Open-Weight Models

Model · Image and text to text

povl-rl-gsapo-200

Cheney Wong

Reinforcement learning round 2, GSAPO, 200 steps. The delivered model of that round. Equal-weight TEDS 0.8473 / TEDS-Struct 0.9078 over 33,545 samples. Fine-tuned from CheneyWong/povl-rl-dapo-400, which is RL round 1 -- the stages form a single linear chain. Base = the unmodified base model. Public datasets only; the project's self-built benchmarks are excluded. (~0.22 here) measures how well the model reproduces empty cells and should be ignored -- a perturbation test shows a structurally perfect prediction that correctly reads the visible text scores 0.33 TEDS / 1.00 TEDS-S. Only the TEDS-S column is meaningful for that dataset. Rejected at the gate. The one substantive gap: every…

Model · Image and text to text

povl-sft-2d-merge

Cheney Wong

Corrective supervised fine-tuning. The core fix is the two-dimensional merge token (a cell spanning both rows and columns): it accounted for only 0.89% of cell tokens in the previous training mixture, and one source had 0.00%. 22,110 samples with two or more 2D merges were added, lifting the rate to 3.92% (1.06 -> 4.96 per table). This also eliminated the runaway-truncation failure mode entirely. Fine-tuned from CheneyWong/povl-sft-full-3ep, which is its own immediate predecessor in the same run. Base = the unmodified base model. Public datasets only; the project's self-built benchmarks are excluded. (~0.22 here) measures how well the model reproduces empty cells and should be ignored -- a…

Model · Image and text to text

povl-sft-full-3ep

Cheney Wong

First-stage supervised fine-tuning: full-parameter, 100,122 samples, 3 epochs. Its job was to stop the model from generating unbounded output on unseen layouts -- average completion length on real photographed pages had grown to 6,610 characters with a 37.3% truncation rate. Fine-tuned from CheneyWong/tablerecognitionpaddleocrvl16 (PaddleOCR-VL-1.6), which is the released base model, unmodified except for a cleaning pass. Base = the unmodified base model. Public datasets only; the project's self-built benchmarks are excluded. (not evaluated) (~0.22 here) measures how well the model reproduces empty cells and should be ignored -- a perturbation test shows a structurally perfect prediction…

Model · Text classification

pplx-decider-v1-27b-EXL3

ramGPT

EXL3 4.00 bpw conversion of perplexity-ai/pplx-decider-v1-27b. pplx-decider-v1-27b does not use the normal LM-head token-generation path for its final answer. It produces a hidden state and applies the model's custom readout.safetensors decision head to obtain option probabilities. The repository therefore includes pplxdeciderexl3.py, which registers the model architecture and performs the custom decision readout. This model is not recommended for standard TabbyAPI usage. A local smoke test with TabbyAPI and ExLlamaV3 1.5.2 failed during startup with: More importantly, simply registering the architecture is not sufficient for normal /v1/chat/completions behavior: TabbyAPI's standard…

Model · Feature extraction

pplx-embed-v2-late-0.6b-ONNX

Nicolas Embleton

Exported to ONNX from perplexity-ai/pplx-embed-v2-late-0.6b (by Perplexity, MIT license). This is an unofficial conversion; all credit for the model goes to the original authors. The model is a late-interaction (multi-vector) retriever: it outputs one L2-normalized 128-d vector per token or merged image patch, and documents are scored against queries with MaxSim. - embedtokens: inputids -> inputsembeds - visionencoder: pixelvalues, gridthw (one image per call) -> image embeddings that replace the image tokens - textmodel: inputsembeds, attentionmask, positionids [3,B,T] -> tokenembeddings [B,T,128] Browser runtime: the fp32, fp16, q4 files loaded and ran with onnxruntime-web 1.30.0 (wasm…

This is a trained model of a PPO agent playing LunarLander-v3 using the stable-baselines3 library.

