Synchronizing local compiler cache.

.gitattributes

@@ -11317,3 +11317,6 @@ neuronxcc-2.21.18209.0+043b1bf7/MODULE_fde1ed73c7f127095e6d+a9d440f5/wrapped_nef
 neuronxcc-2.21.18209.0+043b1bf7/MODULE_6a851b8869f56b4b2062+ed72d204/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.21.18209.0+043b1bf7/MODULE_f66180c31c8cb7fd0124+a9d440f5/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.21.18209.0+043b1bf7/MODULE_f66180c31c8cb7fd0124+a9d440f5/wrapped_neff.hlo filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.21.18209.0+043b1bf7/MODULE_8c86ec94a2915071eb43+ed72d204/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.21.18209.0+043b1bf7/MODULE_abfffd444fb508cfe8b8+a9d440f5/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.21.18209.0+043b1bf7/MODULE_abfffd444fb508cfe8b8+a9d440f5/wrapped_neff.hlo filter=lfs diff=lfs merge=lfs -text

neuronxcc-2.21.18209.0+043b1bf7/0_REGISTRY/0.3.1.dev7/llama/deepseek-ai/DeepSeek-R1-Distill-Llama-8B/9a86f16f5aef60b890f0.json

@@ -0,0 +1,63 @@
+{
+  "_entry_class": "SingleModelCacheEntry",
+  "_model_id": "deepseek-ai/DeepSeek-R1-Distill-Llama-8B",
+  "_task": "text-generation",
+  "architectures": [
+    "LlamaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "intermediate_size": 14336,
+  "max_position_embeddings": 131072,
+  "mlp_bias": false,
+  "model_type": "llama",
+  "neuron": {
+    "_serialized_key": "NxDNeuronConfig",
+    "batch_size": 16,
+    "capacity_factor": null,
+    "checkpoint_id": "deepseek-ai/DeepSeek-R1-Distill-Llama-8B",
+    "checkpoint_revision": "6a6f4aa4197940add57724a7707d069478df56b1",
+    "continuous_batching": true,
+    "enable_bucketing": false,
+    "ep_degree": 1,
+    "fused_qkv": true,
+    "glu_mlp": true,
+    "local_ranks_size": 8,
+    "logical_nc_config": 1,
+    "max_batch_size": 16,
+    "max_context_length": 4096,
+    "max_topk": 256,
+    "n_active_tokens": 4096,
+    "neuronxcc_version": "2.21.18209.0+043b1bf7",
+    "on_device_sampling": true,
+    "optimum_neuron_version": "0.3.1.dev7",
+    "output_logits": false,
+    "pp_degree": 1,
+    "sequence_length": 4096,
+    "speculation_length": 0,
+    "start_rank_id": 0,
+    "target": null,
+    "torch_dtype": "bfloat16",
+    "tp_degree": 8
+  },
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "num_key_value_heads": 8,
+  "pretraining_tp": 1,
+  "rms_norm_eps": 1e-05,
+  "rope_scaling": {
+    "factor": 8.0,
+    "high_freq_factor": 4.0,
+    "low_freq_factor": 1.0,
+    "original_max_position_embeddings": 8192,
+    "rope_type": "llama3"
+  },
+  "rope_theta": 500000.0,
+  "tie_word_embeddings": false,
+  "use_cache": true,
+  "vocab_size": 128256
+}

neuronxcc-2.21.18209.0+043b1bf7/MODULE_8c86ec94a2915071eb43+ed72d204/compile_flags.json

@@ -0,0 +1 @@
+["--target=trn1", "--auto-cast=none", "--model-type=transformer", "--tensorizer-options=--enable-ccop-compute-overlap --cc-pipeline-tiling-factor=2 --vectorize-strided-dma ", "-O2", "--lnc=1", "--logfile=/tmp/nxd_model/context_encoding_model/_tp0_bk0/log-neuron-cc.txt"]

neuronxcc-2.21.18209.0+043b1bf7/MODULE_8c86ec94a2915071eb43+ed72d204/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:09c37f172a403dcc8861eb15c6f20f96da7b22710956d82936c1632b716ed653
+size 972333

neuronxcc-2.21.18209.0+043b1bf7/MODULE_8c86ec94a2915071eb43+ed72d204/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d89733c3319e8fdbdb12be09523152d35f980d9048296a7a8479754b2db87b5f
+size 41892864

neuronxcc-2.21.18209.0+043b1bf7/MODULE_abfffd444fb508cfe8b8+a9d440f5/compile_flags.json

@@ -0,0 +1 @@
+["--target=trn1", "--auto-cast=none", "--model-type=transformer", "--tensorizer-options=--enable-ccop-compute-overlap --cc-pipeline-tiling-factor=2 --vectorize-strided-dma ", "-O2", "--lnc=1", "--logfile=/tmp/nxd_model/token_generation_model/_tp0_bk0/log-neuron-cc.txt", "--enable-internal-neff-wrapper"]

neuronxcc-2.21.18209.0+043b1bf7/MODULE_abfffd444fb508cfe8b8+a9d440f5/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:250d7c8577a9e4f494e177d6aeef843790699c4b6612d80d442ddd6760546b5b
+size 791025

neuronxcc-2.21.18209.0+043b1bf7/MODULE_abfffd444fb508cfe8b8+a9d440f5/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc387edb05cc204cb349460f1d3e5c79af072b93e2435831b24341f9577adb85
+size 5284864

neuronxcc-2.21.18209.0+043b1bf7/MODULE_abfffd444fb508cfe8b8+a9d440f5/wrapped_neff.hlo

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b8c0d08554ce92142ff840b3518481c20138d8a9ffaf3cd5d838ad77c4d89199
+size 5432471