Synchronizing local compiler cache.
Browse files
neuronxcc-2.19.8089.0+8ab9f450/0_REGISTRY/0.3.1.dev3/qwen3_moe/Qwen/Qwen3-30B-A3B-Instruct-2507/7d3ef321e1467f2184e9.json
ADDED
@@ -0,0 +1,79 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_entry_class": "SingleModelCacheEntry",
|
3 |
+
"_model_id": "Qwen/Qwen3-30B-A3B-Instruct-2507",
|
4 |
+
"_task": "text-generation",
|
5 |
+
"architectures": [
|
6 |
+
"Qwen3MoeForCausalLM"
|
7 |
+
],
|
8 |
+
"attention_bias": false,
|
9 |
+
"attention_dropout": 0.0,
|
10 |
+
"decoder_sparse_step": 1,
|
11 |
+
"head_dim": 128,
|
12 |
+
"hidden_act": "silu",
|
13 |
+
"hidden_size": 2048,
|
14 |
+
"initializer_range": 0.02,
|
15 |
+
"intermediate_size": 6144,
|
16 |
+
"max_position_embeddings": 262144,
|
17 |
+
"max_window_layers": 48,
|
18 |
+
"mlp_only_layers": [],
|
19 |
+
"model_type": "qwen3_moe",
|
20 |
+
"moe_intermediate_size": 768,
|
21 |
+
"neuron": {
|
22 |
+
"_serialized_key": "NxDNeuronConfig",
|
23 |
+
"async_mode": false,
|
24 |
+
"attn_kernel_enabled": false,
|
25 |
+
"batch_size": 4,
|
26 |
+
"capacity_factor": null,
|
27 |
+
"cc_pipeline_tiling_factor": 2,
|
28 |
+
"checkpoint_id": "Qwen/Qwen3-30B-A3B-Instruct-2507",
|
29 |
+
"checkpoint_revision": "61082d4deaa4785f64943b443cbc2b5de7524fad",
|
30 |
+
"continuous_batching": false,
|
31 |
+
"enable_bucketing": false,
|
32 |
+
"ep_degree": 1,
|
33 |
+
"flash_decoding_enabled": false,
|
34 |
+
"fused_qkv": false,
|
35 |
+
"glu_mlp": true,
|
36 |
+
"is_chunked_prefill": false,
|
37 |
+
"local_ranks_size": 8,
|
38 |
+
"logical_nc_config": 1,
|
39 |
+
"max_batch_size": 4,
|
40 |
+
"max_context_length": 4096,
|
41 |
+
"max_topk": 256,
|
42 |
+
"mlp_kernel_enabled": false,
|
43 |
+
"mlp_kernel_fuse_residual_add": false,
|
44 |
+
"n_active_tokens": 4096,
|
45 |
+
"neuronxcc_version": "2.19.8089.0+8ab9f450",
|
46 |
+
"num_cores_per_group": 1,
|
47 |
+
"on_device_sampling": false,
|
48 |
+
"optimum_neuron_version": "0.3.1.dev3",
|
49 |
+
"output_logits": false,
|
50 |
+
"padding_side": "right",
|
51 |
+
"pp_degree": 1,
|
52 |
+
"qkv_kernel_enabled": false,
|
53 |
+
"rpl_reduce_dtype": "bfloat16",
|
54 |
+
"sequence_length": 4096,
|
55 |
+
"sequence_parallel_enabled": false,
|
56 |
+
"speculation_length": 0,
|
57 |
+
"start_rank_id": 0,
|
58 |
+
"target": null,
|
59 |
+
"torch_dtype": "bfloat16",
|
60 |
+
"tp_degree": 8,
|
61 |
+
"vocab_parallel": false
|
62 |
+
},
|
63 |
+
"norm_topk_prob": true,
|
64 |
+
"num_attention_heads": 32,
|
65 |
+
"num_experts": 128,
|
66 |
+
"num_experts_per_tok": 8,
|
67 |
+
"num_hidden_layers": 48,
|
68 |
+
"num_key_value_heads": 4,
|
69 |
+
"output_router_logits": false,
|
70 |
+
"rms_norm_eps": 1e-06,
|
71 |
+
"rope_scaling": null,
|
72 |
+
"rope_theta": 10000000,
|
73 |
+
"router_aux_loss_coef": 0.001,
|
74 |
+
"sliding_window": null,
|
75 |
+
"tie_word_embeddings": false,
|
76 |
+
"use_cache": true,
|
77 |
+
"use_sliding_window": false,
|
78 |
+
"vocab_size": 151936
|
79 |
+
}
|