RL-MPQ — Llama 2 13B - a AvoCahDoe Collection

AvoCahDoe 's Collections

Invoice Layout Extraction

RL-MPQ VLM — Qwen2-VL-7B

RL-MPQ VLM — LLaVA-Next Mistral-7B

RL-MPQ VLM — LLaVA-1.5-7B

RL-MPQ VLM — LLaVA-1.5-13B

RL-MPQ LLaVA-1.5-13B VLM Evaluation

RL-MPQ — Qwen 2.5 7B

RL-MPQ — Llama 2 7B

RL-MPQ — Llama 3 8B

RL-MPQ — Llama 2 13B

RL-MPQ — Mistral 7B

RL-MPQ — Gemma 2 9B

RL-MPQ — Llama 3.1 8B

RL-MPQ — JetMoE 8B

RL-MPQ — Llama 2 13B

updated Jun 11

5 RL-MPQ fake-quant Llama 2 13B scenarios: High Fidelity, Conservative, Balanced, Aggressive, Extreme Survival.

AvoCahDoe/llama-2-13b-rlmpq-aggressive

Text Generation • 13B • Updated Jun 11 • 5

Note Aggressive: 3.75 bits/weight, 4.2667x vs FP16, WikiText-2 PPL 4.5724. RL-MPQ PPO fake-quant of meta-llama/Llama-2-13b-hf.
AvoCahDoe/llama-2-13b-rlmpq-balanced

Text Generation • 13B • Updated Jun 11 • 3

Note Balanced: 4.4 bits/weight, 3.6364x vs FP16, WikiText-2 PPL 4.4797. RL-MPQ PPO fake-quant of meta-llama/Llama-2-13b-hf.
AvoCahDoe/llama-2-13b-rlmpq-conservative

Text Generation • 13B • Updated Jun 11 • 4

Note Conservative: 5.2 bits/weight, 3.0769x vs FP16, WikiText-2 PPL 4.4663. RL-MPQ PPO fake-quant of meta-llama/Llama-2-13b-hf.
AvoCahDoe/llama-2-13b-rlmpq-extreme-survival

Text Generation • 13B • Updated Jun 11 • 8

Note Extreme Survival: 2.775 bits/weight, 5.7658x vs FP16, WikiText-2 PPL 6.1148. RL-MPQ PPO fake-quant of meta-llama/Llama-2-13b-hf.
AvoCahDoe/llama-2-13b-rlmpq-high-fidelity

Text Generation • 13B • Updated Jun 11 • 3

Note High Fidelity: 6.7 bits/weight, 2.3881x vs FP16, WikiText-2 PPL 4.4313. RL-MPQ PPO fake-quant of meta-llama/Llama-2-13b-hf.