dacorvo HF staff commited on
Commit
e1149ae
1 Parent(s): c7cae9f

Synchronizing local compiler cache.

Browse files
neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.24.dev0/inference/llama/gorilla-llm/gorilla-openfunctions-v2/aa4259a04e2fbe240935.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 100000, "eos_token_id": 100015, "hidden_act": "silu", "hidden_size": 4096, "initializer_range": 0.02, "intermediate_size": 11008, "max_position_embeddings": 4096, "mlp_bias": false, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 4, "checkpoint_id": "gorilla-llm/gorilla-openfunctions-v2", "checkpoint_revision": "1f6ac3b8bb09e347a875f5ff08cd66798706341e", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 2, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 32, "num_hidden_layers": 30, "num_key_value_heads": 32, "pretraining_tp": 1, "rms_norm_eps": 1e-06, "rope_scaling": null, "rope_theta": 10000.0, "tie_word_embeddings": false, "torch_dtype": "bfloat16", "use_cache": false, "vocab_size": 102400}