config.json

{
  "architectures": [
    "CohereForCausalLM"
  ],
  "attention_axis_name": "sp",
  "attention_bias": false,
  "attention_dropout": 0.0,
  "attn_mechanism": "sharded_vanilla",
  "axis_dims": [
    1,
    -1,
    1,
    1
  ],
  "axis_names": [
    "dp",
    "fsdp",
    "tp",
    "sp"
  ],
  "backend": null,
  "bits": null,
  "block_b": 1,
  "block_k": 128,
  "block_k_dkv": 128,
  "block_k_dq": 128,
  "block_k_major": 128,
  "block_k_major_dkv": 128,
  "block_k_major_dq": 128,
  "block_q": 128,
  "block_q_dkv": 128,
  "block_q_dq": 128,
  "block_q_major_dkv": 128,
  "bos_token_id": 5,
  "easy_method": "train",
  "eos_token_id": 255001,
  "flash_attention_backward_pass_impl": "triton",
  "gradient_checkpointing": "nothing_saveable",
  "hidden_act": "silu",
  "hidden_size": 4096,
  "initializer_range": 0.02,
  "intermediate_size": 14336,
  "layer_norm_eps": 1e-05,
  "logit_scale": 0.0625,
  "max_position_embeddings": 8192,
  "model_type": "cohere",
  "num_attention_heads": 32,
  "num_hidden_layers": 32,
  "num_key_value_heads": 8,
  "pad_token_id": 0,
  "partition_axis": [
    [
      "fsdp",
      "dp"
    ],
    "sp",
    "sp",
    "tp",
    "sp",
    "tp",
    null,
    null,
    null,
    null,
    "tp",
    "sp",
    null
  ],
  "quantize_kv_cache": false,
  "rope_theta": 10000,
  "scan_attention_layers": false,
  "scan_mlp_chunk_size": 1024,
  "scan_ring_attention": true,
  "shard_attention_computation": true,
  "torch_dtype": "float16",
  "transformers_version": "4.41.2",
  "use_cache": true,
  "use_qk_norm": false,
  "use_scan_mlp": false,
  "use_sharded_kv_caching": true,
  "use_sharding_constraint": false,
  "vocab_size": 256000
}