattleSnake ladder smoke over the llama endpoint."
4025
ladder:
4126
- {agent: dummy, name: pinky, branch_init: human/moxuz/pinky-snek} # simple heuristic (port)
4227
- {agent: dummy, name: tr8r, branch_init: human/noahspriggs/tr-8r} # 2016 winner (port)
28+

‎configs/ablations/ladder/corewar__gemini_3_5_flash.yaml‎

Lines changed: 19 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,19 @@
1+
# CoreWar ladder climbed by gemini-3-5-flash. Requires LLAMA_API_KEY in .env.
2+
# uv run codeclash ladder run configs/ablations/ladder/corewar__gemini_3_5_flash.yaml
3+
tournament:
4+
rounds: 5
5+
game:
6+
name: CoreWar
7+
sims_per_round: 2000
8+
player:
9+
agent: mini
10+
name: gemini-3-5-flash
11+
branch_init: human/pspace
12+
config:
13+
agent: !include mini/default.yaml
14+
model: !include mini/models/llama_gemini_3_5_flash.yaml
15+
push: True
16+
prompts:
17+
game_description: |-
18+
Core War ladder
19+
ladder: !include ablations/ladder/rungs/corewar.yaml

‎configs/ablations/ladder/corewar__gpt_5_5.yaml‎

Lines changed: 19 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,19 @@
1+
# CoreWar ladder climbed by gpt-5-5. Requires LLAMA_API_KEY in .env.
2+
# uv run codeclash ladder run configs/ablations/ladder/corewar__gpt_5_5.yaml
3+
tournament:
4+
rounds: 5
5+
game:
6+
name: CoreWar
7+
sims_per_round: 2000
8+
player:
9+
agent: mini
10+
name: gpt-5-5
11+
branch_init: human/pspace
12+
config:
13+
agent: !include mini/default.yaml
14+
model: !include mini/models/llama_gpt_5_5.yaml
15+
push: True
16+
prompts:
17+
game_description: |-
18+
Core War ladder
19+
ladder: !include ablations/ladder/rungs/corewar.yaml

‎configs/ablations/ladder/corewar__opus_4_7.yaml‎

Lines changed: 19 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,19 @@
1+
# CoreWar ladder climbed by opus-4-7. Requires LLAMA_API_KEY in .env.
2+
# uv run codeclash ladder run configs/ablations/ladder/corewar__opus_4_7.yaml
3+
tournament:
4+
rounds: 5
5+
game:
6+
name: CoreWar
7+
sims_per_round: 2000
8+
player:
9+
agent: mini
10+
name: opus-4-7
11+
branch_init: human/pspace
12+
config:
13+
agent: !include mini/default.yaml
14+
model: !include mini/models/llama_opus_4_7.yaml
15+
push: True
16+
prompts:
17+
game_description: |-
18+
Core War ladder
19+
ladder: !include ablations/ladder/rungs/corewar.yaml

‎configs/ablations/ladder/corewar__opus_4_8.yaml‎

Lines changed: 19 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,19 @@
1+
# CoreWar ladder climbed by opus-4-8. Requires LLAMA_API_KEY in .env.
2+
# uv run codeclash ladder run configs/ablations/ladder/corewar__opus_4_8.yaml
3+
tournament:
4+
rounds: 5
5+
game:
6+
name: CoreWar
7+
sims_per_round: 2000
8+
player:
9+
agent: mini
10+
name: opus-4-8
11+
branch_init: human/pspace
12+
config:
13+
agent: !include mini/default.yaml
14+
model: !include mini/models/llama_opus_4_8.yaml
15+
push: True
16+
prompts:
17+
game_description: |-
18+
Core War ladder
19+
ladder: !include ablations/ladder/rungs/corewar.yaml

