From 126663821a704a17af627128340e59459c623118 Mon Sep 17 00:00:00 2001
From: Nathan Lambert <nathanl@allenai.org>
Date: Tue, 15 Oct 2024 16:55:56 -0700
Subject: [PATCH] add configs (#200)

---
 scripts/configs/eval_configs.yaml | 39 ++++++++++++++++++++++++++++++-
 1 file changed, 38 insertions(+), 1 deletion(-)

diff --git a/scripts/configs/eval_configs.yaml b/scripts/configs/eval_configs.yaml
index 14befc6..88f9d83 100644
--- a/scripts/configs/eval_configs.yaml
+++ b/scripts/configs/eval_configs.yaml
@@ -783,4 +783,41 @@ Ray2333/GRM-Llama3-8B-rewardmodel-ft:
   trust_remote_code: False
   dpo: False
   quantized: False
-  
\ No newline at end of file
+# for QRM models, use export ACCELERATE_MIXED_PRECISION=bf16; for best performance
+nicolinho/QRM-Llama3.1-8B:
+  model: nicolinho/QRM-Llama3.1-8B
+  tokenizer: nicolinho/QRM-Llama3.1-8B
+  chat_template: # none for tokenizer
+  batch_size: 8
+  dpo: False
+  trust_remote_code: True
+  quantized: False
+  attention_implementation: flash_attention_2
+nicolinho/QRM-Llama3-8B:
+  model: nicolinho/QRM-Llama3-8B
+  tokenizer: nicolinho/QRM-Llama3-8B
+  chat_template: # none for tokenizer
+  batch_size: 8
+  dpo: False
+  trust_remote_code: True
+  quantized: False
+  attention_implementation: flash_attention_2
+Skywork/Skywork-Reward-Gemma-2-27B-v0.2:
+  model: Skywork/Skywork-Reward-Gemma-2-27B-v0.2
+  tokenizer: Skywork/Skywork-Reward-Gemma-2-27B-v0.2
+  chat_template: # none for tokenizer
+  batch_size: 2
+  dpo: False
+  torch_dtype: bfloat16
+  trust_remote_code: False
+  attention_implementation: flash_attention_2
+  quantized: False
+Skywork/Skywork-Reward-Llama-3.1-8B-v0.2:
+  model: Skywork/Skywork-Reward-Llama-3.1-8B-v0.2
+  tokenizer: Skywork/Skywork-Reward-Llama-3.1-8B-v0.2
+  chat_template: # none for tokenizer
+  batch_size: 8
+  dpo: False
+  torch_dtype: bfloat16
+  trust_remote_code: False
+  quantized: False