CUDA_VISIBLE_DEVICES=0 python test_benchmark_inference.py \
         -d /workspace/models/superhot-13b-8k-no-rlhf-test-32g-GPTQ \
         -ppl \
         -ppl_ds datasets/wikitext2.txt \
         -l 8192 \
         -cpe 4 \
         -ppl_cn 40 \
         -ppl_cs 8192 \
         -ppl_ct 8192
 -- Perplexity:
 -- - Dataset: datasets/wikitext2.txt
 -- - Chunks: 40
 -- - Chunk size: 8192 -> 8192
 -- - Chunk overlap: 0
 -- - Min. chunk size: 50
 -- - Key: text
 -- Tokenizer: /workspace/models/superhot-13b-8k-no-rlhf-test-32g-GPTQ/tokenizer.model
 -- Model config: /workspace/models/superhot-13b-8k-no-rlhf-test-32g-GPTQ/config.json
 -- Model: /workspace/models/superhot-13b-8k-no-rlhf-test-32g-GPTQ/4bit-32g.safetensors
 -- Sequence length: 8192
 -- RoPE compression factor: 4.0
 -- Tuning:
 -- --matmul_recons_thd: 8
 -- --fused_mlp_thd: 2
 -- --sdp_thd: 8
 -- Options: ['perplexity']
 ** Time, Load model: 4.23 seconds
 ** Time, Load tokenizer: 0.01 seconds
 -- Groupsize (inferred): 32
 -- Act-order (inferred): yes
 ** VRAM, Model: [cuda:0] 7,732.62 MB
 -- Loading dataset...
 -- Testing 40 chunks....
 ** Perplexity: 5.4066