| { | |
| "model_type": "shivik_m2", | |
| "vocab_size": 49152, | |
| "d_model": 2048, | |
| "n_layers": 24, | |
| "num_heads": 16, | |
| "kv_heads": 4, | |
| "rotary_dim": 128, | |
| "context_length": 4096, | |
| "use_cache": true, | |
| "architectures": [ | |
| "ShivikM2ForCausalLM" | |
| ] | |
| } |
| { | |
| "model_type": "shivik_m2", | |
| "vocab_size": 49152, | |
| "d_model": 2048, | |
| "n_layers": 24, | |
| "num_heads": 16, | |
| "kv_heads": 4, | |
| "rotary_dim": 128, | |
| "context_length": 4096, | |
| "use_cache": true, | |
| "architectures": [ | |
| "ShivikM2ForCausalLM" | |
| ] | |
| } |