‎configs/ablations/ladder/corewar__sonnet_5.yaml‎

Lines changed: 19 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,19 @@
1+
# CoreWar ladder climbed by sonnet-5. Requires LLAMA_API_KEY in .env.
2+
# uv run codeclash ladder run configs/ablations/ladder/corewar__sonnet_5.yaml
3+
tournament:
4+
rounds: 5
5+
game:
6+
name: CoreWar
7+
sims_per_round: 2000
8+
player:
9+
agent: mini
10+
name: sonnet-5
11+
branch_init: human/pspace
12+
config:
13+
agent: !include mini/default.yaml
14+
model: !include mini/models/llama_sonnet_5.yaml
15+
push: True
16+
prompts:
17+
game_description: |-
18+
Core War ladder
19+
ladder: !include ablations/ladder/rungs/corewar.yaml

‎configs/ablations/ladder/robotrumble__gemini_3_5_flash.yaml‎

Lines changed: 21 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,21 @@
1+
# RobotRumble ladder climbed by gemini-3-5-flash. Requires LLAMA_API_KEY in .env.
2+
# uv run codeclash ladder run configs/ablations/ladder/robotrumble__gemini_3_5_flash.yaml
3+
tournament:
4+
rounds: 5
5+
game:
6+
name: RobotRumble
7+
sims_per_round: 250
8+
args:
9+
raw: false
10+
player:
11+
agent: mini
12+
name: gemini-3-5-flash
13+
branch_init: human/anton/anton3000
14+
config:
15+
agent: !include mini/default.yaml
16+
model: !include mini/models/llama_gemini_3_5_flash.yaml
17+
push: True
18+
prompts:
19+
game_description: |-
20+
RobotRumble ladder
21+
ladder: !include ablations/ladder/rungs/robotrumble.yaml

‎configs/ablations/ladder/robotrumble__gpt_5_5.yaml‎

Lines changed: 21 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,21 @@
1+
# RobotRumble ladder climbed by gpt-5-5. Requires LLAMA_API_KEY in .env.
2+
# uv run codeclash ladder run configs/ablations/ladder/robotrumble__gpt_5_5.yaml
3+
tournament:
4+
rounds: 5
5+
game:
6+
name: RobotRumble
7+
sims_per_round: 250
8+
args:
9+
raw: false
10+
player:
11+
agent: mini
12+
name: gpt-5-5
13+
branch_init: human/anton/anton3000
14+
config:
15+
agent: !include mini/default.yaml
16+
model: !include mini/models/llama_gpt_5_5.yaml
17+
push: True
18+
prompts:
19+
game_description: |-
20+
RobotRumble ladder
21+
ladder: !include ablations/ladder/rungs/robotrumble.yaml

‎configs/ablations/ladder/robotrumble__opus_4_7.yaml‎

Lines changed: 21 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,21 @@
1+
# RobotRumble ladder climbed by opus-4-7. Requires LLAMA_API_KEY in .env.
2+
# uv run codeclash ladder run configs/ablations/ladder/robotrumble__opus_4_7.yaml
3+
tournament:
4+
rounds: 5
5+
game:
6+
name: RobotRumble
7+
sims_per_round: 250
8+
args:
9+
raw: false
10+
player:
11+
agent: mini
12+
name: opus-4-7
13+
branch_init: human/anton/anton3000
14+
config:
15+
agent: !include mini/default.yaml
16+
model: !include mini/models/llama_opus_4_7.yaml
17+
push: True
18+
prompts:
19+
game_description: |-
20+
RobotRumble ladder
21+
ladder: !include ablations/ladder/rungs/robotrumble.yaml

‎configs/ablations/ladder/robotrumble__opus_4_8.yaml‎

Lines changed: 21 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,21 @@
1+
# RobotRumble ladder climbed by opus-4-8. Requires LLAMA_API_KEY in .env.
2+
# uv run codeclash ladder run configs/ablations/ladder/robotrumble__opus_4_8.yaml
3+
tournament:
4+
rounds: 5
5+
game:
6+
name: RobotRumble
7+
sims_per_round: 250
8+
args:
9+
raw: false
10+
player:
11+
agent: mini
12+
name: opus-4-8
13+
branch_init: human/anton/anton3000
14+
config:
15+
agent: !include mini/default.yaml
16+
model: !include mini/models/llama_opus_4_8.yaml
17+
push: True
18+
prompts:
19+
game_description: |-
20+
RobotRumble ladder
21+
ladder: !include ablations/ladder/rungs/robotrumble.yaml

0 commit comments

Comments
 (0)