config.json

{
    "num_layers": 80,
    "hidden_size": 8192,
    "ffn_hidden_size": 28672,
    "layernorm_epsilon": 1e-05,
    "max_sequence_length": 2048,
    "num_attention_heads": 64,
    "num_key_value_heads": 8,
    "max_position_embeddings": 4096,
    "vocab_size": 32016,
    "rope_scaling": null,
    "rope_theta": 10000,
    "transpose_mlp_dense": true,
    "transpose_query_key_value": true,
    "load_platform": "megatron",
    "save_platform": "huggingface",
    "load_ckpt_path": "mumath_code_stage_2_pot_new_config_0310/iter_0025128",
    "save_ckpt_path": "mumath_code_stage_2_pot_new_config_0310/iter_0025128_hf",
    "common_config_path": "config/llama-2-70b_codellama.json",
    "tensor_model_parallel_size": 1,
    "pipeline_model_parallel_size": 1,
    "data_parallel_size": 1,
    "no_load_optim": true,
    "no_save_optim": false,
    "use_distributed_optimizer": true,
    "megatron_path": "/workspace/Megatron-LM/",
    "_name_or_path": "codellama/CodeLlama-70b-Python-hf",
    "architectures": [
        "LlamaForCausalLM"
    ],
    "bos_token_id": 1,
    "eos_token_id": 2,
    "hidden_act": "silu",
    "intermediate_size": 28672,
    "initializer_range": 0.02,
    "model_type": "llama",
    "num_hidden_layers": 80,
    "pad_token_id": 0,
    "rms_norm_eps": 1e-05,
    "tie_word_embeddings": false,
    "pretraining_tp": 1,
    "use_cache": true
}