Merge branch 'kilian-tmp-exp'

Author: klieret

codeclash/agents/minisweagent.py

@@ -10,11 +10,13 @@
 from minisweagent.models.test_models import DeterministicModel
 from minisweagent.run.utils.save import save_traj
 
+from codeclash import REPO_DIR
 from codeclash.agents.player import Player
 from codeclash.agents.utils import GameContext
 from codeclash.utils.environment import copy_to_container
 
 os.environ["MSWEA_MODEL_RETRY_STOP_AFTER_ATTEMPT"] = "90"
+os.environ["LITELLM_MODEL_REGISTRY_PATH"] = str((REPO_DIR / "configs" / "litellm_custom_model_config.yaml").resolve())
 
 
 class ClashAgent(DefaultAgent):

configs/litellm_custom_model_config.yaml

@@ -0,0 +1,51 @@
+{
+    "dashscope/qwen3-coder-plus-2025-09-23": {
+        "litellm_provider": "dashscope",
+        "max_input_tokens": 997952,
+        "max_output_tokens": 65536,
+        "max_tokens": 1000000,
+        "mode": "chat",
+        "source": "https://www.alibabacloud.com/help/en/model-studio/models",
+        "supports_function_calling": true,
+        "supports_reasoning": true,
+        "supports_tool_choice": true,
+        "tiered_pricing": [
+            {
+                "cache_read_input_token_cost": 1e-07,
+                "input_cost_per_token": 1e-06,
+                "output_cost_per_token": 5e-06,
+                "range": [
+                    0,
+                    32000.0
+                ]
+            },
+            {
+                "cache_read_input_token_cost": 1.8e-07,
+                "input_cost_per_token": 1.8e-06,
+                "output_cost_per_token": 9e-06,
+                "range": [
+                    32000.0,
+                    128000.0
+                ]
+            },
+            {
+                "cache_read_input_token_cost": 3e-07,
+                "input_cost_per_token": 3e-06,
+                "output_cost_per_token": 1.5e-05,
+                "range": [
+                    128000.0,
+                    256000.0
+                ]
+            },
+            {
+                "cache_read_input_token_cost": 6e-07,
+                "input_cost_per_token": 6e-06,
+                "output_cost_per_token": 6e-05,
+                "range": [
+                    256000.0,
+                    1000000.0
+                ]
+            }
+        ]
+    }
+}

configs/main/HuskyBench__claude-sonnet-4-20250514__qwen3-coder-plus-2025-09-23__r15__s100.yaml

@@ -20,9 +20,7 @@ players:
   config:
     agent: !include mini/default.yaml
     model:
-      model_name: qwen3-coder-plus-2025-09-23
-      model_class: dashscope
-      litellm_model_name_override: deepinfra/Qwen/Qwen3-Coder-480B-A35B-Instruct
+      model_name: dashscope/qwen3-coder-plus-2025-09-23
       model_kwargs:
         temperature: 0.2
 prompts:

configs/main/HuskyBench__claude-sonnet-4-5-20250929__qwen3-coder-plus-2025-09-23__r15__s100.yaml

@@ -20,9 +20,7 @@ players:
   config:
     agent: !include mini/default.yaml
     model:
-      model_name: qwen3-coder-plus-2025-09-23
-      model_class: dashscope
-      litellm_model_name_override: deepinfra/Qwen/Qwen3-Coder-480B-A35B-Instruct
+      model_name: dashscope/qwen3-coder-plus-2025-09-23
       model_kwargs:
         temperature: 0.2
 prompts:

configs/main/HuskyBench__gemini-2.5-pro__qwen3-coder-plus-2025-09-23__r15__s100.yaml

@@ -19,9 +19,7 @@ players:
   config:
     agent: !include mini/default.yaml
     model:
-      model_name: qwen3-coder-plus-2025-09-23
-      model_class: dashscope
-      litellm_model_name_override: deepinfra/Qwen/Qwen3-Coder-480B-A35B-Instruct
+      model_name: dashscope/qwen3-coder-plus-2025-09-23
       model_kwargs:
         temperature: 0.2
 prompts:

configs/main/HuskyBench__gpt-5-mini__qwen3-coder-plus-2025-09-23__r15__s100.yaml

@@ -17,9 +17,7 @@ players:
   config:
     agent: !include mini/default.yaml
     model:
-      model_name: qwen3-coder-plus-2025-09-23
-      model_class: dashscope
-      litellm_model_name_override: deepinfra/Qwen/Qwen3-Coder-480B-A35B-Instruct
+      model_name: dashscope/qwen3-coder-plus-2025-09-23
       model_kwargs:
         temperature: 0.2
 prompts:

configs/main/HuskyBench__gpt-5__qwen3-coder-plus-2025-09-23__r15__s100.yaml

@@ -17,9 +17,7 @@ players:
   config:
     agent: !include mini/default.yaml
     model:
-      model_name: qwen3-coder-plus-2025-09-23
-      model_class: dashscope
-      litellm_model_name_override: deepinfra/Qwen/Qwen3-Coder-480B-A35B-Instruct
+      model_name: dashscope/qwen3-coder-plus-2025-09-23
       model_kwargs:
         temperature: 0.2
 prompts:

configs/main/HuskyBench__grok-code-fast-1__qwen3-coder-plus-2025-09-23__r15__s100.yaml

@@ -20,9 +20,7 @@ players:
   config:
     agent: !include mini/default.yaml
     model:
-      model_name: qwen3-coder-plus-2025-09-23
-      model_class: dashscope
-      litellm_model_name_override: deepinfra/Qwen/Qwen3-Coder-480B-A35B-Instruct
+      model_name: dashscope/qwen3-coder-plus-2025-09-23
       model_kwargs:
         temperature: 0.2
 prompts:

configs/main/HuskyBench__o3__qwen3-coder-plus-2025-09-23__r15__s100.yaml

@@ -10,9 +10,7 @@ players:
   config:
     agent: !include mini/default.yaml
     model:
-      model_name: qwen3-coder-plus-2025-09-23
-      model_class: dashscope
-      litellm_model_name_override: deepinfra/Qwen/Qwen3-Coder-480B-A35B-Instruct
+      model_name: dashscope/qwen3-coder-plus-2025-09-23
       model_kwargs:
         temperature: 0.2
 - agent: mini

configs/models.yaml

@@ -35,9 +35,12 @@
   #   drop_params: true
   #   reasoning_effort: "high"
   #   verbosity: "medium"
-- model_name: "qwen3-coder-plus-2025-09-23"
-  model_class: dashscope
-  litellm_model_name_override: "deepinfra/Qwen/Qwen3-Coder-480B-A35B-Instruct"
+# - model_name: "qwen3-coder-plus-2025-09-23"
+#   model_class: dashscope
+#   litellm_model_name_override: "deepinfra/Qwen/Qwen3-Coder-480B-A35B-Instruct"
+#   model_kwargs:
+#     temperature: 0.2
+- model_name: "dashscope/qwen3-coder-plus-2025-09-23"
   model_kwargs:
     temperature: 0.2
 - model_name: "@openai/o3"