Merge branch 'agent-traces' of github.com:huggingface/open-r1 into agent-traces

Author: aymeric-roucher

recipes/Qwen2.5-VL-3B-Instruct/sft/config_agent.yaml

@@ -16,7 +16,7 @@ weight_decay: 0.0001
 optim: adamw_torch
 lr_scheduler_type: linear
 warmup_ratio: 0.1
-learning_rate: 2.0e-05
+learning_rate: 3.0e-05
 gradient_accumulation_steps: 8
 per_device_eval_batch_size: 4
 per_device_train_batch_size: 4 # Change this depending on the context length of the model to keep a 500M GBS. 
@@ -53,18 +53,18 @@ dataset_mixture:
       config: mind2web     # Name of the dataset config
       split: train            # Split to use from the dataset
       columns:                  # Columns to keep
-        - images              
-        - texts    
-      weight: 1.             # Fraction of dataset to use
+        - images
+        - texts
+      weight: 1.
     - id: smolagents/aguvis-stage-2
-      config: guiact-web-multi
+      config: guiact-web-single
       split: train
       columns:
         - images
         - texts
       weight: 1.
     - id: smolagents/aguvis-stage-2
-      config: guiact-web-single
+      config: guiact-web-multi
       split: train
       columns:
         - images