Update config.json

config.json

@@ -9,20 +9,13 @@
     "AutoConfig": "configuration_phi.PhiConfig",
     "AutoModelForCausalLM": "modeling_phi.PhiForCausalLM"
   },
-  "bos_token_id": null,
   "embd_pdrop": 0.0,
-  "eos_token_id": null,
   "flash_attn": false,
   "flash_rotary": false,
   "fused_dense": false,
-  "hidden_act": "gelu_new",
-  "hidden_size": 2560,
   "img_processor": null,
   "initializer_range": 0.02,
-  "intermediate_size": 8192,
-  "layer_norm_eps": 1e-05,
   "layer_norm_epsilon": 1e-05,
-  "max_position_embeddings": 2048,
   "model_type": "phi-msft",
   "n_embd": 2560,
   "n_head": 32,
@@ -30,23 +23,12 @@
   "n_inner": null,
   "n_layer": 32,
   "n_positions": 2048,
-  "num_attention_heads": 32,
-  "num_experts_per_tok": 2,
-  "num_hidden_layers": 24,
   "num_experts_per_tok": 2,
   "num_local_experts": 4,
-  "num_key_value_heads": 32,
-  "num_local_experts": 2,
-  "partial_rotary_factor": 0.5,
-  "qk_layernorm": false,
   "resid_pdrop": 0.1,
-  "rope_scaling": null,
-  "rope_theta": 10000.0,
   "rotary_dim": 32,
-  "sliding_window": null,
   "tie_word_embeddings": false,
   "torch_dtype": "float16",
-  "transformers_version": "4.37.1",
-  "use_cache": false,
+  "transformers_version": "4.35.2",
   "vocab_size": 51200
 }