Hugging Face's logo

aws-neuron
/
optimum-neuron-cache

Model card Files
xet
 Community
658
dacorvo HF Staff commited on
Commit
d102244
·
verified ·
1 Parent(s): 467d7ea

Synchronizing local compiler cache.

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. .gitattributes +41 -0
  2. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/0cc25526e2cfc37a8875a3752f33c4d7505d8a07b869d0f3f41915cf6e763b74/31629530ce956e3eb55e.json +59 -0
  3. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/2da7a00f0478d50ae1e7f75f085c5b2773b5f355f427c61cf34cb6febd629d96/6b7aebc14c435ae3db63.json +60 -0
  4. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/441269935591cad8d370e512c0b93cdd2fce6247c40e5a4866d872ee5338b0de/9208f2d66062f6c60fad.json +82 -0
  5. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/441269935591cad8d370e512c0b93cdd2fce6247c40e5a4866d872ee5338b0de/a421ae2ad1770f32b078.json +82 -0
  6. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/441269935591cad8d370e512c0b93cdd2fce6247c40e5a4866d872ee5338b0de/e912866445bcdd7bd532.json +82 -0
  7. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/4ab8140bc7eb4a553d95855c5c2be2cf8c0fbab21b823d76183b6f51e98b6fc5/1171cdc1e0668ecced1b.json +59 -0
  8. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/4cb7aff9e2a15c151396f2b684013e39d6739f0dec83e5c9dabbfe9d5fcf77b7/08703cff93ab144d67ec.json +83 -0
  9. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/4cb7aff9e2a15c151396f2b684013e39d6739f0dec83e5c9dabbfe9d5fcf77b7/21c3bc6d9846044edf76.json +83 -0
  10. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/4cb7aff9e2a15c151396f2b684013e39d6739f0dec83e5c9dabbfe9d5fcf77b7/2c6cc66c4776ff9413f0.json +83 -0
  11. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/60feecaa0c4c075e2f3e46a3f55d9a273f0ddd75a0ecf64e4ae27352e0819506/43be67542ceab3db3486.json +88 -0
  12. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/60feecaa0c4c075e2f3e46a3f55d9a273f0ddd75a0ecf64e4ae27352e0819506/d57c7c9107ac77a642d4.json +88 -0
  13. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/6454afdf3e9d66c7226c13a575b718845c25e53b0699600ba2bb4f883e9d841b/4b6c7416e39455d1ad2e.json +63 -0
  14. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/6454afdf3e9d66c7226c13a575b718845c25e53b0699600ba2bb4f883e9d841b/5d5179e2ac0b724548b0.json +63 -0
  15. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/6454afdf3e9d66c7226c13a575b718845c25e53b0699600ba2bb4f883e9d841b/8f8e8c8f417393bfd04a.json +63 -0
  16. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/73707b485eab9008c7aba7f5dad0ce2384ac685318d5f888c12fa0d81ed90b19/7536b906773aff032659.json +135 -0
  17. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/73707b485eab9008c7aba7f5dad0ce2384ac685318d5f888c12fa0d81ed90b19/87f1aaa5ae1a31bb7adc.json +135 -0
  18. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/73707b485eab9008c7aba7f5dad0ce2384ac685318d5f888c12fa0d81ed90b19/b2c32dabac1b1c034891.json +135 -0
  19. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/7518518c7e077820070186deda960d8cc49db068cdf0ac70664098fa2b6b698c/7e378b918d8aa52c31f9.json +65 -0
  20. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/920f44ce6d3e004d1ce547ae06644f7be262180644b04573153aa15d98742edc/ad0f2997db25e16e085b.json +66 -0
  21. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/929b02754a13cbfdf657d863c3fc6f3bce672879bc6ae48ab45be21e881e9ec2/947cd8633065b84ee800.json +88 -0
  22. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/929b02754a13cbfdf657d863c3fc6f3bce672879bc6ae48ab45be21e881e9ec2/c3b960b90a5e3c592c04.json +88 -0
  23. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/9c24a2080ec26e9c74d5bd90866f74131aa4d5c975415b96f5e6600f081d5647/d244406bccd80288e991.json +82 -0
  24. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/cf6b9a360dcf294104671106bae2adbd9fd291823bb60a351883163684073231/0fa4ed4626958e306002.json +64 -0
  25. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/cf6b9a360dcf294104671106bae2adbd9fd291823bb60a351883163684073231/124119d9f41b205009a0.json +64 -0
  26. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/cf6b9a360dcf294104671106bae2adbd9fd291823bb60a351883163684073231/80f4956a4a5775280772.json +64 -0
  27. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/cf6b9a360dcf294104671106bae2adbd9fd291823bb60a351883163684073231/9df84ad1fbfa85b8d13d.json +64 -0
  28. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/cf6b9a360dcf294104671106bae2adbd9fd291823bb60a351883163684073231/ce4667775a41bf1523e4.json +64 -0
  29. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/d139acf64685f15794bb983ff6eb881bdd31304bae88b0ce1ed20a54c21f2265/4aab0cb19b60948b084e.json +59 -0
  30. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/gemma3_text/unsloth/gemma-3-270m-it/a421ae2ad1770f32b078.json +82 -0
  31. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/granite/hf-internal-testing/tiny-random-GraniteForCausalLM/4aab0cb19b60948b084e.json +59 -0
  32. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/granite/ibm-granite/granite-3.1-2b-instruct/31629530ce956e3eb55e.json +59 -0
  33. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/llama/llamafactory/tiny-random-Llama-3/5d5179e2ac0b724548b0.json +63 -0
  34. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/llama/unsloth/Llama-3.2-1B-Instruct/9df84ad1fbfa85b8d13d.json +64 -0
  35. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/llama4_text/tiny-random/llama-4/d244406bccd80288e991.json +82 -0
  36. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/mixtral/dacorvo/Mixtral-tiny/1171cdc1e0668ecced1b.json +59 -0
  37. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/phi3/yujiepan/phi-4-tiny-random/6b7aebc14c435ae3db63.json +60 -0
  38. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/qwen2/Qwen/Qwen2.5-0.5B/2c6cc66c4776ff9413f0.json +83 -0
  39. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/qwen2/yujiepan/qwen2.5-128k-tiny-random/7e378b918d8aa52c31f9.json +65 -0
  40. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/qwen3/Qwen/Qwen3-0.6B/c3b960b90a5e3c592c04.json +88 -0
  41. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/qwen3/Qwen/Qwen3-Embedding-0.6B/43be67542ceab3db3486.json +88 -0
  42. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/qwen3_moe/optimum-internal-testing/tiny-random-qwen3_moe/ad0f2997db25e16e085b.json +66 -0
  43. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/smollm3/HuggingFaceTB/SmolLM3-3B/b2c32dabac1b1c034891.json +135 -0
  44. neuronxcc-2.21.33363.0+82129205/MODULE_10244305442015770634+e30acd3a/model.neff +0 -0
  45. neuronxcc-2.21.33363.0+82129205/MODULE_10606948783918825529+e30acd3a/model.neff +0 -0
  46. neuronxcc-2.21.33363.0+82129205/MODULE_10645643398657092095+e30acd3a/model.neff +0 -0
  47. neuronxcc-2.21.33363.0+82129205/MODULE_10661660426924300837+e30acd3a/model.neff +0 -0
  48. neuronxcc-2.21.33363.0+82129205/MODULE_10746122569655005679+e30acd3a/model.neff +0 -0
  49. neuronxcc-2.21.33363.0+82129205/MODULE_11086318750207148626+e30acd3a/model.neff +0 -0
  50. neuronxcc-2.21.33363.0+82129205/MODULE_11301370888592189606+e30acd3a/compile_flags.json +1 -0
.gitattributes CHANGED
@@ -17062,3 +17062,44 @@ neuronxcc-2.21.33363.0+82129205/MODULE_f17348c87fa73ad85ce1+909b2bae/model.neff
17062
  neuronxcc-2.21.33363.0+82129205/MODULE_f17348c87fa73ad85ce1+909b2bae/wrapped_neff.hlo filter=lfs diff=lfs merge=lfs -text
17063
  neuronxcc-2.21.33363.0+82129205/MODULE_fdc7603ab4547da30e1f+31905a9b/model.neff filter=lfs diff=lfs merge=lfs -text
17064
  neuronxcc-2.21.33363.0+82129205/MODULE_fdc7603ab4547da30e1f+31905a9b/wrapped_neff.hlo filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
17062
  neuronxcc-2.21.33363.0+82129205/MODULE_f17348c87fa73ad85ce1+909b2bae/wrapped_neff.hlo filter=lfs diff=lfs merge=lfs -text
17063
  neuronxcc-2.21.33363.0+82129205/MODULE_fdc7603ab4547da30e1f+31905a9b/model.neff filter=lfs diff=lfs merge=lfs -text
17064
  neuronxcc-2.21.33363.0+82129205/MODULE_fdc7603ab4547da30e1f+31905a9b/wrapped_neff.hlo filter=lfs diff=lfs merge=lfs -text
17065
+ neuronxcc-2.21.33363.0+82129205/MODULE_11301370888592189606+e30acd3a/model.neff filter=lfs diff=lfs merge=lfs -text
17066
+ neuronxcc-2.21.33363.0+82129205/MODULE_137d2b1e9d185656d517+6170d8e1/model.neff filter=lfs diff=lfs merge=lfs -text
17067
+ neuronxcc-2.21.33363.0+82129205/MODULE_14657035910656097540+e30acd3a/model.neff filter=lfs diff=lfs merge=lfs -text
17068
+ neuronxcc-2.21.33363.0+82129205/MODULE_15128894797204950836+e30acd3a/model.neff filter=lfs diff=lfs merge=lfs -text
17069
+ neuronxcc-2.21.33363.0+82129205/MODULE_18044ed346a94d78250a+24129607/model.neff filter=lfs diff=lfs merge=lfs -text
17070
+ neuronxcc-2.21.33363.0+82129205/MODULE_1adb90886791031c9f74+06b21f21/model.neff filter=lfs diff=lfs merge=lfs -text
17071
+ neuronxcc-2.21.33363.0+82129205/MODULE_1adb90886791031c9f74+06b21f21/wrapped_neff.hlo filter=lfs diff=lfs merge=lfs -text
17072
+ neuronxcc-2.21.33363.0+82129205/MODULE_2060474045114402138+e30acd3a/model.neff filter=lfs diff=lfs merge=lfs -text
17073
+ neuronxcc-2.21.33363.0+82129205/MODULE_2fda55dcd99cd2bf7c01+74e4ea3a/model.neff filter=lfs diff=lfs merge=lfs -text
17074
+ neuronxcc-2.21.33363.0+82129205/MODULE_2fda55dcd99cd2bf7c01+74e4ea3a/wrapped_neff.hlo filter=lfs diff=lfs merge=lfs -text
17075
+ neuronxcc-2.21.33363.0+82129205/MODULE_47bbfc2252c9a9fd3c68+a02c3a36/model.neff filter=lfs diff=lfs merge=lfs -text
17076
+ neuronxcc-2.21.33363.0+82129205/MODULE_47bbfc2252c9a9fd3c68+a02c3a36/wrapped_neff.hlo filter=lfs diff=lfs merge=lfs -text
17077
+ neuronxcc-2.21.33363.0+82129205/MODULE_522781774461263035+e30acd3a/model.neff filter=lfs diff=lfs merge=lfs -text
17078
+ neuronxcc-2.21.33363.0+82129205/MODULE_5620f0d58954f88c2890+24129607/model.neff filter=lfs diff=lfs merge=lfs -text
17079
+ neuronxcc-2.21.33363.0+82129205/MODULE_56637414cad2eda298a8+066be99d/model.neff filter=lfs diff=lfs merge=lfs -text
17080
+ neuronxcc-2.21.33363.0+82129205/MODULE_56637414cad2eda298a8+066be99d/wrapped_neff.hlo filter=lfs diff=lfs merge=lfs -text
17081
+ neuronxcc-2.21.33363.0+82129205/MODULE_5a955c273582832570b1+8ecabf0c/model.neff filter=lfs diff=lfs merge=lfs -text
17082
+ neuronxcc-2.21.33363.0+82129205/MODULE_5a955c273582832570b1+8ecabf0c/wrapped_neff.hlo filter=lfs diff=lfs merge=lfs -text
17083
+ neuronxcc-2.21.33363.0+82129205/MODULE_691f0ef9fca273527dbe+24129607/model.neff filter=lfs diff=lfs merge=lfs -text
17084
+ neuronxcc-2.21.33363.0+82129205/MODULE_71833bfb10ff3967bd3d+6170d8e1/model.neff filter=lfs diff=lfs merge=lfs -text
17085
+ neuronxcc-2.21.33363.0+82129205/MODULE_7247213647965899113+e30acd3a/model.neff filter=lfs diff=lfs merge=lfs -text
17086
+ neuronxcc-2.21.33363.0+82129205/MODULE_7671391308733910418+e30acd3a/model.neff filter=lfs diff=lfs merge=lfs -text
