commit

huggingface · Feb 18, 2025 · 9054c99 · 9054c99
1 parent 6b3976a
commit 9054c99
Show file tree

Hide file tree

Showing 2 changed files with 2 additions and 11 deletions.
diff --git a/examples/model_configs/vllm_model_config.yaml b/examples/model_configs/vllm_model_config.yaml
@@ -1,14 +1,6 @@
 model:
   base_params:
-    model_args: "pretrained=HuggingFaceTB/SmolLM-1.7B,revision=main,dtype=bfloat16" # pretrained=model_name,trust_remote_code=boolean,revision=revision_to_use,model_parallel=True ...
+    model_args: "pretrained=NousResearch/DeepHermes-3-Llama-3-8B-Preview,revision=main,dtype=bfloat16,max_model_length=32000" # pretrained=model_name,trust_remote_code=boolean,revision=revision_to_use,model_parallel=True ...
   generation:
     temperature: 0.3
-    repetition_penalty: 1.0
-    frequency_penalty: 0.0
-    presence_penalty: 0.0
-    seed: 42
-    top_k: -1
-    min_p: 0.0
-    top_p: 0.9
-    max_new_tokens: 256
-    stop_tokens: ["<EOS>", "<PAD>"]
+    max_new_tokens: 12000
diff --git a/src/lighteval/models/vllm/vllm_model.py b/src/lighteval/models/vllm/vllm_model.py
@@ -361,7 +361,6 @@ def run_inference_one_model(model_args: dict, sampling_params: SamplingParams, r
                 if x is not None
             ]
         else:
-            breakpoint()
             outputs = self.model.generate(
                 prompt_token_ids=inputs,
                 sampling_params=sampling_params,