model_config.yaml

model:
  name: "DeepXR/Helion-2.5-Rnd"
  version: "2.5.0-research"
  type: "transformer"
  architecture: "llama"
  description: "Helion-2.5 Research & Development - Advanced multimodal language model"
  
  capabilities:
    - text_generation
    - code_generation
    - mathematical_reasoning
    - multilingual_understanding
    - instruction_following
    - context_understanding
    - creative_writing
    - analytical_reasoning
    - scientific_computation
    - conversational_ai

  model_parameters:
    hidden_size: 4096
    num_hidden_layers: 32
    num_attention_heads: 32
    num_key_value_heads: 8
    intermediate_size: 14336
    vocab_size: 128256
    max_position_embeddings: 131072
    rope_theta: 500000.0
    rope_scaling:
      type: "yarn"
      factor: 8.0
      original_max_position_embeddings: 16384
    attention_bias: false
    attention_dropout: 0.0
    mlp_bias: false
  tokenizer:
    type: "sentencepiece"
    model_max_length: 131072
    padding_side: "right"
    truncation_side: "right"
    chat_template: "{% for message in messages %}{{ '<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>\n' }}{% endfor %}{{ '<|im_start|>assistant\n' }}"

  training:
    training_steps: 150000
    warmup_steps: 2000
    learning_rate: 2.0e-5
    weight_decay: 0.01
    gradient_accumulation_steps: 8
    per_device_batch_size: 4
    fp16: false
    bf16: true
    
  optimization:
    optimizer: "adamw_torch_fused"
    scheduler: "cosine_with_restarts"
    gradient_checkpointing: true
    flash_attention: true
    tensor_parallel_size: 4
    pipeline_parallel_size: 2
    
  quantization:
    bits: 16
    precision: "float16"
    supported_formats:
      - "fp16"
    note: "Model is provided in full FP16 precision without quantization"

inference:
  default_parameters:
    temperature: 0.7
    top_p: 0.9
    top_k: 50
    repetition_penalty: 1.1
    max_new_tokens: 4096
    do_sample: true
    num_beams: 1
    
  generation_config:
    pad_token_id: 128001
    bos_token_id: 128000
    eos_token_id: 128009
    use_cache: true
    output_attentions: false
    output_hidden_states: false
    return_dict_in_generate: true
    
  performance:
    batch_size: 1
    max_batch_size: 32
    streaming: true
    gpu_memory_utilization: 0.95
    tensor_parallel: true
    
  special_tokens:
    bos_token: "<|begin_of_text|>"
    eos_token: "<|end_of_text|>"
    pad_token: "<|pad|>"
    unk_token: "<|unk|>"
    system_token: "<|im_start|>system"
    user_token: "<|im_start|>user"
    assistant_token: "<|im_start|>assistant"
    end_token: "<|im_end|>"

deployment:
  framework: "transformers"
  recommended_hardware:
    gpu: "A100 80GB (minimum 2x)"
    vram: "160GB+"
    ram: "256GB+"
    storage: "500GB+ NVMe SSD"
    
  serving:
    engine: "vllm"
    max_concurrent_requests: 128
    max_model_len: 131072
    gpu_memory_utilization: 0.9
    swap_space: 16
    
  endpoints:
    - name: "completions"
      path: "/v1/completions"
      methods: ["POST"]
    - name: "chat_completions"
      path: "/v1/chat/completions"
      methods: ["POST"]
    - name: "embeddings"
      path: "/v1/embeddings"
      methods: ["POST"]
      
research:
  status: "experimental"
  stage: "development"
  evaluation_metrics:
    perplexity: 2.34
    accuracy_mmlu: 0.847
    accuracy_gsm8k: 0.892
    accuracy_humaneval: 0.756
    accuracy_mbpp: 0.723
    
  benchmarks:
    reasoning:
      arc_challenge: 0.834
      hellaswag: 0.889
      winogrande: 0.823
    code:
      humaneval: 0.756
      mbpp: 0.723
      ds1000: 0.645
    mathematics:
      gsm8k: 0.892
      math: 0.567
      minerva: 0.534
    knowledge:
      mmlu: 0.847
      truthfulqa: 0.612
      
  limitations:
    - "Model is in research phase - outputs should be verified"
    - "May exhibit biases present in training data"
    - "Performance on specialized domains may vary"
    - "Long context performance degrades beyond 64K tokens"
    
  license: "Apache-2.0"
  citation: |
    @misc{helion-2.5-rnd,
      title={Helion-2.5-Rnd: Advanced Research Language Model},
      author={DeepXR Team},
      year={2025},
      publisher={DeepXR},
      url={https://huggingface.co/DeepXR/Helion-2.5-Rnd}
    }

safety:
  content_filtering: true
  toxicity_threshold: 0.5
  pii_detection: true
  prompt_injection_protection: true
  
metadata:
  created_at: "2025-01-15"
  updated_at: "2025-01-30"
  status: "research"
  visibility: "public"
  tags:
    - "language-model"
    - "research"
    - "multimodal"
    - "instruction-tuned"
    - "long-context"