17087
+ neuronxcc-2.21.33363.0+82129205/MODULE_7975285533996118114+e30acd3a/model.neff filter=lfs diff=lfs merge=lfs -text
17088
+ neuronxcc-2.21.33363.0+82129205/MODULE_9176004544586841168+e30acd3a/model.neff filter=lfs diff=lfs merge=lfs -text
17089
+ neuronxcc-2.21.33363.0+82129205/MODULE_95b9072f246645f24461+a02c3a36/model.neff filter=lfs diff=lfs merge=lfs -text
17090
+ neuronxcc-2.21.33363.0+82129205/MODULE_95b9072f246645f24461+a02c3a36/wrapped_neff.hlo filter=lfs diff=lfs merge=lfs -text
17091
+ neuronxcc-2.21.33363.0+82129205/MODULE_9ba42250590aa5875bc2+57a3a5ff/model.neff filter=lfs diff=lfs merge=lfs -text
17092
+ neuronxcc-2.21.33363.0+82129205/MODULE_9ba42250590aa5875bc2+57a3a5ff/wrapped_neff.hlo filter=lfs diff=lfs merge=lfs -text
17093
+ neuronxcc-2.21.33363.0+82129205/MODULE_a3c6a617f4311e09ad49+a02c3a36/model.neff filter=lfs diff=lfs merge=lfs -text
17094
+ neuronxcc-2.21.33363.0+82129205/MODULE_a3c6a617f4311e09ad49+a02c3a36/wrapped_neff.hlo filter=lfs diff=lfs merge=lfs -text
17095
+ neuronxcc-2.21.33363.0+82129205/MODULE_ab293424ba3cf7b684c0+6170d8e1/model.neff filter=lfs diff=lfs merge=lfs -text
17096
+ neuronxcc-2.21.33363.0+82129205/MODULE_b097ceba291117567f12+f6339882/model.neff filter=lfs diff=lfs merge=lfs -text
17097
+ neuronxcc-2.21.33363.0+82129205/MODULE_b097ceba291117567f12+f6339882/wrapped_neff.hlo filter=lfs diff=lfs merge=lfs -text
17098
+ neuronxcc-2.21.33363.0+82129205/MODULE_c68ed9d0d3211fb37165+24129607/model.neff filter=lfs diff=lfs merge=lfs -text
17099
+ neuronxcc-2.21.33363.0+82129205/MODULE_c69743a3b026cb8c0f9e+a02c3a36/model.neff filter=lfs diff=lfs merge=lfs -text
17100
+ neuronxcc-2.21.33363.0+82129205/MODULE_c69743a3b026cb8c0f9e+a02c3a36/wrapped_neff.hlo filter=lfs diff=lfs merge=lfs -text
17101
+ neuronxcc-2.21.33363.0+82129205/MODULE_cafe5e1cc124bf1d35e0+a02c3a36/model.neff filter=lfs diff=lfs merge=lfs -text
17102
+ neuronxcc-2.21.33363.0+82129205/MODULE_cafe5e1cc124bf1d35e0+a02c3a36/wrapped_neff.hlo filter=lfs diff=lfs merge=lfs -text
17103
+ neuronxcc-2.21.33363.0+82129205/MODULE_d50f09930f3886ac4b66+a02c3a36/model.neff filter=lfs diff=lfs merge=lfs -text
17104
+ neuronxcc-2.21.33363.0+82129205/MODULE_d50f09930f3886ac4b66+a02c3a36/wrapped_neff.hlo filter=lfs diff=lfs merge=lfs -text
17105
+ neuronxcc-2.21.33363.0+82129205/MODULE_da7543b73fbd9c77b695+6170d8e1/model.neff filter=lfs diff=lfs merge=lfs -text
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/0cc25526e2cfc37a8875a3752f33c4d7505d8a07b869d0f3f41915cf6e763b74/31629530ce956e3eb55e.json ADDED
@@ -0,0 +1,59 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "ibm-granite/granite-3.1-2b-instruct",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "GraniteForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.1,
10
+ "attention_multiplier": 0.015625,
11
+ "dtype": "bfloat16",
12
+ "embedding_multiplier": 12.0,
13
+ "hidden_act": "silu",
14
+ "hidden_size": 2048,
15
+ "initializer_range": 0.02,
16
+ "intermediate_size": 8192,
17
+ "logits_scaling": 8.0,
18
+ "max_position_embeddings": 131072,
19
+ "mlp_bias": false,
20
+ "model_type": "granite",
21
+ "neuron": {
22
+ "_serialized_key": "NxDNeuronConfig",
23
+ "batch_size": 1,
24
+ "capacity_factor": null,
25
+ "checkpoint_id": "ibm-granite/granite-3.1-2b-instruct",
26
+ "checkpoint_revision": "bbc2aed595bd38bd770263dc3ab831db9794441d",
27
+ "continuous_batching": false,
28
+ "ep_degree": 1,
29
+ "fused_qkv": true,
30
+ "glu_mlp": true,
31
+ "local_ranks_size": 2,
32
+ "max_batch_size": 1,
33
+ "max_context_length": 8192,
34
+ "max_topk": 256,
35
+ "n_active_tokens": 8192,
36
+ "neuronxcc_version": "2.21.33363.0+82129205",
37
+ "on_device_sampling": true,
38
+ "optimum_neuron_version": "0.4.6.dev3",
39
+ "output_logits": false,
40
+ "pp_degree": 1,
41
+ "prefill_chunk_size": 1024,
42
+ "sequence_length": 8192,
43
+ "speculation_length": 0,
44
+ "start_rank_id": 0,
45
+ "target": "trn1",
46
+ "torch_dtype": "bfloat16",
47
+ "tp_degree": 2
48
+ },
49
+ "num_attention_heads": 32,
50
+ "num_hidden_layers": 40,
51
+ "num_key_value_heads": 8,
52
+ "residual_multiplier": 0.22,
53
+ "rms_norm_eps": 1e-05,
54
+ "rope_scaling": null,
55
+ "rope_theta": 5000000.0,
56
+ "tie_word_embeddings": true,
57
+ "use_cache": true,
58
+ "vocab_size": 49155
59
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/2da7a00f0478d50ae1e7f75f085c5b2773b5f355f427c61cf34cb6febd629d96/6b7aebc14c435ae3db63.json ADDED
@@ -0,0 +1,60 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "yujiepan/phi-4-tiny-random",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "Phi3ForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "auto_map": {},
11
+ "dtype": "bfloat16",
12
+ "embd_pdrop": 0.0,
13
+ "hidden_act": "silu",
14
+ "hidden_size": 16,
15
+ "initializer_range": 0.02,
16
+ "intermediate_size": 32,
17
+ "max_position_embeddings": 16384,
18
+ "model_type": "phi3",
19
+ "neuron": {
20
+ "_serialized_key": "NxDNeuronConfig",
21
+ "batch_size": 1,
22
+ "capacity_factor": null,
23
+ "checkpoint_id": "yujiepan/phi-4-tiny-random",
24
+ "checkpoint_revision": "18a9a1168dc97ac6d128f811925670c275610f5a",
25
+ "continuous_batching": false,
26
+ "ep_degree": 1,
27
+ "fused_qkv": true,
28
+ "glu_mlp": true,
29
+ "local_ranks_size": 2,
30
+ "max_batch_size": 1,
31
+ "max_context_length": 1024,
32
+ "max_topk": 256,
33
+ "n_active_tokens": 1024,
34
+ "neuronxcc_version": "2.21.33363.0+82129205",
35
+ "on_device_sampling": true,
36
+ "optimum_neuron_version": "0.4.6.dev3",
37
+ "output_logits": false,
38
+ "pp_degree": 1,
39
+ "prefill_chunk_size": 0,
40
+ "sequence_length": 1024,
41
+ "speculation_length": 0,
42
+ "start_rank_id": 0,
43
+ "target": "trn1",
44
+ "torch_dtype": "bfloat16",
45
+ "tp_degree": 2
46
+ },
47
+ "num_attention_heads": 2,
48
+ "num_hidden_layers": 2,
49
+ "num_key_value_heads": 1,
50
+ "original_max_position_embeddings": 16384,
51
+ "partial_rotary_factor": 1.0,
52
+ "resid_pdrop": 0.0,
53
+ "rms_norm_eps": 1e-05,
54
+ "rope_scaling": null,
55
+ "rope_theta": 250000,
56
+ "sliding_window": null,
57
+ "tie_word_embeddings": false,
58
+ "use_cache": true,
59
+ "vocab_size": 100352
60
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/441269935591cad8d370e512c0b93cdd2fce6247c40e5a4866d872ee5338b0de/9208f2d66062f6c60fad.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "unsloth/gemma-3-270m-it",
4
+ "_sliding_window_pattern": 6,
5
+ "_task": "text-generation",
6
+ "architectures": [
7
+ "Gemma3ForCausalLM"
8
+ ],
9
+ "attention_bias": false,
10
+ "attention_dropout": 0.0,
11
+ "attn_logit_softcapping": null,
12
+ "dtype": "bfloat16",
13
+ "final_logit_softcapping": null,
14
+ "head_dim": 256,
15
+ "hidden_activation": "gelu_pytorch_tanh",
16
+ "hidden_size": 640,
17
+ "initializer_range": 0.02,
18
+ "intermediate_size": 2048,
19
+ "layer_types": [
20
+ "sliding_attention",
21
+ "sliding_attention",
22
+ "sliding_attention",
23
+ "sliding_attention",
24
+ "sliding_attention",
25
+ "full_attention",
26
+ "sliding_attention",
27
+ "sliding_attention",
28
+ "sliding_attention",
29
+ "sliding_attention",
30
+ "sliding_attention",
31
+ "full_attention",
32
+ "sliding_attention",
33
+ "sliding_attention",
34
+ "sliding_attention",
35
+ "sliding_attention",
36
+ "sliding_attention",
37
+ "full_attention"
38
+ ],
39
+ "max_position_embeddings": 32768,
40
+ "model_type": "gemma3_text",
41
+ "neuron": {
42
+ "_serialized_key": "NxDNeuronConfig",
43
+ "batch_size": 1,
44
+ "capacity_factor": null,
45
+ "checkpoint_id": "unsloth/gemma-3-270m-it",
46
+ "checkpoint_revision": "23cf460f6bb16954176b3ddcc8d4f250501458a9",
47
+ "continuous_batching": false,
48
+ "ep_degree": 1,
49
+ "fused_qkv": true,
50
+ "glu_mlp": true,
51
+ "local_ranks_size": 2,
52
+ "max_batch_size": 1,
53
+ "max_context_length": 8192,
54
+ "max_topk": 256,
55
+ "n_active_tokens": 8192,
56
+ "neuronxcc_version": "2.21.33363.0+82129205",
57
+ "on_device_sampling": true,
58
+ "optimum_neuron_version": "0.4.6.dev3",
59
+ "output_logits": false,
60
+ "pp_degree": 1,
61
+ "prefill_chunk_size": 0,
62
+ "sequence_length": 8192,
63
+ "speculation_length": 0,
64
+ "start_rank_id": 0,
65
+ "target": "trn1",
66
+ "torch_dtype": "bfloat16",
67
+ "tp_degree": 2
68
+ },
69
+ "num_attention_heads": 4,
70
+ "num_hidden_layers": 18,
71
+ "num_key_value_heads": 1,
72
+ "query_pre_attn_scalar": 256,
73
+ "rms_norm_eps": 1e-06,
74
+ "rope_local_base_freq": 10000.0,
75
+ "rope_scaling": null,
76
+ "rope_theta": 1000000.0,
77
+ "sliding_window": 512,
78
+ "unsloth_fixed": true,
79
+ "use_bidirectional_attention": false,
80
+ "use_cache": true,
81
+ "vocab_size": 262144
82
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/441269935591cad8d370e512c0b93cdd2fce6247c40e5a4866d872ee5338b0de/a421ae2ad1770f32b078.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "unsloth/gemma-3-270m-it",
4
+ "_sliding_window_pattern": 6,
5
+ "_task": "text-generation",
6
+ "architectures": [
7
+ "Gemma3ForCausalLM"
8
+ ],
9
+ "attention_bias": false,
10
+ "attention_dropout": 0.0,
11
+ "attn_logit_softcapping": null,
12
+ "dtype": "bfloat16",
13
+ "final_logit_softcapping": null,
14
+ "head_dim": 256,
15
+ "hidden_activation": "gelu_pytorch_tanh",
16
+ "hidden_size": 640,
17
+ "initializer_range": 0.02,
18
+ "intermediate_size": 2048,
19
+ "layer_types": [
20
+ "sliding_attention",
21
+ "sliding_attention",
22
+ "sliding_attention",
23
+ "sliding_attention",
24
+ "sliding_attention",
25
+ "full_attention",
26
+ "sliding_attention",
27
+ "sliding_attention",
28
+ "sliding_attention",
29
+ "sliding_attention",
30
+ "sliding_attention",
31
+ "full_attention",
32
+ "sliding_attention",
33
+ "sliding_attention",
34
+ "sliding_attention",
35
+ "sliding_attention",
36
+ "sliding_attention",
37
+ "full_attention"
38
+ ],
39
+ "max_position_embeddings": 32768,
40
+ "model_type": "gemma3_text",
41
+ "neuron": {
42
+ "_serialized_key": "NxDNeuronConfig",
43
+ "batch_size": 4,
44
+ "capacity_factor": null,
45
+ "checkpoint_id": "unsloth/gemma-3-270m-it",
46
+ "checkpoint_revision": "23cf460f6bb16954176b3ddcc8d4f250501458a9",
47
+ "continuous_batching": true,
48
+ "ep_degree": 1,
49
+ "fused_qkv": true,
50
+ "glu_mlp": true,
51
+ "local_ranks_size": 2,
52
+ "max_batch_size": 4,
53
+ "max_context_length": 1024,
54
+ "max_topk": 256,
55
+ "n_active_tokens": 1024,
56
+ "neuronxcc_version": "2.21.33363.0+82129205",
57
+ "on_device_sampling": true,
58
+ "optimum_neuron_version": "0.4.6.dev3",
59
+ "output_logits": false,
60
+ "pp_degree": 1,
61
+ "prefill_chunk_size": 0,
62
+ "sequence_length": 1024,
63
+ "speculation_length": 0,
64
+ "start_rank_id": 0,
65
+ "target": "trn1",
66
+ "torch_dtype": "bfloat16",
67
+ "tp_degree": 2
68
+ },
69
+ "num_attention_heads": 4,
70
+ "num_hidden_layers": 18,
71
