Invoice document layout detection: annotated dataset, fine-tuned YOLOv8 models (nano & medium), and an interactive Gradio demo.
🤝 Open to Collab
El boubkraoui Farid
AvoCahDoe
·
AI & ML interests
None yet
Recent Activity
updated a Space about 5 hours ago
AvoCahDoe/bo-ocr-router updated a Space about 5 hours ago
AvoCahDoe/morocco-bo-latin-ocr-demo published a Space about 5 hours ago
AvoCahDoe/bo-ocr-routerOrganizations
None yet
RL-MPQ fake-quant LLaVA-Next Mistral-7B. Aggressive @ 2.81 bpw. VLMEvalKit MMMU/MMBench/ScienceQA.
RL-MPQ fake-quant LLaVA-1.5-13B (Llama-2-13B backbone). High Fidelity, Balanced, Aggressive. VLMEvalKit MMMU/MMBench/ScienceQA.
-
AvoCahDoe/llava-1.5-13b-rlmpq-high-fidelity
Image-Text-to-Text • 13B • Updated • 4 -
AvoCahDoe/llava-1.5-13b-rlmpq-balanced
Image-Text-to-Text • 13B • Updated • 7 -
AvoCahDoe/llava-1.5-13b-rlmpq-aggressive
Image-Text-to-Text • 13B • Updated • 5 -
AvoCahDoe/llava-15-rlmpq-vlm-eval-results
Viewer • Updated • 14 • 91
5 RL-MPQ fake-quant Qwen 2.5 7B scenarios. Mixed-precision fake-quant checkpoints for thesis.
5 RL-MPQ fake-quant Llama 3 8B scenarios: High Fidelity through Extreme Survival. Thesis checkpoints.
5 RL-MPQ fake-quant Mistral 7B scenarios. Per-layer mixed precision via PPO-trained policies.
5 RL-MPQ fake-quant Llama 3.1 8B scenarios: High Fidelity through Extreme Survival. Extension checkpoints.
-
AvoCahDoe/llama-3-1-8b-rlmpq-aggressive
Text Generation • 8B • Updated • 6 -
AvoCahDoe/llama-3-1-8b-rlmpq-balanced
Text Generation • 8B • Updated • 5 -
AvoCahDoe/llama-3-1-8b-rlmpq-conservative
Text Generation • 8B • Updated • 5 -
AvoCahDoe/llama-3-1-8b-rlmpq-extreme-survival
Text Generation • 8B • Updated • 5
RL-MPQ fake-quant Qwen2-VL-7B (qwen2_5_7b Balanced policy). 3.39 bpw. VLMEvalKit exact-match eval.
RL-MPQ fake-quant LLaVA-1.5-7B (Llama-2-7B backbone). Extreme_Survival @ 2.97 bpw. VLMEvalKit exact-match eval.
RL-MPQ fake-quantized LLaVA-1.5-13B vision-language models and their VLMEvalKit benchmark results (MMMU, MMBench, ScienceQA).
-
AvoCahDoe/llava-1.5-13b-rlmpq-high-fidelity
Image-Text-to-Text • 13B • Updated • 4 -
AvoCahDoe/llava-1.5-13b-rlmpq-balanced
Image-Text-to-Text • 13B • Updated • 7 -
AvoCahDoe/llava-1.5-13b-rlmpq-aggressive
Image-Text-to-Text • 13B • Updated • 5 -
AvoCahDoe/llava-15-rlmpq-vlm-eval-results
Viewer • Updated • 14 • 91
5 RL-MPQ fake-quant Llama 2 7B scenarios: High Fidelity, Conservative, Balanced, Aggressive, Extreme Survival.
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 • 5 -
AvoCahDoe/llama-2-13b-rlmpq-balanced
Text Generation • 13B • Updated • 3 -
AvoCahDoe/llama-2-13b-rlmpq-conservative
Text Generation • 13B • Updated • 4 -
AvoCahDoe/llama-2-13b-rlmpq-extreme-survival
Text Generation • 13B • Updated • 8
5 RL-MPQ fake-quant Gemma 2 9B scenarios. Per-layer bit-width policies from Phase 3 PPO.
5 RL-MPQ fake-quant JetMoE 8B scenarios. Mixed-precision MoE fake-quant extension checkpoints.
Invoice document layout detection: annotated dataset, fine-tuned YOLOv8 models (nano & medium), and an interactive Gradio demo.
