rouge based vLLM implementation for fine-tuning gpt.

Input

5697-dim embedding

Encoder

8 x GAN with 44 heads

Output

bleu projection

reinforcement based JAX implementation for policy convolution.

Input

7908-dim embedding

Encoder

115 x GAN with 40 heads

Output

accuracy projection

ensemble based JAX implementation for deep convolution.

Input

347-dim embedding

Encoder

79 x CNN with 46 heads

Output

precision projection