Text Generation
Transformers
Safetensors
sdar
feature-extraction
diffusion-language-model
reinforcement-learning
mathematical-reasoning
remasking
drpo
conversational
custom_code
Eval Results (legacy)
Instructions to use Leotsia/DRPO with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Leotsia/DRPO with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="Leotsia/DRPO", trust_remote_code=True) messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("Leotsia/DRPO", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use Leotsia/DRPO with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "Leotsia/DRPO" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Leotsia/DRPO", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/Leotsia/DRPO
- SGLang
How to use Leotsia/DRPO with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "Leotsia/DRPO" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Leotsia/DRPO", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "Leotsia/DRPO" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Leotsia/DRPO", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use Leotsia/DRPO with Docker Model Runner:
docker model run hf.co/Leotsia/DRPO
File size: 3,418 Bytes
f836ab9 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 | {
"architectures": [
"SDARForCausalLM"
],
"attention_bias": false,
"attention_dropout": 0.0,
"auto_map": {
"AutoConfig": "configuration_sdar.SDARConfig",
"AutoModel": "modeling_sdar.SDARForCausalLM",
"AutoModelForCausalLM": "modeling_sdar.SDARForCausalLM"
},
"block_size": 4,
"bos_token_id": 151643,
"debug": false,
"eos_token_id": 151643,
"ep_size": 1,
"fuse_cross_entropy": true,
"gap_curriculum_enable": true,
"gap_curriculum_min_coupling": 0.25,
"gap_curriculum_warmup_ratio": 0.15,
"gap_diffusion_loss_weight": 0.0,
"gap_enable": true,
"gap_global_loss_weight": 0.0,
"gap_grpo_action_granularity": "threshold",
"gap_grpo_advantage_eps": 0.0001,
"gap_grpo_anchor_bias_power": 2.0,
"gap_grpo_branch_debug_interval": 5,
"gap_grpo_branch_debug_max_branches": 0,
"gap_grpo_branch_debug_max_chars": 600,
"gap_grpo_branch_debug_max_examples": 1,
"gap_grpo_branch_debug_once": false,
"gap_grpo_candidate_window_blocks": 4,
"gap_grpo_clip_eps": 0.15,
"gap_grpo_correct_threshold": 0.5,
"gap_grpo_dense_reward_weight": 0.2,
"gap_grpo_diverse_rollbacks": false,
"gap_grpo_entropy_coef": 0.05,
"gap_grpo_fixed_visible_blocks": true,
"gap_grpo_format_reward_weight": 0.0,
"gap_grpo_initial_rollout_temperature": 0.9,
"gap_grpo_kl_coef": 0.02,
"gap_grpo_loss_weight": 0.35,
"gap_grpo_max_correct_count": -1,
"gap_grpo_min_correct_count": 1,
"gap_grpo_min_visible_blocks": 3,
"gap_grpo_mixed_terminal_filter": false,
"gap_grpo_num_parents": 1,
"gap_grpo_num_samples": 8,
"gap_grpo_ref_model_path": "/work/leotsia0416/projects/SDAR/training/model/SDAR-1.7B-Chat",
"gap_grpo_remask_penalty": 0.0,
"gap_grpo_rollout_temperature": 0.9,
"gap_grpo_rollout_top_k": 128,
"gap_grpo_rollout_top_p": 0.98,
"gap_grpo_sample_prob_eps": 0.0001,
"gap_grpo_sft_ce_weight": 0.0,
"gap_grpo_terminal_reward_weight": 2.0,
"gap_grpo_terminal_rollout_scope": "all",
"gap_grpo_terminal_rollout_strategy": "low_confidence_dynamic",
"gap_grpo_use_baseline_branch": false,
"gap_grpo_value_baseline_weight": 0.0,
"gap_grpo_value_loss_weight": 0.0,
"gap_loss_scope": "all",
"gap_min_reveal_tokens": 1,
"gap_noisy_context_loss_weight": 0.05,
"gap_projected_aux_loss_weight": 0.0,
"gap_remask_adv_max_candidates": 1,
"gap_remask_adv_warmup_ratio": 0.1,
"gap_remask_default_p_mask": 1.0,
"gap_remask_loss_weight": 0.0,
"gap_remask_stage_mode": "two_stage",
"gap_remask_supervision": "adv_bce",
"gap_remask_threshold": 0.45,
"gap_reveal_ratio": 0.25,
"gap_rollout_confidence_threshold": 0.92,
"gap_rollout_scope": "frontier_block",
"gap_rollout_steps": 3,
"gap_rollout_strategy": "sequential",
"gap_training_mode": "remask",
"head_dim": 128,
"hidden_act": "silu",
"hidden_size": 2048,
"initializer_range": 0.02,
"intermediate_size": 6144,
"mask_token_id": 151669,
"max_position_embeddings": 32768,
"max_window_layers": 28,
"micro_forward": false,
"model_type": "sdar",
"num_attention_heads": 16,
"num_hidden_layers": 28,
"num_key_value_heads": 8,
"rms_norm_eps": 1e-06,
"rope_scaling": null,
"rope_theta": 1000000,
"skip_checkpoint": false,
"sliding_window": null,
"tie_word_embeddings": false,
"torch_dtype": "float32",
"transformers_version": "4.52.4",
"use_cache": false,
"use_deepep": false,
"use_sliding_window": false,
"vocab_size": 151936
}
|