Update config.json

config.json

@@ -0,0 +1,61 @@
+{
+  "model_type": "steerling",
+  "auto_map": {
+    "AutoConfig": "configuration_steerling.SteerlingConfig",
+    "AutoModel": "modeling_steerling.SteerlingForCausalLM",
+    "AutoModelForCausalLM": "modeling_steerling.SteerlingForCausalLM",
+    "AutoTokenizer": ["tokenization_steerling.SteerlingTokenizer", null]
+  },
+  "architectures": ["SteerlingForCausalLM"],
+  "interpretable": true,
+  "vocab_size": 100284,
+  "n_layers": 32,
+  "n_head": 32,
+  "n_embd": 4096,
+  "n_kv_heads": 4,
+  "block_size": 4096,
+  "diff_block_size": 64,
+  "use_rms_norm": true,
+  "norm_eps": 1e-05,
+  "norm_order": "post",
+  "use_qk_norm": true,
+  "use_rope": true,
+  "rope_base": 500000.0,
+  "rope_full_precision": true,
+  "clip_qkv": 10.0,
+  "mlp_type": "swiglu",
+  "activation": "gelu",
+  "mlp_ratio": 4,
+  "intermediate_size": null,
+  "use_bias": false,
+  "weight_sharing": true,
+  "pad_token_id": 100277,
+  "bos_token_id": 100278,
+  "eos_token_id": 100257,
+  "mask_token_id": 100280,
+  "endofchunk_token_id": 100279,
+  "n_concepts": 33732,
+  "n_unknown_concepts": 101196,
+  "concept_dim": 4096,
+  "use_attention_known": false,
+  "use_attention_unknown": false,
+  "topk_known": 32,
+  "topk_known_features": 32,
+  "unknown_topk": 128,
+  "use_unknown": true,
+  "apply_topk_to_unknown": true,
+  "topk_on_logits": false,
+  "factorize_unknown": true,
+  "factorize_rank": 256,
+  "use_epsilon_correction": true,
+  "concept_block_size": 4096,
+  "pad_multiple": 16,
+  "store_unknown_weights": false,
+  "inject_layer": 16,
+  "inject_alpha": 1.0,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.48.0",
+  "start_header_id": 100281,
+  "end_header_id": 100282,
+  "eot_id": 100283
+}