{ "baseline": "modal-projects/spindle@e58ebef130b4cd02de6880e62d27ec358ff7132c:src/spindle/backends/megatron_runtime/common/forward_backward.py::_loss(loss_name='ppo', clip 0.8/1.2) \u2014 Spindle's stock Tinker 'ppo' loss (tinker-cookbook 0.5.9 has no client-side ppo; loss_fn='ppo' is server stock)", "reference": "sail-sg/Stable-RL@c6af1d119c0fe30d16dd9a837d35fdf695acd226:verl/trainer/ppo/core_algos.py::compute_policy_loss_with_mask(loss_mode='dppo_binary_tv', clip_ratio=0.15, clip_ratio_c=10)", "implementation": "code/losses.py sha256 36e744853b410170a874b80e523914ec5b0f04fa7781d96b9d04abd0e10271d3", "sdk": { "tinker": "0.37.0", "tinker-cookbook": "0.5.9", "datasets": "5.1.0", "torch": "2.15.0", "sympy": "1.14.0", "transformers": "5.22.1" }, "no_fork": true }