This is a trained model of a PPO agent playing LunarLander-v2 using the stable-baselines3 library. Try this policy in the Lunar Lander Sandbox. Choose Human Pilot for direct keyboard control, AI Pilot to let the policy choose every action, or Autopilot with Manual Override to let the policy fly until you intervene. 4 AI Simulations runs four independent policy-controlled flights together, with separate rewards and outcomes. The browser runs an ONNX export of this checkpoint with ONNX Runtime Web. Its Sensor Fault setting (OFF by default) can substitute the previous physics-step observation for an AI decision's current input. Thruster Fault (also OFF by default) can drop a requested thruster…

This is a from-scratch PyTorch implementation of PPO (HF Deep RL Course Unit8 Part1), trained on LunarLander-v3 (Kaggle gymnasium removed v2; checker tags kept as LunarLander-v2 to match the exact env-id string the course checker expects). Architecture: CleanRL-style PPO, 2-layer 64-unit actor/critic MLPs, GAE, clipped surrogate objective, orthogonal init. Training: 16 parallel envs, 3,000,000 total steps, ~639s on a Kaggle T4 GPU. Evaluation: 20 deterministic (argmax) episodes, meanreward=219.32, stdreward=23.46. Load pposcratchmodel.pt into the Agent class defined in pposcratch.py (also included in this repo) and run deterministic argmax actions on the actor logits.

video demo: https://huggingface.co/KevinLXu/ppo-LunarLander-v3/resolve/main/trained-lunarlander-step-0-to-step-1000.mp4 This is a trained model of a PPO agent playing LunarLander-v3 using the stable-baselines3 library. https://huggingface.co/learn/deep-rl-course/unit1/hands-on

This is a trained model of a PPO agent playing LunarLander-v3 using the stable-baselines3 library.

This is a trained model of a PPO agent playing LunarLander-v3 using the stable-baselines3 library.

This is a trained model of a PPO agent playing LunarLander-v3 using the stable-baselines3 library.

This is a trained model of a ppo agent playing Pyramids using the Unity ML-Agents Library (HF Deep RL Course Unit5). Final training (run-id Pyramids1): Step 990000, Mean Reward 1.662, Std of Reward 0.400 (checkpoint exported at Step 1000160). Download the.onnx file and use with the ML-Agents Pyramids demo, or watch it play at https://huggingface.co/spaces/unity/ML-Agents-Pyramids

This is a trained model of a ppo agent playing SnowballTarget using the Unity ML-Agents Library (HF Deep RL Course Unit5). Download the.onnx file and use with the ML-Agents SnowballTarget demo, or watch it play at https://huggingface.co/spaces/ThomasSimonini/ML-Agents-SnowballTarget

This is a trained model of a PPO agent playing LunarLander-v3 using the stable-baselines3 library.

This model is a fine-tuned version of EleutherAI/pythia-160m on an unknown dataset. The following hyperparameters were used during training: - learningrate: 0.001 - trainbatchsize: 16 - evalbatchsize: 16 - gradientaccumulationsteps: 2 - totaltrainbatchsize: 32 - lrschedulertype: cosinewithminlr - lrschedulerwarmupsteps: 13 - trainingsteps: 250 - Transformers 5.4.0 - Pytorch 2.8.0+cu128 - Datasets 3.2.0 - Tokenizers 0.22.1

This model is a fine-tuned version of EleutherAI/pythia-160m on an unknown dataset. The following hyperparameters were used during training: - learningrate: 0.001 - trainbatchsize: 16 - evalbatchsize: 16 - gradientaccumulationsteps: 2 - totaltrainbatchsize: 32 - lrschedulertype: cosinewithminlr - lrschedulerwarmupsteps: 13 - trainingsteps: 250 - Transformers 5.4.0 - Pytorch 2.8.0+cu128 - Datasets 3.2.0 - Tokenizers 0.22.1

This model is a fine-tuned version of EleutherAI/pythia-160m on an unknown dataset. The following hyperparameters were used during training: - learningrate: 0.001 - trainbatchsize: 16 - evalbatchsize: 16 - gradientaccumulationsteps: 2 - totaltrainbatchsize: 32 - lrschedulertype: cosinewithminlr - lrschedulerwarmupsteps: 13 - trainingsteps: 250 - Transformers 5.4.0 - Pytorch 2.8.0+cu128 - Datasets 3.2.0 - Tokenizers 0.22.1