+ "num_key_value_heads": 1,
72
+ "query_pre_attn_scalar": 256,
73
+ "rms_norm_eps": 1e-06,
74
+ "rope_local_base_freq": 10000.0,
75
+ "rope_scaling": null,
76
+ "rope_theta": 1000000.0,
77
+ "sliding_window": 512,
78
+ "unsloth_fixed": true,
79
+ "use_bidirectional_attention": false,
80
+ "use_cache": true,
81
+ "vocab_size": 262144
82
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/441269935591cad8d370e512c0b93cdd2fce6247c40e5a4866d872ee5338b0de/e912866445bcdd7bd532.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "unsloth/gemma-3-270m-it",
4
+ "_sliding_window_pattern": 6,
5
+ "_task": "text-generation",
6
+ "architectures": [
7
+ "Gemma3ForCausalLM"
8
+ ],
9
+ "attention_bias": false,
10
+ "attention_dropout": 0.0,
11
+ "attn_logit_softcapping": null,
12
+ "dtype": "bfloat16",
13
+ "final_logit_softcapping": null,
14
+ "head_dim": 256,
15
+ "hidden_activation": "gelu_pytorch_tanh",
16
+ "hidden_size": 640,
17
+ "initializer_range": 0.02,
18
+ "intermediate_size": 2048,
19
+ "layer_types": [
20
+ "sliding_attention",
21
+ "sliding_attention",
22
+ "sliding_attention",
23
+ "sliding_attention",
24
+ "sliding_attention",
25
+ "full_attention",
26
+ "sliding_attention",
27
+ "sliding_attention",
28
+ "sliding_attention",
29
+ "sliding_attention",
30
+ "sliding_attention",
31
+ "full_attention",
32
+ "sliding_attention",
33
+ "sliding_attention",
34
+ "sliding_attention",
35
+ "sliding_attention",
36
+ "sliding_attention",
37
+ "full_attention"
38
+ ],
39
+ "max_position_embeddings": 32768,
40
+ "model_type": "gemma3_text",
41
+ "neuron": {
42
+ "_serialized_key": "NxDNeuronConfig",
43
+ "batch_size": 1,
44
+ "capacity_factor": null,
45
+ "checkpoint_id": "unsloth/gemma-3-270m-it",
46
+ "checkpoint_revision": "23cf460f6bb16954176b3ddcc8d4f250501458a9",
47
+ "continuous_batching": false,
48
+ "ep_degree": 1,
49
+ "fused_qkv": true,
50
+ "glu_mlp": true,
51
+ "local_ranks_size": 2,
52
+ "max_batch_size": 1,
53
+ "max_context_length": 1024,
54
+ "max_topk": 256,
55
+ "n_active_tokens": 1024,
56
+ "neuronxcc_version": "2.21.33363.0+82129205",
57
+ "on_device_sampling": true,
58
+ "optimum_neuron_version": "0.4.6.dev3",
59
+ "output_logits": false,
60
+ "pp_degree": 1,
61
+ "prefill_chunk_size": 0,
62
+ "sequence_length": 1024,
63
+ "speculation_length": 0,
64
+ "start_rank_id": 0,
65
+ "target": "trn1",
66
+ "torch_dtype": "bfloat16",
67
+ "tp_degree": 2
68
+ },
69
+ "num_attention_heads": 4,
70
+ "num_hidden_layers": 18,
71
+ "num_key_value_heads": 1,
72
+ "query_pre_attn_scalar": 256,
73
+ "rms_norm_eps": 1e-06,
74
+ "rope_local_base_freq": 10000.0,
75
+ "rope_scaling": null,
76
+ "rope_theta": 1000000.0,
77
+ "sliding_window": 512,
78
+ "unsloth_fixed": true,
79
+ "use_bidirectional_attention": false,
80
+ "use_cache": true,
81
+ "vocab_size": 262144
82
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/4ab8140bc7eb4a553d95855c5c2be2cf8c0fbab21b823d76183b6f51e98b6fc5/1171cdc1e0668ecced1b.json ADDED
@@ -0,0 +1,59 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "dacorvo/Mixtral-tiny",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "MixtralForCausalLM"
7
+ ],
8
+ "attention_dropout": 0.0,
9
+ "dtype": "float16",
10
+ "head_dim": 32,
11
+ "hidden_act": "silu",
12
+ "hidden_size": 1024,
13
+ "initializer_range": 0.02,
14
+ "intermediate_size": 3584,
15
+ "max_position_embeddings": 1024,
16
+ "model_type": "mixtral",
17
+ "neuron": {
18
+ "_serialized_key": "NxDNeuronConfig",
19
+ "batch_size": 1,
20
+ "capacity_factor": null,
21
+ "checkpoint_id": "dacorvo/Mixtral-tiny",
22
+ "checkpoint_revision": "c557ba205ddff6ea911f4719e0d543d6c08356b6",
23
+ "continuous_batching": false,
24
+ "ep_degree": 1,
25
+ "fused_qkv": false,
26
+ "glu_mlp": true,
27
+ "local_ranks_size": 2,
28
+ "max_batch_size": 1,
29
+ "max_context_length": 1024,
30
+ "max_topk": 256,
31
+ "n_active_tokens": 1024,
32
+ "neuronxcc_version": "2.21.33363.0+82129205",
33
+ "on_device_sampling": false,
34
+ "optimum_neuron_version": "0.4.6.dev3",
35
+ "output_logits": false,
36
+ "pp_degree": 1,
37
+ "prefill_chunk_size": 0,
38
+ "sequence_length": 1024,
39
+ "speculation_length": 0,
40
+ "start_rank_id": 0,
41
+ "target": "trn1",
42
+ "torch_dtype": "float16",
43
+ "tp_degree": 2
44
+ },
45
+ "num_attention_heads": 32,
46
+ "num_experts_per_tok": 2,
47
+ "num_hidden_layers": 2,
48
+ "num_key_value_heads": 8,
49
+ "num_local_experts": 8,
50
+ "output_router_logits": false,
51
+ "rms_norm_eps": 1e-05,
52
+ "rope_theta": 10000.0,
53
+ "router_aux_loss_coef": 0.001,
54
+ "router_jitter_noise": 0.0,
55
+ "sliding_window": 4096,
56
+ "tie_word_embeddings": false,
57
+ "use_cache": true,
58
+ "vocab_size": 32000
59
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/4cb7aff9e2a15c151396f2b684013e39d6739f0dec83e5c9dabbfe9d5fcf77b7/08703cff93ab144d67ec.json ADDED
@@ -0,0 +1,83 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "Qwen/Qwen2.5-0.5B",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "Qwen2ForCausalLM"
7
+ ],
8
+ "attention_dropout": 0.0,
9
+ "dtype": "bfloat16",
10
+ "hidden_act": "silu",
11
+ "hidden_size": 896,
12
+ "initializer_range": 0.02,
13
+ "intermediate_size": 4864,
14
+ "layer_types": [
15
+ "full_attention",
16
+ "full_attention",
17
+ "full_attention",
18
+ "full_attention",
19
+ "full_attention",
20
+ "full_attention",
21
+ "full_attention",
22
+ "full_attention",
23
+ "full_attention",
24
+ "full_attention",
25
+ "full_attention",
26
+ "full_attention",
27
+ "full_attention",
28
+ "full_attention",
29
+ "full_attention",
30
+ "full_attention",
31
+ "full_attention",
32
+ "full_attention",
33
+ "full_attention",
34
+ "full_attention",
35
+ "full_attention",
36
+ "full_attention",
37
+ "full_attention",
38
+ "full_attention"
39
+ ],
40
+ "max_position_embeddings": 32768,
41
+ "max_window_layers": 24,
42
+ "model_type": "qwen2",
43
+ "neuron": {
44
+ "_serialized_key": "NxDNeuronConfig",
45
+ "batch_size": 1,
46
+ "capacity_factor": null,
47
+ "checkpoint_id": "Qwen/Qwen2.5-0.5B",
48
+ "checkpoint_revision": "060db6499f32faf8b98477b0a26969ef7d8b9987",
49
+ "continuous_batching": false,
50
+ "ep_degree": 1,
51
+ "fused_qkv": false,
52
+ "glu_mlp": true,
53
+ "local_ranks_size": 2,
54
+ "max_batch_size": 1,
55
+ "max_context_length": 8192,
56
+ "max_topk": 256,
57
+ "n_active_tokens": 8192,
58
+ "neuronxcc_version": "2.21.33363.0+82129205",
59
+ "on_device_sampling": true,
60
+ "optimum_neuron_version": "0.4.6.dev3",
61
+ "output_logits": false,
62
+ "pp_degree": 1,
63
+ "prefill_chunk_size": 1024,
64
+ "sequence_length": 8192,
65
+ "speculation_length": 0,
66
+ "start_rank_id": 0,
67
+ "target": "trn1",
68
+ "torch_dtype": "bfloat16",
69
+ "tp_degree": 2
70
+ },
71
+ "num_attention_heads": 14,
72
+ "num_hidden_layers": 24,
73
+ "num_key_value_heads": 2,
74
+ "rms_norm_eps": 1e-06,
75
+ "rope_scaling": null,
76
+ "rope_theta": 1000000.0,
77
+ "sliding_window": null,
78
+ "tie_word_embeddings": true,
79
+ "use_cache": true,
80
+ "use_mrope": false,
81
+ "use_sliding_window": false,
82
+ "vocab_size": 151936
83
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/4cb7aff9e2a15c151396f2b684013e39d6739f0dec83e5c9dabbfe9d5fcf77b7/21c3bc6d9846044edf76.json ADDED
@@ -0,0 +1,83 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "Qwen/Qwen2.5-0.5B",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "Qwen2ForCausalLM"
7
+ ],
8
+ "attention_dropout": 0.0,
9
+ "dtype": "bfloat16",
10
+ "hidden_act": "silu",
11
+ "hidden_size": 896,
12
+ "initializer_range": 0.02,
13
+ "intermediate_size": 4864,
14
+ "layer_types": [
15
+ "full_attention",
16
+ "full_attention",
17
+ "full_attention",
18
+ "full_attention",
19
+ "full_attention",
20
+ "full_attention",
21
+ "full_attention",
22
+ "full_attention",
23
+ "full_attention",
24
+ "full_attention",
25
+ "full_attention",
26
+ "full_attention",
27
+ "full_attention",
28
+ "full_attention",
29
+ "full_attention",
30
+ "full_attention",
31
+ "full_attention",
32
+ "full_attention",
33
+ "full_attention",
34
+ "full_attention",
35
+ "full_attention",
36
+ "full_attention",
37
+ "full_attention",
38
+ "full_attention"
39
+ ],
40
+ "max_position_embeddings": 32768,
41
+ "max_window_layers": 24,
42
+ "model_type": "qwen2",
43
+ "neuron": {
44
+ "_serialized_key": "NxDNeuronConfig",
45
+ "batch_size": 2,
46
+ "capacity_factor": null,
47
+ "checkpoint_id": "Qwen/Qwen2.5-0.5B",
48
+ "checkpoint_revision": "060db6499f32faf8b98477b0a26969ef7d8b9987",
49
+ "continuous_batching": true,
50
+ "ep_degree": 1,
51
+ "fused_qkv": false,
52
+ "glu_mlp": true,
53
+ "local_ranks_size": 2,
54
+ "max_batch_size": 2,
55
+ "max_context_length": 1024,
56
+ "max_topk": 256,
57
+ "n_active_tokens": 1024,
58
+ "neuronxcc_version": "2.21.33363.0+82129205",
59
+ "on_device_sampling": false,
60
+ "optimum_neuron_version": "0.4.6.dev3",
61
+ "output_logits": false,
62
+ "pp_degree": 1,
63
+ "prefill_chunk_size": 0,
64
+ "sequence_length": 1024,
65
+ "speculation_length": 0,
66
+ "start_rank_id": 0,
67
+ "target": "trn1",
68
+ "torch_dtype": "bfloat16",
69
+ "tp_degree": 2
70
+ },
71
+ "num_attention_heads": 14,
72
+ "num_hidden_layers": 24,
73
+ "num_key_value_heads": 2,
74
+ "rms_norm_eps": 1e-06,
75
+ "rope_scaling": null,
76
+ "rope_theta": 1000000.0,
77
+ "sliding_window": null,
78
+ "tie_word_embeddings": true,
79
+ "use_cache": true,
80
+ "use_mrope": false,
81
+ "use_sliding_window": false,
82
+ "vocab_size": 151936
83
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/4cb7aff9e2a15c151396f2b684013e39d6739f0dec83e5c9dabbfe9d5fcf77b7/2c6cc66c4776ff9413f0.json ADDED
@@ -0,0 +1,83 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "Qwen/Qwen2.5-0.5B",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "Qwen2ForCausalLM"
7
+ ],
8
+ "attention_dropout": 0.0,
9
+ "dtype": "bfloat16",
10
+ "hidden_act": "silu",
11
+ "hidden_size": 896,
12
+ "initializer_range": 0.02,
13
+ "intermediate_size": 4864,
14
+ "layer_types": [
15
+ "full_attention",
16
+ "full_attention",
17
+ "full_attention",
18
+ "full_attention",
19
+ "full_attention",
20
+ "full_attention",
21
+ "full_attention",
22
+ "full_attention",
23
+ "full_attention",
24
+ "full_attention",
25
+ "full_attention",
26
+ "full_attention",
27
+ "full_attention",
28
+ "full_attention",
29
+ "full_attention",
30
+ "full_attention",
31
+ "full_attention",
32
+ "full_attention",
33
+ "full_attention",
34
+ "full_attention",
35
+ "full_attention",
36
+ "full_attention",
37
+ "full_attention",
38
+ "full_attention"
39
+ ],
40
+ "max_position_embeddings": 32768,
41
+ "max_window_layers": 24,
42
+ "model_type": "qwen2",
43
+ "neuron": {
44
+ "_serialized_key": "NxDNeuronConfig",
45
+ "batch_size": 4,
46
+ "capacity_factor": null,
47
+ "checkpoint_id": "Qwen/Qwen2.5-0.5B",
48
+ "checkpoint_revision": "060db6499f32faf8b98477b0a26969ef7d8b9987",
