Initial commit

b84161d1 · jerrrrry · b84161d1 · b84161d1 · b84161d1 · b84161d1
Commit b84161d1 authored May 20, 2025 by jerrrrry
20 changed files
--- a/examples/tuning/14b/qwen2_14b_grpo_4_h800_fsdp_vllm.sh
+++ b/examples/tuning/14b/qwen2_14b_grpo_4_h800_fsdp_vllm.sh
--- a/examples/tuning/70b/qwen2-70b_grpo_32_h20_fsdp_vllm.sh
+++ b/examples/tuning/70b/qwen2-70b_grpo_32_h20_fsdp_vllm.sh
--- a/examples/tuning/70b/qwen2-70b_grpo_32_h800_fsdp_vllm.sh
+++ b/examples/tuning/70b/qwen2-70b_grpo_32_h800_fsdp_vllm.sh
--- a/examples/tuning/7b/qwen2-7b_grpo_2_h800_fsdp_vllm.sh
+++ b/examples/tuning/7b/qwen2-7b_grpo_2_h800_fsdp_vllm.sh
--- a/patches/megatron_v4.patch
+++ b/patches/megatron_v4.patch
--- a/pyproject.toml
+++ b/pyproject.toml
--- a/recipe/dapo/README.md
+++ b/recipe/dapo/README.md
--- a/recipe/dapo/prepare_dapo_data.sh
+++ b/recipe/dapo/prepare_dapo_data.sh
--- a/recipe/dapo/run_dapo_early_qwen2.5_32b.sh
+++ b/recipe/dapo/run_dapo_early_qwen2.5_32b.sh
--- a/recipe/dapo/run_dapo_qwen2.5_32b.sh
+++ b/recipe/dapo/run_dapo_qwen2.5_32b.sh
--- a/recipe/dapo/src/config/dapo_trainer.yaml
+++ b/recipe/dapo/src/config/dapo_trainer.yaml
--- a/recipe/dapo/src/dapo_ray_trainer.py
+++ b/recipe/dapo/src/dapo_ray_trainer.py
--- a/recipe/dapo/src/main_dapo.py
+++ b/recipe/dapo/src/main_dapo.py
--- a/recipe/dapo/test_dapo_7b.sh
+++ b/recipe/dapo/test_dapo_7b.sh
--- a/recipe/prime/__init__.py
+++ b/recipe/prime/__init__.py
--- a/recipe/prime/config/prime_trainer.yaml
+++ b/recipe/prime/config/prime_trainer.yaml
--- a/recipe/prime/main_prime.py
+++ b/recipe/prime/main_prime.py
--- a/recipe/prime/prime_core_algos.py
+++ b/recipe/prime/prime_core_algos.py
--- a/recipe/prime/prime_dp_rm.py
+++ b/recipe/prime/prime_dp_rm.py
--- a/recipe/prime/prime_fsdp_workers.py
+++ b/recipe/prime/prime_fsdp_workers.py