lightrft.models

• lightrft.models.actor_al
  • ActorAL
    • ActorAL.forward()
    • ActorAL.generate()
    • ActorAL.gradient_checkpointing_disable()
    • ActorAL.gradient_checkpointing_enable()
    • ActorAL.modality
    • ActorAL.print_trainable_parameters()
    • ActorAL.process_sequences()
• lightrft.models.actor_language
  • ActorLanguage
    • ActorLanguage.forward()
    • ActorLanguage.generate()
    • ActorLanguage.gradient_checkpointing_disable()
    • ActorLanguage.gradient_checkpointing_enable()
    • ActorLanguage.modality
    • ActorLanguage.print_trainable_parameters()
    • ActorLanguage.process_sequences()
• lightrft.models.actor_vl
  • ActorVL
    • ActorVL.forward()
    • ActorVL.generate()
    • ActorVL.gradient_checkpointing_disable()
    • ActorVL.gradient_checkpointing_enable()
    • ActorVL.modality
    • ActorVL.print_trainable_parameters()
    • ActorVL.process_sequences()
• lightrft.models.grm_vl
  • GenerativeRewardModelVL
    • GenerativeRewardModelVL.forward()
    • GenerativeRewardModelVL.generate()
    • GenerativeRewardModelVL.gradient_checkpointing_disable()
    • GenerativeRewardModelVL.gradient_checkpointing_enable()
    • GenerativeRewardModelVL.print_trainable_parameters()
• lightrft.models.loss
  • DPOLoss
    • DPOLoss.forward()
  • GPTLMLoss
    • GPTLMLoss.forward()
  • HPSLoss
    • HPSLoss.forward()
  • KDLoss
    • KDLoss.forward()
  • KTOLoss
    • KTOLoss.forward()
  • LogExpLoss
    • LogExpLoss.forward()
  • LogSigmoidLoss
    • LogSigmoidLoss.forward()
  • PRMLoss
    • PRMLoss.forward()
  • PairWiseLoss
    • PairWiseLoss.forward()
  • PolicyLoss
    • PolicyLoss.forward()
  • ValueLoss
    • ValueLoss.forward()
  • VanillaKTOLoss
    • VanillaKTOLoss.forward()
• lightrft.models.monkey_patch
  • lightrft.models.monkey_patch.apply
    • apply_monkey_patch_to_llama()
    • apply_monkey_patch_to_qwen2()
  • Hugging Face generation patch
  • lightrft.models.monkey_patch.llama
    • llama_attn_forward()
  • lightrft.models.monkey_patch.qwen
    • qwen2_attn_forward()
• lightrft.models.srm_al
  • ScalarRewardModelAL
    • ScalarRewardModelAL.forward()
    • ScalarRewardModelAL.gradient_checkpointing_disable()
    • ScalarRewardModelAL.gradient_checkpointing_enable()
    • ScalarRewardModelAL.print_trainable_parameters()
• lightrft.models.srm_vl
  • ScalarRewardModelVL
    • ScalarRewardModelVL.forward()
    • ScalarRewardModelVL.gradient_checkpointing_disable()
    • ScalarRewardModelVL.gradient_checkpointing_enable()
    • ScalarRewardModelVL.print_trainable_parameters()
• lightrft.models.utils
  • log_probs_from_logits
    • log_probs_from_logits()