49
+ "continuous_batching": true,
50
+ "ep_degree": 1,
51
+ "fused_qkv": false,
52
+ "glu_mlp": true,
53
+ "local_ranks_size": 2,
54
+ "max_batch_size": 4,
55
+ "max_context_length": 1024,
56
+ "max_topk": 256,
57
+ "n_active_tokens": 1024,
58
+ "neuronxcc_version": "2.21.33363.0+82129205",
59
+ "on_device_sampling": false,
60
+ "optimum_neuron_version": "0.4.6.dev3",
61
+ "output_logits": false,
62
+ "pp_degree": 1,
63
+ "prefill_chunk_size": 0,
64
+ "sequence_length": 1024,
65
+ "speculation_length": 0,
66
+ "start_rank_id": 0,
67
+ "target": "trn1",
68
+ "torch_dtype": "bfloat16",
69
+ "tp_degree": 2
70
+ },
71
+ "num_attention_heads": 14,
72
+ "num_hidden_layers": 24,
73
+ "num_key_value_heads": 2,
74
+ "rms_norm_eps": 1e-06,
75
+ "rope_scaling": null,
76
+ "rope_theta": 1000000.0,
77
+ "sliding_window": null,
78
+ "tie_word_embeddings": true,
79
+ "use_cache": true,
80
+ "use_mrope": false,
81
+ "use_sliding_window": false,
82
+ "vocab_size": 151936
83
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/60feecaa0c4c075e2f3e46a3f55d9a273f0ddd75a0ecf64e4ae27352e0819506/43be67542ceab3db3486.json ADDED
@@ -0,0 +1,88 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "Qwen/Qwen3-Embedding-0.6B",
4
+ "_task": "feature-extraction",
5
+ "architectures": [
6
+ "Qwen3ForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "dtype": "bfloat16",
11
+ "head_dim": 128,
12
+ "hidden_act": "silu",
13
+ "hidden_size": 1024,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 3072,
16
+ "layer_types": [
17
+ "full_attention",
18
+ "full_attention",
19
+ "full_attention",
20
+ "full_attention",
21
+ "full_attention",
22
+ "full_attention",
23
+ "full_attention",
24
+ "full_attention",
25
+ "full_attention",
26
+ "full_attention",
27
+ "full_attention",
28
+ "full_attention",
29
+ "full_attention",
30
+ "full_attention",
31
+ "full_attention",
32
+ "full_attention",
33
+ "full_attention",
34
+ "full_attention",
35
+ "full_attention",
36
+ "full_attention",
37
+ "full_attention",
38
+ "full_attention",
39
+ "full_attention",
40
+ "full_attention",
41
+ "full_attention",
42
+ "full_attention",
43
+ "full_attention",
44
+ "full_attention"
45
+ ],
46
+ "max_position_embeddings": 32768,
47
+ "max_window_layers": 28,
48
+ "model_type": "qwen3",
49
+ "neuron": {
50
+ "_serialized_key": "NxDNeuronConfig",
51
+ "batch_size": 4,
52
+ "capacity_factor": null,
53
+ "checkpoint_id": "Qwen/Qwen3-Embedding-0.6B",
54
+ "checkpoint_revision": "c54f2e6e80b2d7b7de06f51cec4959f6b3e03418",
55
+ "continuous_batching": false,
56
+ "ep_degree": 1,
57
+ "fused_qkv": true,
58
+ "glu_mlp": true,
59
+ "local_ranks_size": 2,
60
+ "max_batch_size": 4,
61
+ "max_context_length": 8192,
62
+ "max_topk": 256,
63
+ "n_active_tokens": 8192,
64
+ "neuronxcc_version": "2.21.33363.0+82129205",
65
+ "on_device_sampling": false,
66
+ "optimum_neuron_version": "0.4.6.dev3",
67
+ "output_logits": false,
68
+ "pp_degree": 1,
69
+ "prefill_chunk_size": 0,
70
+ "sequence_length": 8192,
71
+ "speculation_length": 0,
72
+ "start_rank_id": 0,
73
+ "target": "trn1",
74
+ "torch_dtype": "bfloat16",
75
+ "tp_degree": 2
76
+ },
77
+ "num_attention_heads": 16,
78
+ "num_hidden_layers": 28,
79
+ "num_key_value_heads": 8,
80
+ "rms_norm_eps": 1e-06,
81
+ "rope_scaling": null,
82
+ "rope_theta": 1000000,
83
+ "sliding_window": null,
84
+ "tie_word_embeddings": true,
85
+ "use_cache": true,
86
+ "use_sliding_window": false,
87
+ "vocab_size": 151669
88
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/60feecaa0c4c075e2f3e46a3f55d9a273f0ddd75a0ecf64e4ae27352e0819506/d57c7c9107ac77a642d4.json ADDED
@@ -0,0 +1,88 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "Qwen/Qwen3-Embedding-0.6B",
4
+ "_task": "feature-extraction",
5
+ "architectures": [
6
+ "Qwen3ForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "dtype": "bfloat16",
11
+ "head_dim": 128,
12
+ "hidden_act": "silu",
13
+ "hidden_size": 1024,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 3072,
16
+ "layer_types": [
17
+ "full_attention",
18
+ "full_attention",
19
+ "full_attention",
20
+ "full_attention",
21
+ "full_attention",
22
+ "full_attention",
23
+ "full_attention",
24
+ "full_attention",
25
+ "full_attention",
26
+ "full_attention",
27
+ "full_attention",
28
+ "full_attention",
29
+ "full_attention",
30
+ "full_attention",
31
+ "full_attention",
32
+ "full_attention",
33
+ "full_attention",
34
+ "full_attention",
35
+ "full_attention",
36
+ "full_attention",
37
+ "full_attention",
38
+ "full_attention",
39
+ "full_attention",
40
+ "full_attention",
41
+ "full_attention",
42
+ "full_attention",
43
+ "full_attention",
44
+ "full_attention"
45
+ ],
46
+ "max_position_embeddings": 32768,
47
+ "max_window_layers": 28,
48
+ "model_type": "qwen3",
49
+ "neuron": {
50
+ "_serialized_key": "NxDNeuronConfig",
51
+ "batch_size": 6,
52
+ "capacity_factor": null,
53
+ "checkpoint_id": "Qwen/Qwen3-Embedding-0.6B",
54
+ "checkpoint_revision": "c54f2e6e80b2d7b7de06f51cec4959f6b3e03418",
55
+ "continuous_batching": false,
56
+ "ep_degree": 1,
57
+ "fused_qkv": true,
58
+ "glu_mlp": true,
59
+ "local_ranks_size": 2,
60
+ "max_batch_size": 6,
61
+ "max_context_length": 8192,
62
+ "max_topk": 256,
63
+ "n_active_tokens": 8192,
64
+ "neuronxcc_version": "2.21.33363.0+82129205",
65
+ "on_device_sampling": false,
66
+ "optimum_neuron_version": "0.4.6.dev3",
67
+ "output_logits": false,
68
+ "pp_degree": 1,
69
+ "prefill_chunk_size": 0,
70
+ "sequence_length": 8192,
71
+ "speculation_length": 0,
72
+ "start_rank_id": 0,
73
+ "target": "trn1",
74
+ "torch_dtype": "bfloat16",
75
+ "tp_degree": 2
76
+ },
77
+ "num_attention_heads": 16,
78
+ "num_hidden_layers": 28,
79
+ "num_key_value_heads": 8,
80
+ "rms_norm_eps": 1e-06,
81
+ "rope_scaling": null,
82
+ "rope_theta": 1000000,
83
+ "sliding_window": null,
84
+ "tie_word_embeddings": true,
85
+ "use_cache": true,
86
+ "use_sliding_window": false,
87
+ "vocab_size": 151669
88
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/6454afdf3e9d66c7226c13a575b718845c25e53b0699600ba2bb4f883e9d841b/4b6c7416e39455d1ad2e.json ADDED
@@ -0,0 +1,63 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "llamafactory/tiny-random-Llama-3",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "LlamaForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "dtype": "float16",
11
+ "head_dim": 4,
12
+ "hidden_act": "silu",
13
+ "hidden_size": 16,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 64,
16
+ "max_position_embeddings": 131072,
17
+ "mlp_bias": false,
18
+ "model_type": "llama",
19
+ "neuron": {
20
+ "_serialized_key": "NxDNeuronConfig",
21
+ "batch_size": 1,
22
+ "capacity_factor": null,
23
+ "checkpoint_id": "llamafactory/tiny-random-Llama-3",
24
+ "checkpoint_revision": "bf2a2e3bf199ad2ee96f02a3c00246c608db22a8",
25
+ "continuous_batching": false,
26
+ "ep_degree": 1,
27
+ "fused_qkv": true,
28
+ "glu_mlp": true,
29
+ "local_ranks_size": 2,
30
+ "max_batch_size": 1,
31
+ "max_context_length": 512,
32
+ "max_topk": 256,
33
+ "n_active_tokens": 512,
34
+ "neuronxcc_version": "2.21.33363.0+82129205",
35
+ "on_device_sampling": true,
36
+ "optimum_neuron_version": "0.4.6.dev3",
37
+ "output_logits": false,
38
+ "pp_degree": 1,
39
+ "prefill_chunk_size": 0,
40
+ "sequence_length": 512,
41
+ "speculation_length": 0,
42
+ "start_rank_id": 0,
43
+ "target": "trn1",
44
+ "torch_dtype": "float16",
45
+ "tp_degree": 2
46
+ },
47
+ "num_attention_heads": 4,
48
+ "num_hidden_layers": 2,
49
+ "num_key_value_heads": 4,
50
+ "pretraining_tp": 1,
51
+ "rms_norm_eps": 1e-05,
52
+ "rope_scaling": {
53
+ "factor": 8.0,
54
+ "high_freq_factor": 4.0,
55
+ "low_freq_factor": 1.0,
56
+ "original_max_position_embeddings": 8192,
57
+ "rope_type": "llama3"
58
+ },
59
+ "rope_theta": 500000.0,
60
+ "tie_word_embeddings": false,
61
+ "use_cache": true,
62
+ "vocab_size": 128256
63
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/6454afdf3e9d66c7226c13a575b718845c25e53b0699600ba2bb4f883e9d841b/5d5179e2ac0b724548b0.json ADDED
@@ -0,0 +1,63 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "llamafactory/tiny-random-Llama-3",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "LlamaForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "dtype": "float16",
11
+ "head_dim": 4,
12
+ "hidden_act": "silu",
13
+ "hidden_size": 16,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 64,
16
+ "max_position_embeddings": 131072,
17
+ "mlp_bias": false,
18
+ "model_type": "llama",
19
+ "neuron": {
20
+ "_serialized_key": "NxDNeuronConfig",
21
+ "batch_size": 1,
22
+ "capacity_factor": null,
23
+ "checkpoint_id": "llamafactory/tiny-random-Llama-3",
24
+ "checkpoint_revision": "bf2a2e3bf199ad2ee96f02a3c00246c608db22a8",
25
+ "continuous_batching": false,
26
+ "ep_degree": 1,
27
+ "fused_qkv": true,
28
+ "glu_mlp": true,
29
+ "local_ranks_size": 2,
30
+ "max_batch_size": 1,
31
+ "max_context_length": 1024,
32
+ "max_topk": 256,
33
+ "n_active_tokens": 1024,
34
+ "neuronxcc_version": "2.21.33363.0+82129205",
35
+ "on_device_sampling": true,
36
+ "optimum_neuron_version": "0.4.6.dev3",
37
+ "output_logits": false,
38
+ "pp_degree": 1,
39
+ "prefill_chunk_size": 0,
40
+ "sequence_length": 1024,
41
+ "speculation_length": 0,
42
+ "start_rank_id": 0,
43
+ "target": "trn1",
44
+ "torch_dtype": "float16",
45
+ "tp_degree": 2
46
+ },
47
+ "num_attention_heads": 4,
48
+ "num_hidden_layers": 2,
49
+ "num_key_value_heads": 4,
50
+ "pretraining_tp": 1,
51
+ "rms_norm_eps": 1e-05,
52
+ "rope_scaling": {
53
+ "factor": 8.0,
54
+ "high_freq_factor": 4.0,
55
+ "low_freq_factor": 1.0,
56
+ "original_max_position_embeddings": 8192,
57
+ "rope_type": "llama3"
58
+ },
59
+ "rope_theta": 500000.0,
60
+ "tie_word_embeddings": false,
61
+ "use_cache": true,
62
+ "vocab_size": 128256
63
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/6454afdf3e9d66c7226c13a575b718845c25e53b0699600ba2bb4f883e9d841b/8f8e8c8f417393bfd04a.json ADDED
@@ -0,0 +1,63 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "llamafactory/tiny-random-Llama-3",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "LlamaForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "dtype": "float16",
11
+ "head_dim": 4,
12
+ "hidden_act": "silu",
13
+ "hidden_size": 16,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 64,
16
+ "max_position_embeddings": 131072,
17
+ "mlp_bias": false,
18
+ "model_type": "llama",
19
+ "neuron": {
20
+ "_serialized_key": "NxDNeuronConfig",
21
+ "batch_size": 2,
22
+ "capacity_factor": null,
23
+ "checkpoint_id": "llamafactory/tiny-random-Llama-3",
24
+ "checkpoint_revision": "bf2a2e3bf199ad2ee96f02a3c00246c608db22a8",
25
+ "continuous_batching": true,
26
+ "ep_degree": 1,
27
+ "fused_qkv": true,
28
+ "glu_mlp": true,
29
+ "local_ranks_size": 2,
30
+ "max_batch_size": 2,
31
+ "max_context_length": 128,
