Synchronizing local compiler cache.

.gitattributes

@@ -4712,3 +4712,5 @@ neuronxcc-2.15.128.0+56dc5a86/MODULE_9bb23b192495fde8d26c+39f12043/model.neff fi
 neuronxcc-2.15.128.0+56dc5a86/MODULE_a589a7a0e5eda0008c84+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.128.0+56dc5a86/MODULE_acfa3ff7ac7a29ae97c5+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.128.0+56dc5a86/MODULE_db69e5780be32d666926+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.128.0+56dc5a86/MODULE_111e74a8db1fe63fcd68+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.128.0+56dc5a86/MODULE_fa2e02356ff9d45f20e3+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text

neuronxcc-2.15.128.0+56dc5a86/0_REGISTRY/0.0.25.dev0/inference/llama/meta-llama/Meta-Llama-3-70B/48b2793659454a05dd2a.json

@@ -0,0 +1 @@
+{"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 128000, "eos_token_id": 128001, "hidden_act": "silu", "hidden_size": 8192, "initializer_range": 0.02, "intermediate_size": 28672, "max_position_embeddings": 8192, "mlp_bias": false, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 4, "checkpoint_id": "meta-llama/Meta-Llama-3-70B", "checkpoint_revision": "b4d08b7db49d488da3ac49adf25a6b9ac01ae338", "compiler_type": "neuronx-cc", "compiler_version": "2.15.128.0+56dc5a86", "num_cores": 24, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 64, "num_hidden_layers": 80, "num_key_value_heads": 8, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 500000.0, "tie_word_embeddings": false, "torch_dtype": "bfloat16", "use_cache": true, "vocab_size": 128256}

neuronxcc-2.15.128.0+56dc5a86/MODULE_111e74a8db1fe63fcd68+39f12043/compile_flags.json

@@ -0,0 +1 @@
+["--model-type=transformer", "--auto-cast=none", "--execute-repetition=1"]

neuronxcc-2.15.128.0+56dc5a86/MODULE_111e74a8db1fe63fcd68+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:18bf24a8fd241e998d5469758124099f1189e0c524d0164ea4f2ff970a9e07fb
+size 1152262

neuronxcc-2.15.128.0+56dc5a86/MODULE_111e74a8db1fe63fcd68+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8fe2d0d2f333a5393fa63114dbf425ff32ec5c79911e1af4b656bdbb9d921056
+size 2673664

neuronxcc-2.15.128.0+56dc5a86/MODULE_fa2e02356ff9d45f20e3+39f12043/compile_flags.json

@@ -0,0 +1 @@
+["--model-type=transformer", "--auto-cast=none", "--execute-repetition=1"]

neuronxcc-2.15.128.0+56dc5a86/MODULE_fa2e02356ff9d45f20e3+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:176d7998607262c14c0cb3aa6479a8feeb4ea76d3ae93fc1401c0a8d094b252b
+size 940916

neuronxcc-2.15.128.0+56dc5a86/MODULE_fa2e02356ff9d45f20e3+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:96d049d2b9f466efc157a7e32fe0dcbfdb54d6b7cca13b186cbc7f56e6571b75
+size 10886144