RL-MPQ fake-quant Qwen2-VL-7B (qwen2_5_7b Balanced policy). 3.39 bpw. VLMEvalKit exact-match eval.
RL-MPQ fake-quant LLaVA-Next Mistral-7B. Aggressive @ 2.81 bpw. VLMEvalKit MMMU/MMBench/ScienceQA.
RL-MPQ fake-quant LLaVA-1.5-7B (Llama-2-7B backbone). Extreme_Survival @ 2.97 bpw. VLMEvalKit exact-match eval.
RL-MPQ fake-quant LLaVA-1.5-13B (Llama-2-13B backbone). High Fidelity, Balanced, Aggressive. VLMEvalKit MMMU/MMBench/ScienceQA.
-
AvoCahDoe/llava-1.5-13b-rlmpq-high-fidelity
Image-Text-to-Text • 13B • Updated • 4 -
AvoCahDoe/llava-1.5-13b-rlmpq-balanced
Image-Text-to-Text • 13B • Updated • 7 -
AvoCahDoe/llava-1.5-13b-rlmpq-aggressive
Image-Text-to-Text • 13B • Updated • 5 -
AvoCahDoe/llava-15-rlmpq-vlm-eval-results
Viewer • Updated • 14 • 91
RL-MPQ fake-quantized LLaVA-1.5-13B vision-language models and their VLMEvalKit benchmark results (MMMU, MMBench, ScienceQA).
-
AvoCahDoe/llava-1.5-13b-rlmpq-high-fidelity
Image-Text-to-Text • 13B • Updated • 4 -
AvoCahDoe/llava-1.5-13b-rlmpq-balanced
Image-Text-to-Text • 13B • Updated • 7 -
AvoCahDoe/llava-1.5-13b-rlmpq-aggressive
Image-Text-to-Text • 13B • Updated • 5 -
AvoCahDoe/llava-15-rlmpq-vlm-eval-results
Viewer • Updated • 14 • 91
5 RL-MPQ fake-quant Qwen 2.5 7B scenarios. Mixed-precision fake-quant checkpoints for thesis.
5 RL-MPQ fake-quant Llama 2 7B scenarios: High Fidelity, Conservative, Balanced, Aggressive, Extreme Survival.
5 RL-MPQ fake-quant Llama 3 8B scenarios: High Fidelity through Extreme Survival. Thesis checkpoints.
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 • 5 -
AvoCahDoe/llama-2-13b-rlmpq-balanced
Text Generation • 13B • Updated • 3 -
AvoCahDoe/llama-2-13b-rlmpq-conservative
Text Generation • 13B • Updated • 4 -
AvoCahDoe/llama-2-13b-rlmpq-extreme-survival
Text Generation • 13B • Updated • 8
5 RL-MPQ fake-quant Mistral 7B scenarios. Per-layer mixed precision via PPO-trained policies.
5 RL-MPQ fake-quant Gemma 2 9B scenarios. Per-layer bit-width policies from Phase 3 PPO.
5 RL-MPQ fake-quant Llama 3.1 8B scenarios: High Fidelity through Extreme Survival. Extension checkpoints.
-
AvoCahDoe/llama-3-1-8b-rlmpq-aggressive
Text Generation • 8B • Updated • 6 -
AvoCahDoe/llama-3-1-8b-rlmpq-balanced
Text Generation • 8B • Updated • 5 -
AvoCahDoe/llama-3-1-8b-rlmpq-conservative
Text Generation • 8B • Updated • 5 -
AvoCahDoe/llama-3-1-8b-rlmpq-extreme-survival
Text Generation • 8B • Updated • 5
5 RL-MPQ fake-quant JetMoE 8B scenarios. Mixed-precision MoE fake-quant extension checkpoints.