32
+ "max_topk": 256,
33
+ "n_active_tokens": 128,
34
+ "neuronxcc_version": "2.21.33363.0+82129205",
35
+ "on_device_sampling": true,
36
+ "optimum_neuron_version": "0.4.6.dev3",
37
+ "output_logits": false,
38
+ "pp_degree": 1,
39
+ "prefill_chunk_size": 0,
40
+ "sequence_length": 128,
41
+ "speculation_length": 0,
42
+ "start_rank_id": 0,
43
+ "target": "trn2",
44
+ "torch_dtype": "float16",
45
+ "tp_degree": 2
46
+ },
47
+ "num_attention_heads": 4,
48
+ "num_hidden_layers": 2,
49
+ "num_key_value_heads": 4,
50
+ "pretraining_tp": 1,
51
+ "rms_norm_eps": 1e-05,
52
+ "rope_scaling": {
53
+ "factor": 8.0,
54
+ "high_freq_factor": 4.0,
55
+ "low_freq_factor": 1.0,
56
+ "original_max_position_embeddings": 8192,
57
+ "rope_type": "llama3"
58
+ },
59
+ "rope_theta": 500000.0,
60
+ "tie_word_embeddings": false,
61
+ "use_cache": true,
62
+ "vocab_size": 128256
63
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/73707b485eab9008c7aba7f5dad0ce2384ac685318d5f888c12fa0d81ed90b19/7536b906773aff032659.json ADDED
@@ -0,0 +1,135 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "HuggingFaceTB/SmolLM3-3B",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "SmolLM3ForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "dtype": "bfloat16",
11
+ "hidden_act": "silu",
12
+ "hidden_size": 2048,
13
+ "initializer_range": 0.02,
14
+ "intermediate_size": 11008,
15
+ "layer_types": [
16
+ "full_attention",
17
+ "full_attention",
18
+ "full_attention",
19
+ "full_attention",
20
+ "full_attention",
21
+ "full_attention",
22
+ "full_attention",
23
+ "full_attention",
24
+ "full_attention",
25
+ "full_attention",
26
+ "full_attention",
27
+ "full_attention",
28
+ "full_attention",
29
+ "full_attention",
30
+ "full_attention",
31
+ "full_attention",
32
+ "full_attention",
33
+ "full_attention",
34
+ "full_attention",
35
+ "full_attention",
36
+ "full_attention",
37
+ "full_attention",
38
+ "full_attention",
39
+ "full_attention",
40
+ "full_attention",
41
+ "full_attention",
42
+ "full_attention",
43
+ "full_attention",
44
+ "full_attention",
45
+ "full_attention",
46
+ "full_attention",
47
+ "full_attention",
48
+ "full_attention",
49
+ "full_attention",
50
+ "full_attention",
51
+ "full_attention"
52
+ ],
53
+ "max_position_embeddings": 65536,
54
+ "max_window_layers": 28,
55
+ "mlp_bias": false,
56
+ "model_type": "smollm3",
57
+ "neuron": {
58
+ "_serialized_key": "NxDNeuronConfig",
59
+ "batch_size": 1,
60
+ "capacity_factor": null,
61
+ "checkpoint_id": "HuggingFaceTB/SmolLM3-3B",
62
+ "checkpoint_revision": "a07cc9a04f16550a088caea529712d1d335b0ac1",
63
+ "continuous_batching": false,
64
+ "ep_degree": 1,
65
+ "fused_qkv": true,
66
+ "glu_mlp": true,
67
+ "local_ranks_size": 2,
68
+ "max_batch_size": 1,
69
+ "max_context_length": 8192,
70
+ "max_topk": 256,
71
+ "n_active_tokens": 8192,
72
+ "neuronxcc_version": "2.21.33363.0+82129205",
73
+ "on_device_sampling": true,
74
+ "optimum_neuron_version": "0.4.6.dev3",
75
+ "output_logits": false,
76
+ "pp_degree": 1,
77
+ "prefill_chunk_size": 1024,
78
+ "sequence_length": 8192,
79
+ "speculation_length": 0,
80
+ "start_rank_id": 0,
81
+ "target": "trn1",
82
+ "torch_dtype": "bfloat16",
83
+ "tp_degree": 2
84
+ },
85
+ "no_rope_layer_interval": 4,
86
+ "no_rope_layers": [
87
+ 1,
88
+ 1,
89
+ 1,
90
+ 0,
91
+ 1,
92
+ 1,
93
+ 1,
94
+ 0,
95
+ 1,
96
+ 1,
97
+ 1,
98
+ 0,
99
+ 1,
100
+ 1,
101
+ 1,
102
+ 0,
103
+ 1,
104
+ 1,
105
+ 1,
106
+ 0,
107
+ 1,
108
+ 1,
109
+ 1,
110
+ 0,
111
+ 1,
112
+ 1,
113
+ 1,
114
+ 0,
115
+ 1,
116
+ 1,
117
+ 1,
118
+ 0,
119
+ 1,
120
+ 1,
121
+ 1,
122
+ 0
123
+ ],
124
+ "num_attention_heads": 16,
125
+ "num_hidden_layers": 36,
126
+ "num_key_value_heads": 4,
127
+ "pretraining_tp": 2,
128
+ "rms_norm_eps": 1e-06,
129
+ "rope_scaling": null,
130
+ "rope_theta": 5000000.0,
131
+ "sliding_window": null,
132
+ "use_cache": false,
133
+ "use_sliding_window": false,
134
+ "vocab_size": 128256
135
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/73707b485eab9008c7aba7f5dad0ce2384ac685318d5f888c12fa0d81ed90b19/87f1aaa5ae1a31bb7adc.json ADDED
@@ -0,0 +1,135 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "HuggingFaceTB/SmolLM3-3B",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "SmolLM3ForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "dtype": "bfloat16",
11
+ "hidden_act": "silu",
12
+ "hidden_size": 2048,
13
+ "initializer_range": 0.02,
14
+ "intermediate_size": 11008,
15
+ "layer_types": [
16
+ "full_attention",
17
+ "full_attention",
18
+ "full_attention",
19
+ "full_attention",
20
+ "full_attention",
21
+ "full_attention",
22
+ "full_attention",
23
+ "full_attention",
24
+ "full_attention",
25
+ "full_attention",
26
+ "full_attention",
27
+ "full_attention",
28
+ "full_attention",
29
+ "full_attention",
30
+ "full_attention",
31
+ "full_attention",
32
+ "full_attention",
33
+ "full_attention",
34
+ "full_attention",
35
+ "full_attention",
36
+ "full_attention",
37
+ "full_attention",
38
+ "full_attention",
39
+ "full_attention",
40
+ "full_attention",
41
+ "full_attention",
42
+ "full_attention",
43
+ "full_attention",
44
+ "full_attention",
45
+ "full_attention",
46
+ "full_attention",
47
+ "full_attention",
48
+ "full_attention",
49
+ "full_attention",
50
+ "full_attention",
51
+ "full_attention"
52
+ ],
53
+ "max_position_embeddings": 65536,
54
+ "max_window_layers": 28,
55
+ "mlp_bias": false,
56
+ "model_type": "smollm3",
57
+ "neuron": {
58
+ "_serialized_key": "NxDNeuronConfig",
59
+ "batch_size": 1,
60
+ "capacity_factor": null,
61
+ "checkpoint_id": "HuggingFaceTB/SmolLM3-3B",
62
+ "checkpoint_revision": "a07cc9a04f16550a088caea529712d1d335b0ac1",
63
+ "continuous_batching": false,
64
+ "ep_degree": 1,
65
+ "fused_qkv": true,
66
+ "glu_mlp": true,
67
+ "local_ranks_size": 2,
68
+ "max_batch_size": 1,
69
+ "max_context_length": 1024,
70
+ "max_topk": 256,
71
+ "n_active_tokens": 1024,
72
+ "neuronxcc_version": "2.21.33363.0+82129205",
73
+ "on_device_sampling": true,
74
+ "optimum_neuron_version": "0.4.6.dev3",
75
+ "output_logits": false,
76
+ "pp_degree": 1,
77
+ "prefill_chunk_size": 0,
78
+ "sequence_length": 1024,
79
+ "speculation_length": 0,
80
+ "start_rank_id": 0,
81
+ "target": "trn1",
82
+ "torch_dtype": "bfloat16",
83
+ "tp_degree": 2
84
+ },
85
+ "no_rope_layer_interval": 4,
86
+ "no_rope_layers": [
87
+ 1,
88
+ 1,
89
+ 1,
90
+ 0,
91
+ 1,
92
+ 1,
93
+ 1,
94
+ 0,
95
+ 1,
96
+ 1,
97
+ 1,
98
+ 0,
99
+ 1,
100
+ 1,
101
+ 1,
102
+ 0,
103
+ 1,
104
+ 1,
105
+ 1,
106
+ 0,
107
+ 1,
108
+ 1,
109
+ 1,
110
+ 0,
111
+ 1,
112
+ 1,
113
+ 1,
114
+ 0,
115
+ 1,
116
+ 1,
117
+ 1,
118
+ 0,
119
+ 1,
120
+ 1,
121
+ 1,
122
+ 0
123
+ ],
124
+ "num_attention_heads": 16,
125
+ "num_hidden_layers": 36,
126
+ "num_key_value_heads": 4,
127
+ "pretraining_tp": 2,
128
+ "rms_norm_eps": 1e-06,
129
+ "rope_scaling": null,
130
+ "rope_theta": 5000000.0,
131
+ "sliding_window": null,
132
+ "use_cache": false,
133
+ "use_sliding_window": false,
134
+ "vocab_size": 128256
135
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/73707b485eab9008c7aba7f5dad0ce2384ac685318d5f888c12fa0d81ed90b19/b2c32dabac1b1c034891.json ADDED
@@ -0,0 +1,135 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "HuggingFaceTB/SmolLM3-3B",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "SmolLM3ForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "dtype": "bfloat16",
11
+ "hidden_act": "silu",
12
+ "hidden_size": 2048,
13
+ "initializer_range": 0.02,
14
+ "intermediate_size": 11008,
15
+ "layer_types": [
16
+ "full_attention",
17
+ "full_attention",
18
+ "full_attention",
19
+ "full_attention",
20
+ "full_attention",
21
+ "full_attention",
22
+ "full_attention",
23
+ "full_attention",
24
+ "full_attention",
25
+ "full_attention",
26
+ "full_attention",
27
+ "full_attention",
28
+ "full_attention",
29
+ "full_attention",
30
+ "full_attention",
31
+ "full_attention",
32
+ "full_attention",
33
+ "full_attention",
34
+ "full_attention",
35
+ "full_attention",
36
+ "full_attention",
37
+ "full_attention",
38
+ "full_attention",
39
+ "full_attention",
40
+ "full_attention",
41
+ "full_attention",
42
+ "full_attention",
43
+ "full_attention",
44
+ "full_attention",
45
+ "full_attention",
46
+ "full_attention",
47
+ "full_attention",
48
+ "full_attention",
49
+ "full_attention",
50
+ "full_attention",
51
+ "full_attention"
52
+ ],
53
+ "max_position_embeddings": 65536,
54
+ "max_window_layers": 28,
55
+ "mlp_bias": false,
56
+ "model_type": "smollm3",
57
+ "neuron": {
58
+ "_serialized_key": "NxDNeuronConfig",
59
+ "batch_size": 4,
60
+ "capacity_factor": null,
61
+ "checkpoint_id": "HuggingFaceTB/SmolLM3-3B",
62
+ "checkpoint_revision": "a07cc9a04f16550a088caea529712d1d335b0ac1",
63
+ "continuous_batching": true,
64
+ "ep_degree": 1,
65
+ "fused_qkv": true,
66
+ "glu_mlp": true,
67
+ "local_ranks_size": 2,
68
+ "max_batch_size": 4,
69
+ "max_context_length": 1024,
70
+ "max_topk": 256,
71
+ "n_active_tokens": 1024,
72
+ "neuronxcc_version": "2.21.33363.0+82129205",
73
+ "on_device_sampling": true,
74
+ "optimum_neuron_version": "0.4.6.dev3",
75
+ "output_logits": false,
76
+ "pp_degree": 1,
77
+ "prefill_chunk_size": 0,
78
+ "sequence_length": 1024,
79
+ "speculation_length": 0,
80
+ "start_rank_id": 0,
81
+ "target": "trn1",
82
+ "torch_dtype": "bfloat16",
83
+ "tp_degree": 2
84
+ },
85
+ "no_rope_layer_interval": 4,
86
+ "no_rope_layers": [
87
+ 1,
88
+ 1,
89
+ 1,
90
+ 0,
91
+ 1,
92
+ 1,
93
+ 1,
94
+ 0,
95
+ 1,
96
+ 1,
97
+ 1,
98
+ 0,
99
+ 1,
100
+ 1,
101
+ 1,
102
+ 0,
103
+ 1,
104
+ 1,
105
+ 1,
106
+ 0,
107
+ 1,
108
+ 1,
109
+ 1,
110
+ 0,
111
+ 1,
112
+ 1,
113
+ 1,
114
+ 0,
115
+ 1,
116
+ 1,
117
+ 1,
118
+ 0,
119
+ 1,
120
+ 1,
121
+ 1,
122
+ 0
123
+ ],
124
+ "num_attention_heads": 16,
125
+ "num_hidden_layers": 36,
126
+ "num_key_value_heads": 4,
127
+ "pretraining_tp": 2,
128
+ "rms_norm_eps": 1e-06,
129
+ "rope_scaling": null,
130
+ "rope_theta": 5000000.0,
131
+ "sliding_window": null,
132
+ "use_cache": false,
133
+ "use_sliding_window": false,
134
+ "vocab_size": 128256
135
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/7518518c7e077820070186deda960d8cc49db068cdf0ac70664098fa2b6b698c/7e378b918d8aa52c31f9.json ADDED
@@ -0,0 +1,65 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "yujiepan/qwen2.5-128k-tiny-random",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "Qwen2ForCausalLM"
7
+ ],
8
+ "attention_dropout": 0.0,
9
+ "dtype": "bfloat16",
10
+ "hidden_act": "silu",
11
+ "hidden_size": 8,
12
+ "initializer_range": 0.02,
13
+ "intermediate_size": 16,
14
+ "layer_types": [
15
+ "full_attention",
16
+ "full_attention"
17
+ ],
18
+ "max_position_embeddings": 32768,
19
+ "max_window_layers": 1,
20
+ "model_type": "qwen2",
21
+ "neuron": {
22
+ "_serialized_key": "NxDNeuronConfig",
23
+ "batch_size": 1,
24
