Synchronizing local compiler cache.

.gitattributes

@@ -9788,3 +9788,7 @@ neuronxcc-2.19.8089.0+8ab9f450/MODULE_13139ab478d0ea6f3a81+ed72d204/model.neff f
 neuronxcc-2.19.8089.0+8ab9f450/MODULE_6e12eb6101ea12a7390a+a9d440f5/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.19.8089.0+8ab9f450/MODULE_6e12eb6101ea12a7390a+a9d440f5/wrapped_neff.hlo filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.19.8089.0+8ab9f450/MODULE_b01dd9ad474ff258b312+431f5505/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.19.8089.0+8ab9f450/MODULE_5ff114c824c56c9999f2+ed72d204/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.19.8089.0+8ab9f450/MODULE_6c58b851f8f08ff7fef9+a9d440f5/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.19.8089.0+8ab9f450/MODULE_6c58b851f8f08ff7fef9+a9d440f5/wrapped_neff.hlo filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.19.8089.0+8ab9f450/MODULE_ea2adeef395abde1f271+431f5505/model.neff filter=lfs diff=lfs merge=lfs -text

neuronxcc-2.19.8089.0+8ab9f450/0_REGISTRY/0.3.0.dev5/qwen3/Qwen/Qwen3-8B/001436812f32fabc8b7f.json

@@ -0,0 +1,72 @@
+{
+  "_entry_class": "SingleModelCacheEntry",
+  "_model_id": "Qwen/Qwen3-8B",
+  "_task": "text-generation",
+  "architectures": [
+    "Qwen3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "intermediate_size": 12288,
+  "max_position_embeddings": 40960,
+  "max_window_layers": 36,
+  "model_type": "qwen3",
+  "neuron": {
+    "_serialized_key": "NxDNeuronConfig",
+    "async_mode": false,
+    "attn_kernel_enabled": false,
+    "batch_size": 8,
+    "capacity_factor": null,
+    "cc_pipeline_tiling_factor": 2,
+    "checkpoint_id": "Qwen/Qwen3-8B",
+    "checkpoint_revision": "9c925d64d72725edaf899c6cb9c377fd0709d9c5",
+    "continuous_batching": true,
+    "enable_bucketing": false,
+    "ep_degree": 1,
+    "flash_decoding_enabled": false,
+    "fused_qkv": true,
+    "glu_mlp": true,
+    "is_chunked_prefill": false,
+    "local_ranks_size": 2,
+    "logical_nc_config": 1,
+    "max_batch_size": 8,
+    "max_context_length": 4096,
+    "max_topk": 256,
+    "mlp_kernel_enabled": false,
+    "mlp_kernel_fuse_residual_add": false,
+    "n_active_tokens": 4096,
+    "neuronxcc_version": "2.19.8089.0+8ab9f450",
+    "num_cores_per_group": 1,
+    "on_device_sampling": false,
+    "optimum_neuron_version": "0.3.0.dev5",
+    "output_logits": false,
+    "padding_side": "right",
+    "pp_degree": 1,
+    "qk_layernorm": false,
+    "qkv_kernel_enabled": false,
+    "rpl_reduce_dtype": "bfloat16",
+    "sequence_length": 4096,
+    "sequence_parallel_enabled": false,
+    "speculation_length": 0,
+    "start_rank_id": 0,
+    "target": null,
+    "torch_dtype": "bfloat16",
+    "tp_degree": 2,
+    "vocab_parallel": false
+  },
+  "num_attention_heads": 32,
+  "num_hidden_layers": 36,
+  "num_key_value_heads": 8,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 1000000,
+  "sliding_window": null,
+  "tie_word_embeddings": false,
+  "use_cache": true,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}

neuronxcc-2.19.8089.0+8ab9f450/MODULE_5ff114c824c56c9999f2+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-2.19.8089.0+8ab9f450/MODULE_5ff114c824c56c9999f2+ed72d204/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a09c958328e565bc208f8402f09e024339c8a1e8923c1057fb984f3ea43b50c5
+size 1086874

neuronxcc-2.19.8089.0+8ab9f450/MODULE_5ff114c824c56c9999f2+ed72d204/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ebcc13a1150998cce01c194a7f162beac7c3f979b9e3ee8ac206965721e82030
+size 7732224

neuronxcc-2.19.8089.0+8ab9f450/MODULE_6c58b851f8f08ff7fef9+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-2.19.8089.0+8ab9f450/MODULE_6c58b851f8f08ff7fef9+a9d440f5/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4fe38e19b502d032251ef8e40378b225f9d20552b9c247cc8df27cd15cf5a00b
+size 905051

neuronxcc-2.19.8089.0+8ab9f450/MODULE_6c58b851f8f08ff7fef9+a9d440f5/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c8df22d005dda780135fa622ee10562a0a62acf9a5888a1a1d7581561986c559
+size 9339904

neuronxcc-2.19.8089.0+8ab9f450/MODULE_6c58b851f8f08ff7fef9+a9d440f5/wrapped_neff.hlo

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c0ee6656211ed951ade994e508a98cd4ab2112c6fb9f1b3f8610e452b979c1b
+size 9504709

neuronxcc-2.19.8089.0+8ab9f450/MODULE_ea2adeef395abde1f271+431f5505/compile_flags.json

@@ -0,0 +1 @@
+"--model-type=transformer -O1 --lnc=1 --internal-hlo2tensorizer-options=--experimental-unsafe-fp8e4m3fn-as-fp8e4m3 --logfile=/tmp/nxd_model/layout_opt/log-neuron-cc.txt"

neuronxcc-2.19.8089.0+8ab9f450/MODULE_ea2adeef395abde1f271+431f5505/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:992b3fadc07c9c4737f882821fdb994671761cd7d74f38f0c46e415099107076
+size 166709

neuronxcc-2.19.8089.0+8ab9f450/MODULE_ea2adeef395abde1f271+431f5505/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c902d7a1a12c0a5881f32502e110377d776e47a6453d7179849e9bca907a1501
+size 5755904