| { |
| "model_type": "multiscreen", |
| "architectures": [ |
| "ObsidianMultiscreenForCausalLM" |
| ], |
|
|
| "model_name": "ObsidianSmall-Base", |
| "model_family": "ObsidianSmall", |
| "library_name": "litgpt", |
|
|
| "vocab_size": 4096, |
| "hidden_size": 256, |
| "num_hidden_layers": 9, |
|
|
| "num_attention_heads": 8, |
| "head_dim": 32, |
|
|
| "max_position_embeddings": 1024, |
| "context_length": 1024, |
|
|
| "multiscreen": { |
| "enabled": true, |
| "type": "pure", |
|
|
| "num_heads": 8, |
| "query_chunk_size": 128, |
|
|
| "key_dim": 16, |
| "value_dim": 128, |
|
|
| "window_threshold": 256, |
| "triton_kernel": true, |
|
|
| "kv_cache": false, |
| "full_sequence_decoding": true |
| }, |
|
|
| "mlp": { |
| "enabled": false |
| }, |
|
|
| "attention": { |
| "enabled": false |
| }, |
|
|
| "normalization": "RMSNorm", |
|
|
| "parameters": 8716506, |
|
|
| "torch_dtype": "bfloat16", |
|
|
| "tokenizer_class": "GPT2Tokenizer", |
|
|
| "auto_map": { |
| "AutoConfig": "configuration_multiscreen.ObsidianMultiscreenConfig", |
| "AutoModelForCausalLM": "modeling_multiscreen.ObsidianMultiscreenForCausalLM" |
| }, |
|
|
| "transformers_compatible": false, |
|
|
| "license": "apache-2.0" |
| } |