+ "capacity_factor": null,
25
+ "checkpoint_id": "yujiepan/qwen2.5-128k-tiny-random",
26
+ "checkpoint_revision": "c8296d4ca3f87782876d2382fbb6481d1beb8ef0",
27
+ "continuous_batching": false,
28
+ "ep_degree": 1,
29
+ "fused_qkv": false,
30
+ "glu_mlp": true,
31
+ "local_ranks_size": 2,
32
+ "max_batch_size": 1,
33
+ "max_context_length": 1024,
34
+ "max_topk": 256,
35
+ "n_active_tokens": 1024,
36
+ "neuronxcc_version": "2.21.33363.0+82129205",
37
+ "on_device_sampling": true,
38
+ "optimum_neuron_version": "0.4.6.dev3",
39
+ "output_logits": false,
40
+ "pp_degree": 1,
41
+ "prefill_chunk_size": 0,
42
+ "sequence_length": 1024,
43
+ "speculation_length": 0,
44
+ "start_rank_id": 0,
45
+ "target": "trn1",
46
+ "torch_dtype": "bfloat16",
47
+ "tp_degree": 2
48
+ },
49
+ "num_attention_heads": 4,
50
+ "num_hidden_layers": 2,
51
+ "num_key_value_heads": 2,
52
+ "rms_norm_eps": 1e-06,
53
+ "rope_scaling": {
54
+ "factor": 4.0,
55
+ "original_max_position_embeddings": 32768,
56
+ "rope_type": "yarn",
57
+ "type": "yarn"
58
+ },
59
+ "rope_theta": 1000000.0,
60
+ "sliding_window": null,
61
+ "tie_word_embeddings": false,
62
+ "use_cache": true,
63
+ "use_sliding_window": false,
64
+ "vocab_size": 152064
65
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/920f44ce6d3e004d1ce547ae06644f7be262180644b04573153aa15d98742edc/ad0f2997db25e16e085b.json ADDED
@@ -0,0 +1,66 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "optimum-internal-testing/tiny-random-qwen3_moe",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "Qwen3MoeForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "decoder_sparse_step": 2,
11
+ "dtype": "float32",
12
+ "head_dim": 32,
13
+ "hidden_act": "silu",
14
+ "hidden_size": 64,
15
+ "initializer_range": 0.02,
16
+ "intermediate_size": 128,
17
+ "max_position_embeddings": 40960,
18
+ "max_window_layers": 1,
19
+ "mlp_only_layers": [],
20
+ "model_type": "qwen3_moe",
21
+ "moe_intermediate_size": 128,
22
+ "neuron": {
23
+ "_serialized_key": "NxDNeuronConfig",
24
+ "batch_size": 1,
25
+ "capacity_factor": null,
26
+ "checkpoint_id": "optimum-internal-testing/tiny-random-qwen3_moe",
27
+ "checkpoint_revision": "e0230be2839556b44b7400a233c73c74b4abb7af",
28
+ "continuous_batching": false,
29
+ "ep_degree": 1,
30
+ "fused_qkv": false,
31
+ "glu_mlp": true,
32
+ "local_ranks_size": 2,
33
+ "max_batch_size": 1,
34
+ "max_context_length": 1024,
35
+ "max_topk": 256,
36
+ "n_active_tokens": 1024,
37
+ "neuronxcc_version": "2.21.33363.0+82129205",
38
+ "on_device_sampling": true,
39
+ "optimum_neuron_version": "0.4.6.dev3",
40
+ "output_logits": false,
41
+ "pp_degree": 1,
42
+ "prefill_chunk_size": 0,
43
+ "sequence_length": 1024,
44
+ "speculation_length": 0,
45
+ "start_rank_id": 0,
46
+ "target": "trn1",
47
+ "torch_dtype": "float32",
48
+ "tp_degree": 2
49
+ },
50
+ "norm_topk_prob": true,
51
+ "num_attention_heads": 2,
52
+ "num_experts": 8,
53
+ "num_experts_per_tok": 2,
54
+ "num_hidden_layers": 2,
55
+ "num_key_value_heads": 1,
56
+ "output_router_logits": false,
57
+ "rms_norm_eps": 1e-06,
58
+ "rope_scaling": null,
59
+ "rope_theta": 1000000.0,
60
+ "router_aux_loss_coef": 0.001,
61
+ "sliding_window": null,
62
+ "tie_word_embeddings": true,
63
+ "use_cache": true,
64
+ "use_sliding_window": false,
65
+ "vocab_size": 151936
66
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/929b02754a13cbfdf657d863c3fc6f3bce672879bc6ae48ab45be21e881e9ec2/947cd8633065b84ee800.json ADDED
@@ -0,0 +1,88 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "Qwen/Qwen3-0.6B",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "Qwen3ForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "dtype": "bfloat16",
11
+ "head_dim": 128,
12
+ "hidden_act": "silu",
13
+ "hidden_size": 1024,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 3072,
16
+ "layer_types": [
17
+ "full_attention",
18
+ "full_attention",
19
+ "full_attention",
20
+ "full_attention",
21
+ "full_attention",
22
+ "full_attention",
23
+ "full_attention",
24
+ "full_attention",
25
+ "full_attention",
26
+ "full_attention",
27
+ "full_attention",
28
+ "full_attention",
29
+ "full_attention",
30
+ "full_attention",
31
+ "full_attention",
32
+ "full_attention",
33
+ "full_attention",
34
+ "full_attention",
35
+ "full_attention",
36
+ "full_attention",
37
+ "full_attention",
38
+ "full_attention",
39
+ "full_attention",
40
+ "full_attention",
41
+ "full_attention",
42
+ "full_attention",
43
+ "full_attention",
44
+ "full_attention"
45
+ ],
46
+ "max_position_embeddings": 40960,
47
+ "max_window_layers": 28,
48
+ "model_type": "qwen3",
49
+ "neuron": {
50
+ "_serialized_key": "NxDNeuronConfig",
51
+ "batch_size": 1,
52
+ "capacity_factor": null,
53
+ "checkpoint_id": "Qwen/Qwen3-0.6B",
54
+ "checkpoint_revision": "c1899de289a04d12100db370d81485cdf75e47ca",
55
+ "continuous_batching": false,
56
+ "ep_degree": 1,
57
+ "fused_qkv": true,
58
+ "glu_mlp": true,
59
+ "local_ranks_size": 2,
60
+ "max_batch_size": 1,
61
+ "max_context_length": 8192,
62
+ "max_topk": 256,
63
+ "n_active_tokens": 8192,
64
+ "neuronxcc_version": "2.21.33363.0+82129205",
65
+ "on_device_sampling": true,
66
+ "optimum_neuron_version": "0.4.6.dev3",
67
+ "output_logits": false,
68
+ "pp_degree": 1,
69
+ "prefill_chunk_size": 1024,
70
+ "sequence_length": 8192,
71
+ "speculation_length": 0,
72
+ "start_rank_id": 0,
73
+ "target": "trn1",
74
+ "torch_dtype": "bfloat16",
75
+ "tp_degree": 2
76
+ },
77
+ "num_attention_heads": 16,
78
+ "num_hidden_layers": 28,
79
+ "num_key_value_heads": 8,
80
+ "rms_norm_eps": 1e-06,
81
+ "rope_scaling": null,
82
+ "rope_theta": 1000000,
83
+ "sliding_window": null,
84
+ "tie_word_embeddings": true,
85
+ "use_cache": true,
86
+ "use_sliding_window": false,
87
+ "vocab_size": 151936
88
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/929b02754a13cbfdf657d863c3fc6f3bce672879bc6ae48ab45be21e881e9ec2/c3b960b90a5e3c592c04.json ADDED
@@ -0,0 +1,88 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "Qwen/Qwen3-0.6B",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "Qwen3ForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "dtype": "bfloat16",
11
+ "head_dim": 128,
12
+ "hidden_act": "silu",
13
+ "hidden_size": 1024,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 3072,
16
+ "layer_types": [
17
+ "full_attention",
18
+ "full_attention",
19
+ "full_attention",
20
+ "full_attention",
21
+ "full_attention",
22
+ "full_attention",
23
+ "full_attention",
24
+ "full_attention",
25
+ "full_attention",
26
+ "full_attention",
27
+ "full_attention",
28
+ "full_attention",
29
+ "full_attention",
30
+ "full_attention",
31
+ "full_attention",
32
+ "full_attention",
33
+ "full_attention",
34
+ "full_attention",
35
+ "full_attention",
36
+ "full_attention",
37
+ "full_attention",
38
+ "full_attention",
39
+ "full_attention",
40
+ "full_attention",
41
+ "full_attention",
42
+ "full_attention",
43
+ "full_attention",
44
+ "full_attention"
45
+ ],
46
+ "max_position_embeddings": 40960,
47
+ "max_window_layers": 28,
48
+ "model_type": "qwen3",
49
+ "neuron": {
50
+ "_serialized_key": "NxDNeuronConfig",
51
+ "batch_size": 4,
52
+ "capacity_factor": null,
53
+ "checkpoint_id": "Qwen/Qwen3-0.6B",
54
+ "checkpoint_revision": "c1899de289a04d12100db370d81485cdf75e47ca",
55
+ "continuous_batching": true,
56
+ "ep_degree": 1,
57
+ "fused_qkv": true,
58
+ "glu_mlp": true,
59
+ "local_ranks_size": 2,
60
+ "max_batch_size": 4,
61
+ "max_context_length": 1024,
62
+ "max_topk": 256,
63
+ "n_active_tokens": 1024,
64
+ "neuronxcc_version": "2.21.33363.0+82129205",
65
+ "on_device_sampling": false,
66
+ "optimum_neuron_version": "0.4.6.dev3",
67
+ "output_logits": false,
68
+ "pp_degree": 1,
69
+ "prefill_chunk_size": 0,
70
+ "sequence_length": 1024,
71
+ "speculation_length": 0,
72
+ "start_rank_id": 0,
73
+ "target": "trn1",
74
+ "torch_dtype": "bfloat16",
75
+ "tp_degree": 2
76
+ },
77
+ "num_attention_heads": 16,
78
+ "num_hidden_layers": 28,
79
+ "num_key_value_heads": 8,
80
+ "rms_norm_eps": 1e-06,
81
+ "rope_scaling": null,
82
+ "rope_theta": 1000000,
83
+ "sliding_window": null,
84
+ "tie_word_embeddings": true,
85
+ "use_cache": true,
86
+ "use_sliding_window": false,
87
+ "vocab_size": 151936
88
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/9c24a2080ec26e9c74d5bd90866f74131aa4d5c975415b96f5e6600f081d5647/d244406bccd80288e991.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "tiny-random/llama-4",
4
+ "_task": "text-generation",
5
+ "attention_bias": false,
6
+ "attention_chunk_size": 128,
7
+ "attention_dropout": 0.0,
8
+ "attn_scale": 0.1,
9
+ "attn_temperature_tuning": 4,
10
+ "cache_implementation": "hybrid",
11
+ "dtype": "bfloat16",
12
+ "floor_scale": 8192,
13
+ "for_llm_compressor": false,
14
+ "head_dim": 32,
15
+ "hidden_act": "silu",
16
+ "hidden_size": 32,
17
+ "initializer_range": 0.02,
18
+ "interleave_moe_layer_step": 2,
19
+ "intermediate_size": 64,
20
+ "intermediate_size_mlp": 128,
21
+ "layer_types": [
22
+ "chunked_attention",
23
+ "chunked_attention",
24
+ "chunked_attention",
25
+ "full_attention"
26
+ ],
27
+ "max_position_embeddings": 1048576,
28
+ "model_type": "llama4_text",
29
+ "moe_layers": [
30
+ 1,
31
+ 3
32
+ ],
33
+ "neuron": {
34
+ "_serialized_key": "NxDNeuronConfig",
35
+ "batch_size": 1,
36
+ "capacity_factor": null,
37
+ "checkpoint_id": "tiny-random/llama-4",
38
+ "checkpoint_revision": "9e716f5d4d1ffe0a44a15f46f4a12b840439aba4",
39
+ "continuous_batching": false,
40
+ "ep_degree": 1,
41
+ "fused_qkv": false,
42
+ "glu_mlp": true,
43
+ "local_ranks_size": 2,
44
+ "max_batch_size": 1,
45
+ "max_context_length": 1024,
46
+ "max_topk": 256,
47
+ "n_active_tokens": 1024,
48
+ "neuronxcc_version": "2.21.33363.0+82129205",
49
+ "on_device_sampling": true,
50
+ "optimum_neuron_version": "0.4.6.dev3",
51
+ "output_logits": false,
52
+ "pp_degree": 1,
53
+ "prefill_chunk_size": 0,
54
+ "sequence_length": 1024,
55
+ "speculation_length": 0,
56
+ "start_rank_id": 0,
57
+ "target": "trn1",
58
+ "torch_dtype": "bfloat16",
59
+ "tp_degree": 2
60
+ },
61
+ "no_rope_layers": [
62
+ 1,
63
+ 1,
64
+ 1,
65
+ 0
66
+ ],
67
+ "num_attention_heads": 1,
68
+ "num_experts_per_tok": 1,
69
+ "num_hidden_layers": 4,
70
+ "num_key_value_heads": 1,
71
+ "num_local_experts": 8,
72
+ "output_router_logits": false,
73
+ "rms_norm_eps": 1e-05,
74
+ "rope_scaling": null,
75
+ "rope_theta": 500000.0,
76
+ "router_aux_loss_coef": 0.001,
77
+ "router_jitter_noise": 0.0,
78
+ "tie_word_embeddings": true,
79
+ "use_cache": true,
80
+ "use_qk_norm": true,
81
+ "vocab_size": 202048
82
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/cf6b9a360dcf294104671106bae2adbd9fd291823bb60a351883163684073231/0fa4ed4626958e306002.json ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "unsloth/Llama-3.2-1B-Instruct",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "LlamaForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "dtype": "bfloat16",
