From bcd31cf2456a0584765158106292e9b777e84570 Mon Sep 17 00:00:00 2001
From: Yuchen Han <42163912+hannlp@users.noreply.github.com>
Date: Fri, 17 Nov 2023 00:15:51 -0800
Subject: [PATCH 1/4] Update finetuning_args.py

Former-commit-id: 30e3430553f1f7e09cd57ef2c9843b549746c618
---
 src/llmtuner/hparams/finetuning_args.py | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/src/llmtuner/hparams/finetuning_args.py b/src/llmtuner/hparams/finetuning_args.py
index cfdc8b24..d39812c7 100644
--- a/src/llmtuner/hparams/finetuning_args.py
+++ b/src/llmtuner/hparams/finetuning_args.py
@@ -74,6 +74,10 @@ class RLHFArguments:
         default=None,
         metadata={"help": "Log with either 'wandb' or 'tensorboard' in PPO training."}
     )
+    ppo_epochs: Optional[int] = field(
+        default=4,
+        metadata={"help": "Number of optimisation epochs per batch of samples"},
+    )
     ppo_score_norm: Optional[bool] = field(
         default=False,
         metadata={"help": "Use score normalization in PPO training."}

From 6af7107938cce678ae04b534e9d1132d322fd10a Mon Sep 17 00:00:00 2001
From: Yuchen Han <42163912+hannlp@users.noreply.github.com>
Date: Fri, 17 Nov 2023 00:16:27 -0800
Subject: [PATCH 2/4] Update workflow.py

Former-commit-id: f70b7ffe6442217a222e0ef797c407f259a13886
---
 src/llmtuner/train/ppo/workflow.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/llmtuner/train/ppo/workflow.py b/src/llmtuner/train/ppo/workflow.py
index 41a99e2c..4fb9d593 100644
--- a/src/llmtuner/train/ppo/workflow.py
+++ b/src/llmtuner/train/ppo/workflow.py
@@ -45,7 +45,7 @@ def run_ppo(
         mini_batch_size=training_args.per_device_train_batch_size,
         batch_size=training_args.per_device_train_batch_size * training_args.gradient_accumulation_steps,
         gradient_accumulation_steps=training_args.gradient_accumulation_steps,
-        ppo_epochs=1,
+        ppo_epochs=finetuning_args.ppo_epochs,
         max_grad_norm=training_args.max_grad_norm,
         seed=training_args.seed,
         optimize_device_cache=True,

From cab80a3c568e136757677bcc3ca5df8e32d8c4b5 Mon Sep 17 00:00:00 2001
From: Yuchen Han <42163912+hannlp@users.noreply.github.com>
Date: Fri, 17 Nov 2023 00:17:36 -0800
Subject: [PATCH 3/4] Update README.md

Former-commit-id: c1532dc6fe5d5b427011bd5509a2bc44ee16d951
---
 README.md | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/README.md b/README.md
index 7ba2b748..88a820da 100644
--- a/README.md
+++ b/README.md
@@ -288,6 +288,8 @@ CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
     --per_device_train_batch_size 2 \
     --gradient_accumulation_steps 4 \
     --lr_scheduler_type cosine \
+    --top_k 0 \
+    --top_p 0.9 \
     --logging_steps 10 \
     --save_steps 1000 \
     --learning_rate 1e-5 \

From 992be39f904ddbd29acc9f217de934fbedb7cbe4 Mon Sep 17 00:00:00 2001
From: Yuchen Han <42163912+hannlp@users.noreply.github.com>
Date: Fri, 17 Nov 2023 00:18:07 -0800
Subject: [PATCH 4/4] Update README_zh.md

Former-commit-id: 3e8a17c92d700bcafbe6559ea689dc4c0ad0481a
---
 README_zh.md | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/README_zh.md b/README_zh.md
index dbcee80a..f03f27f0 100644
--- a/README_zh.md
+++ b/README_zh.md
@@ -288,6 +288,8 @@ CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
     --per_device_train_batch_size 2 \
     --gradient_accumulation_steps 4 \
     --lr_scheduler_type cosine \
+    --top_k 0 \
+    --top_p 0.9 \
     --logging_steps 10 \
     --save_steps 1000 \
     --learning_rate 1e-5 \