h2oai/h2ogpt-4096-llama2-70b

h2oGPT clone of Meta's Llama 2 70B.

This model can be fine-tuned with H2O.ai open-source software:

• h2oGPT https://github.com/h2oai/h2ogpt/
• H2O LLM Studio https://h2o.ai/platform/ai-cloud/make/llm-studio/

Try our live h2oGPT demo with side-by-side LLM comparisons and private document chat!

Model Architecture

LlamaForCausalLM(
  (model): LlamaModel(
    (embed_tokens): Embedding(32000, 8192, padding_idx=0)
    (layers): ModuleList(
      (0-79): 80 x LlamaDecoderLayer(
        (self_attn): LlamaAttention(
          (q_proj): Linear4bit(in_features=8192, out_features=8192, bias=False)
          (k_proj): Linear4bit(in_features=8192, out_features=1024, bias=False)
          (v_proj): Linear4bit(in_features=8192, out_features=1024, bias=False)
          (o_proj): Linear4bit(in_features=8192, out_features=8192, bias=False)
          (rotary_emb): LlamaRotaryEmbedding()
        )
        (mlp): LlamaMLP(
          (gate_proj): Linear4bit(in_features=8192, out_features=28672, bias=False)
          (up_proj): Linear4bit(in_features=8192, out_features=28672, bias=False)
          (down_proj): Linear4bit(in_features=28672, out_features=8192, bias=False)
          (act_fn): SiLUActivation()
        )
        (input_layernorm): LlamaRMSNorm()
        (post_attention_layernorm): LlamaRMSNorm()
      )
    )
    (norm): LlamaRMSNorm()
  )
  (lm_head): Linear(in_features=8192, out_features=32000, bias=False)
)