11
+ "head_dim": 64,
12
+ "hidden_act": "silu",
13
+ "hidden_size": 2048,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 8192,
16
+ "max_position_embeddings": 131072,
17
+ "mlp_bias": false,
18
+ "model_type": "llama",
19
+ "neuron": {
20
+ "_serialized_key": "NxDNeuronConfig",
21
+ "batch_size": 1,
22
+ "capacity_factor": null,
23
+ "checkpoint_id": "unsloth/Llama-3.2-1B-Instruct",
24
+ "checkpoint_revision": "5a8abab4a5d6f164389b1079fb721cfab8d7126c",
25
+ "continuous_batching": false,
26
+ "ep_degree": 1,
27
+ "fused_qkv": true,
28
+ "glu_mlp": true,
29
+ "local_ranks_size": 2,
30
+ "max_batch_size": 1,
31
+ "max_context_length": 8192,
32
+ "max_topk": 256,
33
+ "n_active_tokens": 8192,
34
+ "neuronxcc_version": "2.21.33363.0+82129205",
35
+ "on_device_sampling": true,
36
+ "optimum_neuron_version": "0.4.6.dev3",
37
+ "output_logits": false,
38
+ "pp_degree": 1,
39
+ "prefill_chunk_size": 1024,
40
+ "sequence_length": 8192,
41
+ "speculation_length": 0,
42
+ "start_rank_id": 0,
43
+ "target": "trn1",
44
+ "torch_dtype": "bfloat16",
45
+ "tp_degree": 2
46
+ },
47
+ "num_attention_heads": 32,
48
+ "num_hidden_layers": 16,
49
+ "num_key_value_heads": 8,
50
+ "pretraining_tp": 1,
51
+ "rms_norm_eps": 1e-05,
52
+ "rope_scaling": {
53
+ "factor": 32.0,
54
+ "high_freq_factor": 4.0,
55
+ "low_freq_factor": 1.0,
56
+ "original_max_position_embeddings": 8192,
57
+ "rope_type": "llama3"
58
+ },
59
+ "rope_theta": 500000.0,
60
+ "tie_word_embeddings": true,
61
+ "unsloth_fixed": true,
62
+ "use_cache": true,
63
+ "vocab_size": 128256
64
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/cf6b9a360dcf294104671106bae2adbd9fd291823bb60a351883163684073231/124119d9f41b205009a0.json ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "unsloth/Llama-3.2-1B-Instruct",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "LlamaForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "dtype": "bfloat16",
11
+ "head_dim": 64,
12
+ "hidden_act": "silu",
13
+ "hidden_size": 2048,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 8192,
16
+ "max_position_embeddings": 131072,
17
+ "mlp_bias": false,
18
+ "model_type": "llama",
19
+ "neuron": {
20
+ "_serialized_key": "NxDNeuronConfig",
21
+ "batch_size": 1,
22
+ "capacity_factor": null,
23
+ "checkpoint_id": "unsloth/Llama-3.2-1B-Instruct",
24
+ "checkpoint_revision": "5a8abab4a5d6f164389b1079fb721cfab8d7126c",
25
+ "continuous_batching": false,
26
+ "ep_degree": 1,
27
+ "fused_qkv": true,
28
+ "glu_mlp": true,
29
+ "local_ranks_size": 2,
30
+ "max_batch_size": 1,
31
+ "max_context_length": 4096,
32
+ "max_topk": 256,
33
+ "n_active_tokens": 4096,
34
+ "neuronxcc_version": "2.21.33363.0+82129205",
35
+ "on_device_sampling": true,
36
+ "optimum_neuron_version": "0.4.6.dev3",
37
+ "output_logits": false,
38
+ "pp_degree": 1,
39
+ "prefill_chunk_size": 1024,
40
+ "sequence_length": 4096,
41
+ "speculation_length": 0,
42
+ "start_rank_id": 0,
43
+ "target": "trn1",
44
+ "torch_dtype": "bfloat16",
45
+ "tp_degree": 2
46
+ },
47
+ "num_attention_heads": 32,
48
+ "num_hidden_layers": 16,
49
+ "num_key_value_heads": 8,
50
+ "pretraining_tp": 1,
51
+ "rms_norm_eps": 1e-05,
52
+ "rope_scaling": {
53
+ "factor": 32.0,
54
+ "high_freq_factor": 4.0,
55
+ "low_freq_factor": 1.0,
56
+ "original_max_position_embeddings": 8192,
57
+ "rope_type": "llama3"
58
+ },
59
+ "rope_theta": 500000.0,
60
+ "tie_word_embeddings": true,
61
+ "unsloth_fixed": true,
62
+ "use_cache": true,
63
+ "vocab_size": 128256
64
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/cf6b9a360dcf294104671106bae2adbd9fd291823bb60a351883163684073231/80f4956a4a5775280772.json ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "unsloth/Llama-3.2-1B-Instruct",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "LlamaForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "dtype": "bfloat16",
11
+ "head_dim": 64,
12
+ "hidden_act": "silu",
13
+ "hidden_size": 2048,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 8192,
16
+ "max_position_embeddings": 131072,
17
+ "mlp_bias": false,
18
+ "model_type": "llama",
19
+ "neuron": {
20
+ "_serialized_key": "NxDNeuronConfig",
21
+ "batch_size": 1,
22
+ "capacity_factor": null,
23
+ "checkpoint_id": "unsloth/Llama-3.2-1B-Instruct",
24
+ "checkpoint_revision": null,
25
+ "continuous_batching": false,
26
+ "ep_degree": 1,
27
+ "fused_qkv": false,
28
+ "glu_mlp": true,
29
+ "local_ranks_size": 2,
30
+ "max_batch_size": 1,
31
+ "max_context_length": 4096,
32
+ "max_topk": 256,
33
+ "n_active_tokens": 4096,
34
+ "neuronxcc_version": "2.21.33363.0+82129205",
35
+ "on_device_sampling": false,
36
+ "optimum_neuron_version": "0.4.6.dev3",
37
+ "output_logits": false,
38
+ "pp_degree": 1,
39
+ "prefill_chunk_size": 0,
40
+ "sequence_length": 4096,
41
+ "speculation_length": 5,
42
+ "start_rank_id": 0,
43
+ "target": "trn1",
44
+ "torch_dtype": "bfloat16",
45
+ "tp_degree": 2
46
+ },
47
+ "num_attention_heads": 32,
48
+ "num_hidden_layers": 16,
49
+ "num_key_value_heads": 8,
50
+ "pretraining_tp": 1,
51
+ "rms_norm_eps": 1e-05,
52
+ "rope_scaling": {
53
+ "factor": 32.0,
54
+ "high_freq_factor": 4.0,
55
+ "low_freq_factor": 1.0,
56
+ "original_max_position_embeddings": 8192,
57
+ "rope_type": "llama3"
58
+ },
59
+ "rope_theta": 500000.0,
60
+ "tie_word_embeddings": true,
61
+ "unsloth_fixed": true,
62
+ "use_cache": true,
63
+ "vocab_size": 128256
64
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/cf6b9a360dcf294104671106bae2adbd9fd291823bb60a351883163684073231/9df84ad1fbfa85b8d13d.json ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "unsloth/Llama-3.2-1B-Instruct",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "LlamaForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "dtype": "bfloat16",
11
+ "head_dim": 64,
12
+ "hidden_act": "silu",
13
+ "hidden_size": 2048,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 8192,
16
+ "max_position_embeddings": 131072,
17
+ "mlp_bias": false,
18
+ "model_type": "llama",
19
+ "neuron": {
20
+ "_serialized_key": "NxDNeuronConfig",
21
+ "batch_size": 4,
22
+ "capacity_factor": null,
23
+ "checkpoint_id": "unsloth/Llama-3.2-1B-Instruct",
24
+ "checkpoint_revision": "5a8abab4a5d6f164389b1079fb721cfab8d7126c",
25
+ "continuous_batching": true,
26
+ "ep_degree": 1,
27
+ "fused_qkv": true,
28
+ "glu_mlp": true,
29
+ "local_ranks_size": 2,
30
+ "max_batch_size": 4,
31
+ "max_context_length": 1024,
32
+ "max_topk": 256,
33
+ "n_active_tokens": 1024,
34
+ "neuronxcc_version": "2.21.33363.0+82129205",
35
+ "on_device_sampling": true,
36
+ "optimum_neuron_version": "0.4.6.dev3",
37
+ "output_logits": false,
38
+ "pp_degree": 1,
39
+ "prefill_chunk_size": 0,
40
+ "sequence_length": 1024,
41
+ "speculation_length": 0,
42
+ "start_rank_id": 0,
43
+ "target": "trn1",
44
+ "torch_dtype": "bfloat16",
45
+ "tp_degree": 2
46
+ },
47
+ "num_attention_heads": 32,
48
+ "num_hidden_layers": 16,
49
+ "num_key_value_heads": 8,
50
+ "pretraining_tp": 1,
51
+ "rms_norm_eps": 1e-05,
52
+ "rope_scaling": {
53
+ "factor": 32.0,
54
+ "high_freq_factor": 4.0,
55
+ "low_freq_factor": 1.0,
56
+ "original_max_position_embeddings": 8192,
57
+ "rope_type": "llama3"
58
+ },
59
+ "rope_theta": 500000.0,
60
+ "tie_word_embeddings": true,
61
+ "unsloth_fixed": true,
62
+ "use_cache": true,
63
+ "vocab_size": 128256
64
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/cf6b9a360dcf294104671106bae2adbd9fd291823bb60a351883163684073231/ce4667775a41bf1523e4.json ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "unsloth/Llama-3.2-1B-Instruct",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "LlamaForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "dtype": "bfloat16",
11
+ "head_dim": 64,
12
+ "hidden_act": "silu",
13
+ "hidden_size": 2048,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 8192,
16
+ "max_position_embeddings": 131072,
17
+ "mlp_bias": false,
18
+ "model_type": "llama",
19
+ "neuron": {
20
+ "_serialized_key": "NxDNeuronConfig",
21
+ "batch_size": 1,
22
+ "capacity_factor": null,
23
+ "checkpoint_id": "unsloth/Llama-3.2-1B-Instruct",
24
+ "checkpoint_revision": null,
25
+ "continuous_batching": false,
26
+ "ep_degree": 1,
27
+ "fused_qkv": false,
28
+ "glu_mlp": true,
29
+ "local_ranks_size": 2,
30
+ "max_batch_size": 1,
31
+ "max_context_length": 4096,
32
+ "max_topk": 256,
33
+ "n_active_tokens": 4096,
34
+ "neuronxcc_version": "2.21.33363.0+82129205",
35
+ "on_device_sampling": false,
36
+ "optimum_neuron_version": "0.4.6.dev3",
37
+ "output_logits": false,
38
+ "pp_degree": 1,
39
+ "prefill_chunk_size": 0,
40
+ "sequence_length": 4096,
41
+ "speculation_length": 0,
42
+ "start_rank_id": 0,
43
+ "target": "trn1",
44
+ "torch_dtype": "bfloat16",
45
+ "tp_degree": 2
46
+ },
47
+ "num_attention_heads": 32,
48
+ "num_hidden_layers": 16,
49
+ "num_key_value_heads": 8,
50
+ "pretraining_tp": 1,
51
+ "rms_norm_eps": 1e-05,
52
+ "rope_scaling": {
53
+ "factor": 32.0,
54
+ "high_freq_factor": 4.0,
55
+ "low_freq_factor": 1.0,
56
+ "original_max_position_embeddings": 8192,
57
+ "rope_type": "llama3"
58
+ },
59
+ "rope_theta": 500000.0,
60
+ "tie_word_embeddings": true,
61
+ "unsloth_fixed": true,
62
+ "use_cache": true,
63
+ "vocab_size": 128256
64
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/d139acf64685f15794bb983ff6eb881bdd31304bae88b0ce1ed20a54c21f2265/4aab0cb19b60948b084e.json ADDED
@@ -0,0 +1,59 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "hf-internal-testing/tiny-random-GraniteForCausalLM",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "GraniteForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "attention_multiplier": 1.0,
11
+ "dtype": "float32",
12
+ "embedding_multiplier": 1.0,
13
+ "hidden_act": "silu",
14
+ "hidden_size": 32,
15
+ "initializer_range": 0.02,
16
+ "intermediate_size": 64,
17
+ "logits_scaling": 1.0,
18
+ "max_position_embeddings": 2048,
19
+ "mlp_bias": false,
20
+ "model_type": "granite",
21
+ "neuron": {
22
+ "_serialized_key": "NxDNeuronConfig",
23
+ "batch_size": 1,
24
+ "capacity_factor": null,
25
+ "checkpoint_id": "hf-internal-testing/tiny-random-GraniteForCausalLM",
26
+ "checkpoint_revision": "c3074ebc0ac2fe545305f5e5f6cce2cc9b2aa0c5",
27
+ "continuous_batching": false,
28
+ "ep_degree": 1,
29
+ "fused_qkv": true,
30
+ "glu_mlp": true,
31
+ "local_ranks_size": 2,
32
+ "max_batch_size": 1,
33
+ "max_context_length": 1024,
34
+ "max_topk": 256,
35
+ "n_active_tokens": 1024,
36
+ "neuronxcc_version": "2.21.33363.0+82129205",
37
+ "on_device_sampling": true,
38
+ "optimum_neuron_version": "0.4.6.dev3",
39
+ "output_logits": false,
40
+ "pp_degree": 1,
41
+ "prefill_chunk_size": 0,
42
+ "sequence_length": 1024,
43
