Fix PPO/RLOO examples (#2100)

huggingface · Sep 23, 2024 · 6859e04 · 6859e04
1 parent 92eea1f
commit 6859e04
Show file tree

Hide file tree

Showing 2 changed files with 2 additions and 4 deletions.
diff --git a/examples/scripts/ppo/ppo.py b/examples/scripts/ppo/ppo.py
@@ -50,7 +50,6 @@
     --sft_model_path EleutherAI/pythia-1b-deduped \
     --reward_model_path EleutherAI/pythia-1b-deduped \
     --local_rollout_forward_batch_size 1 \
-    --deepspeed3 \
     --missing_eos_penalty 1.0
 """
 
@@ -88,7 +87,7 @@
     # Dataset
     ################
     dataset = load_dataset("trl-internal-testing/descriptiveness-sentiment-trl-style", split="descriptiveness")
-    eval_samples = 20
+    eval_samples = 100
     train_dataset = dataset.select(range(len(dataset) - eval_samples))
     eval_dataset = dataset.select(range(len(dataset) - eval_samples, len(dataset)))
     dataset_text_field = "prompt"

diff --git a/examples/scripts/rloo/rloo.py b/examples/scripts/rloo/rloo.py
@@ -54,7 +54,6 @@
     --sft_model_path EleutherAI/pythia-1b-deduped \
     --reward_model_path EleutherAI/pythia-1b-deduped \
     --local_rollout_forward_batch_size 1 \
-    --deepspeed3 \
     --missing_eos_penalty 1.0
 """
 
@@ -89,7 +88,7 @@
     # Dataset
     ################
     dataset = load_dataset("trl-internal-testing/descriptiveness-sentiment-trl-style", split="descriptiveness")
-    eval_samples = 20
+    eval_samples = 100
     train_dataset = dataset.select(range(len(dataset) - eval_samples))
     eval_dataset = dataset.select(range(len(dataset) - eval_samples, len(dataset)))
     dataset_text_field = "prompt"