Synchronizing local compiler cache.

neuronxcc-2.16.372.0+4a9b2326/0_REGISTRY/0.2.0.dev2/inference/phi3/microsoft/Phi-3-mini-4k-instruct/bf6ec338cb34d5dded16.json

@@ -0,0 +1 @@
+{"_entry_class": "SingleModelCacheEntry", "_model_id": "microsoft/Phi-3-mini-4k-instruct", "_task": "text-generation", "architectures": ["Phi3ForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "auto_map": {"AutoConfig": "microsoft/Phi-3-mini-4k-instruct--configuration_phi3.Phi3Config", "AutoModelForCausalLM": "microsoft/Phi-3-mini-4k-instruct--modeling_phi3.Phi3ForCausalLM"}, "embd_pdrop": 0.0, "hidden_act": "silu", "hidden_size": 3072, "initializer_range": 0.02, "intermediate_size": 8192, "max_position_embeddings": 4096, "model_type": "phi3", "neuron": {"all_reduce_dtype": null, "allow_flash_attention": false, "attention_layout": "HSB", "attn_output_transposed": false, "auto_cast_type": "bf16", "backend": "hlo", "batch_size": 4, "checkpoint_id": "microsoft/Phi-3-mini-4k-instruct", "checkpoint_revision": "0a67737cc96d2554230f90338b163bc6380a2a85", "collectives_layout": "HSB", "continuous_batching": true, "fuse_qkv": true, "group_query_attention": null, "log_softmax_scores": false, "neuronxcc_version": "2.16.372.0+4a9b2326", "optimum_neuron_version": "0.2.0.dev2", "output_all_logits": false, "sequence_length": 4096, "tp_degree": 2}, "num_attention_heads": 32, "num_hidden_layers": 32, "num_key_value_heads": 32, "original_max_position_embeddings": 4096, "partial_rotary_factor": 1.0, "resid_pdrop": 0.0, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 10000.0, "sliding_window": 2047, "tie_word_embeddings": false, "use_cache": true, "vocab_size": 32064}