amang1802
/

Llama3.2-1B-summary-length-exp2

Text Generation

text-generation-inference

Model card Files Files and versions

Model Card for Model ID

Summary Length PPO experiment #2
No KL divergence in loss

Model Details

Dataset size: 1024
Epochs: 2
Batch Size: 4 * 8 (using Grad Accu)

Optimizer args: Torch AdamW default, except

LR = 0.0001

Downloads last month: -

Safetensors

Model size

1.24B params

Tensor type

BF16

·

Model tree for amang1802/Llama3.2-1B-summary-length-exp2

Quantizations

1 model

Collection including amang1802/Llama3.2-1B-summary-length-exp2

PPO experiments

Using PPO with simpler reward functions • 8 items • Updated Jan 23