+ "speculation_length": 0,
44
+ "start_rank_id": 0,
45
+ "target": "trn1",
46
+ "torch_dtype": "float32",
47
+ "tp_degree": 2
48
+ },
49
+ "num_attention_heads": 4,
50
+ "num_hidden_layers": 2,
51
+ "num_key_value_heads": 4,
52
+ "residual_multiplier": 1.0,
53
+ "rms_norm_eps": 1e-06,
54
+ "rope_scaling": null,
55
+ "rope_theta": 10000.0,
56
+ "tie_word_embeddings": false,
57
+ "use_cache": true,
58
+ "vocab_size": 49152
59
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/gemma3_text/unsloth/gemma-3-270m-it/a421ae2ad1770f32b078.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "unsloth/gemma-3-270m-it",
4
+ "_sliding_window_pattern": 6,
5
+ "_task": "text-generation",
6
+ "architectures": [
7
+ "Gemma3ForCausalLM"
8
+ ],
9
+ "attention_bias": false,
10
+ "attention_dropout": 0.0,
11
+ "attn_logit_softcapping": null,
12
+ "dtype": "bfloat16",
13
+ "final_logit_softcapping": null,
14
+ "head_dim": 256,
15
+ "hidden_activation": "gelu_pytorch_tanh",
16
+ "hidden_size": 640,
17
+ "initializer_range": 0.02,
18
+ "intermediate_size": 2048,
19
+ "layer_types": [
20
+ "sliding_attention",
21
+ "sliding_attention",
22
+ "sliding_attention",
23
+ "sliding_attention",
24
+ "sliding_attention",
25
+ "full_attention",
26
+ "sliding_attention",
27
+ "sliding_attention",
28
+ "sliding_attention",
29
+ "sliding_attention",
30
+ "sliding_attention",
31
+ "full_attention",
32
+ "sliding_attention",
33
+ "sliding_attention",
34
+ "sliding_attention",
35
+ "sliding_attention",
36
+ "sliding_attention",
37
+ "full_attention"
38
+ ],
39
+ "max_position_embeddings": 32768,
40
+ "model_type": "gemma3_text",
41
+ "neuron": {
42
+ "_serialized_key": "NxDNeuronConfig",
43
+ "batch_size": 4,
44
+ "capacity_factor": null,
45
+ "checkpoint_id": "unsloth/gemma-3-270m-it",
46
+ "checkpoint_revision": "23cf460f6bb16954176b3ddcc8d4f250501458a9",
47
+ "continuous_batching": true,
48
+ "ep_degree": 1,
49
+ "fused_qkv": true,
50
+ "glu_mlp": true,
51
+ "local_ranks_size": 2,
52
+ "max_batch_size": 4,
53
+ "max_context_length": 1024,
54
+ "max_topk": 256,
55
+ "n_active_tokens": 1024,
56
+ "neuronxcc_version": "2.21.33363.0+82129205",
57
+ "on_device_sampling": true,
58
+ "optimum_neuron_version": "0.4.6.dev3",
59
+ "output_logits": false,
60
+ "pp_degree": 1,
61
+ "prefill_chunk_size": 0,
62
+ "sequence_length": 1024,
63
+ "speculation_length": 0,
64
+ "start_rank_id": 0,
65
+ "target": "trn1",
66
+ "torch_dtype": "bfloat16",
67
+ "tp_degree": 2
68
+ },
69
+ "num_attention_heads": 4,
70
+ "num_hidden_layers": 18,
71
+ "num_key_value_heads": 1,
72
+ "query_pre_attn_scalar": 256,
73
+ "rms_norm_eps": 1e-06,
74
+ "rope_local_base_freq": 10000.0,
75
+ "rope_scaling": null,
76
+ "rope_theta": 1000000.0,
77
+ "sliding_window": 512,
78
+ "unsloth_fixed": true,
79
+ "use_bidirectional_attention": false,
80
+ "use_cache": true,
81
+ "vocab_size": 262144
82
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/granite/hf-internal-testing/tiny-random-GraniteForCausalLM/4aab0cb19b60948b084e.json ADDED
@@ -0,0 +1,59 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "hf-internal-testing/tiny-random-GraniteForCausalLM",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "GraniteForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "attention_multiplier": 1.0,
11
+ "dtype": "float32",
12
+ "embedding_multiplier": 1.0,
13
+ "hidden_act": "silu",
14
+ "hidden_size": 32,
15
+ "initializer_range": 0.02,
16
+ "intermediate_size": 64,
17
+ "logits_scaling": 1.0,
18
+ "max_position_embeddings": 2048,
19
+ "mlp_bias": false,
20
+ "model_type": "granite",
21
+ "neuron": {
22
+ "_serialized_key": "NxDNeuronConfig",
23
+ "batch_size": 1,
24
+ "capacity_factor": null,
25
+ "checkpoint_id": "hf-internal-testing/tiny-random-GraniteForCausalLM",
26
+ "checkpoint_revision": "c3074ebc0ac2fe545305f5e5f6cce2cc9b2aa0c5",
27
+ "continuous_batching": false,
28
+ "ep_degree": 1,
29
+ "fused_qkv": true,
30
+ "glu_mlp": true,
31
+ "local_ranks_size": 2,
32
+ "max_batch_size": 1,
33
+ "max_context_length": 1024,
34
+ "max_topk": 256,
35
+ "n_active_tokens": 1024,
36
+ "neuronxcc_version": "2.21.33363.0+82129205",
37
+ "on_device_sampling": true,
38
+ "optimum_neuron_version": "0.4.6.dev3",
39
+ "output_logits": false,
40
+ "pp_degree": 1,
41
+ "prefill_chunk_size": 0,
42
+ "sequence_length": 1024,
43
+ "speculation_length": 0,
44
+ "start_rank_id": 0,
45
+ "target": "trn1",
46
+ "torch_dtype": "float32",
47
+ "tp_degree": 2
48
+ },
49
+ "num_attention_heads": 4,
50
+ "num_hidden_layers": 2,
51
+ "num_key_value_heads": 4,
52
+ "residual_multiplier": 1.0,
53
+ "rms_norm_eps": 1e-06,
54
+ "rope_scaling": null,
55
+ "rope_theta": 10000.0,
56
+ "tie_word_embeddings": false,
57
+ "use_cache": true,
58
+ "vocab_size": 49152
59
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/granite/ibm-granite/granite-3.1-2b-instruct/31629530ce956e3eb55e.json ADDED
@@ -0,0 +1,59 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "ibm-granite/granite-3.1-2b-instruct",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "GraniteForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.1,
10
+ "attention_multiplier": 0.015625,
11
+ "dtype": "bfloat16",
12
+ "embedding_multiplier": 12.0,
13
+ "hidden_act": "silu",
14
+ "hidden_size": 2048,
15
+ "initializer_range": 0.02,
16
+ "intermediate_size": 8192,
17
+ "logits_scaling": 8.0,
18
+ "max_position_embeddings": 131072,
19
+ "mlp_bias": false,
20
+ "model_type": "granite",
21
+ "neuron": {
22
+ "_serialized_key": "NxDNeuronConfig",
23
+ "batch_size": 1,
24
+ "capacity_factor": null,
25
+ "checkpoint_id": "ibm-granite/granite-3.1-2b-instruct",
26
+ "checkpoint_revision": "bbc2aed595bd38bd770263dc3ab831db9794441d",
27
+ "continuous_batching": false,
28
+ "ep_degree": 1,
29
+ "fused_qkv": true,
30
+ "glu_mlp": true,
31
+ "local_ranks_size": 2,
32
+ "max_batch_size": 1,
33
+ "max_context_length": 8192,
34
+ "max_topk": 256,
35
+ "n_active_tokens": 8192,
36
+ "neuronxcc_version": "2.21.33363.0+82129205",
37
+ "on_device_sampling": true,
38
+ "optimum_neuron_version": "0.4.6.dev3",
39
+ "output_logits": false,
40
+ "pp_degree": 1,
41
+ "prefill_chunk_size": 1024,
42
+ "sequence_length": 8192,
43
+ "speculation_length": 0,
44
+ "start_rank_id": 0,
45
+ "target": "trn1",
46
+ "torch_dtype": "bfloat16",
47
+ "tp_degree": 2
48
+ },
49
+ "num_attention_heads": 32,
50
+ "num_hidden_layers": 40,
51
+ "num_key_value_heads": 8,
52
+ "residual_multiplier": 0.22,
53
+ "rms_norm_eps": 1e-05,
54
+ "rope_scaling": null,
55
+ "rope_theta": 5000000.0,
56
+ "tie_word_embeddings": true,
57
+ "use_cache": true,
58
+ "vocab_size": 49155
59
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/llama/llamafactory/tiny-random-Llama-3/5d5179e2ac0b724548b0.json ADDED
@@ -0,0 +1,63 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "llamafactory/tiny-random-Llama-3",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "LlamaForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "dtype": "float16",
11
+ "head_dim": 4,
12
+ "hidden_act": "silu",
13
+ "hidden_size": 16,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 64,
16
+ "max_position_embeddings": 131072,
17
+ "mlp_bias": false,
18
+ "model_type": "llama",
19
+ "neuron": {
20
+ "_serialized_key": "NxDNeuronConfig",
21
+ "batch_size": 1,
22
+ "capacity_factor": null,
23
+ "checkpoint_id": "llamafactory/tiny-random-Llama-3",
24
+ "checkpoint_revision": "bf2a2e3bf199ad2ee96f02a3c00246c608db22a8",
25
+ "continuous_batching": false,
26
+ "ep_degree": 1,
27
+ "fused_qkv": true,
28
+ "glu_mlp": true,
29
+ "local_ranks_size": 2,
30
+ "max_batch_size": 1,
31
+ "max_context_length": 1024,
32
+ "max_topk": 256,
33
+ "n_active_tokens": 1024,
34
+ "neuronxcc_version": "2.21.33363.0+82129205",
35
+ "on_device_sampling": true,
36
+ "optimum_neuron_version": "0.4.6.dev3",
37
+ "output_logits": false,
38
+ "pp_degree": 1,
39
+ "prefill_chunk_size": 0,
40
+ "sequence_length": 1024,
41
+ "speculation_length": 0,
42
+ "start_rank_id": 0,
43
+ "target": "trn1",
44
+ "torch_dtype": "float16",
45
+ "tp_degree": 2
46
+ },
47
+ "num_attention_heads": 4,
48
+ "num_hidden_layers": 2,
49
+ "num_key_value_heads": 4,
50
+ "pretraining_tp": 1,
51
+ "rms_norm_eps": 1e-05,
52
+ "rope_scaling": {
53
+ "factor": 8.0,
54
+ "high_freq_factor": 4.0,
55
+ "low_freq_factor": 1.0,
56
+ "original_max_position_embeddings": 8192,
57
+ "rope_type": "llama3"
58
+ },
59
+ "rope_theta": 500000.0,
60
+ "tie_word_embeddings": false,
61
+ "use_cache": true,
62
+ "vocab_size": 128256
63
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/llama/unsloth/Llama-3.2-1B-Instruct/9df84ad1fbfa85b8d13d.json ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "unsloth/Llama-3.2-1B-Instruct",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "LlamaForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "dtype": "bfloat16",
11
+ "head_dim": 64,
12
+ "hidden_act": "silu",
13
+ "hidden_size": 2048,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 8192,
16
+ "max_position_embeddings": 131072,
17
+ "mlp_bias": false,
18
+ "model_type": "llama",
19
+ "neuron": {
20
+ "_serialized_key": "NxDNeuronConfig",
21
+ "batch_size": 4,
22
+ "capacity_factor": null,
23
+ "checkpoint_id": "unsloth/Llama-3.2-1B-Instruct",
24
+ "checkpoint_revision": "5a8abab4a5d6f164389b1079fb721cfab8d7126c",
25
+ "continuous_batching": true,
26
+ "ep_degree": 1,
27
+ "fused_qkv": true,
28
+ "glu_mlp": true,
29
+ "local_ranks_size": 2,
30
+ "max_batch_size": 4,
31
+ "max_context_length": 1024,
32
+ "max_topk": 256,
33
+ "n_active_tokens": 1024,
34
+ "neuronxcc_version": "2.21.33363.0+82129205",
35
+ "on_device_sampling": true,
36
+ "optimum_neuron_version": "0.4.6.dev3",
37
+ "output_logits": false,
38
+ "pp_degree": 1,
39
+ "prefill_chunk_size": 0,
40
+ "sequence_length": 1024,
41
+ "speculation_length": 0,
42
+ "start_rank_id": 0,
43
+ "target": "trn1",
44
+ "torch_dtype": "bfloat16",
45
+ "tp_degree": 2
46
+ },
47
+ "num_attention_heads": 32,
48
+ "num_hidden_layers": 16,
49
+ "num_key_value_heads": 8,
50
+ "pretraining_tp": 1,
51
+ "rms_norm_eps": 1e-05,
52
+ "rope_scaling": {
53
+ "factor": 32.0,
54
+ "high_freq_factor": 4.0,
55
+ "low_freq_factor": 1.0,
56
+ "original_max_position_embeddings": 8192,
57
+ "rope_type": "llama3"
58
+ },
59
+ "rope_theta": 500000.0,
60
+ "tie_word_embeddings": true,
61
+ "unsloth_fixed": true,
62
+ "use_cache": true,
63
+ "vocab_size": 128256
64
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/llama4_text/tiny-random/llama-4/d244406bccd80288e991.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "tiny-random/llama-4",
4
+ "_task": "text-generation",
5
+ "attention_bias": false,
6
+ "attention_chunk_size": 128,
7
+ "attention_dropout": 0.0,
8
+ "attn_scale": 0.1,
9
+ "attn_temperature_tuning": 4,
10
+ "cache_implementation": "hybrid",
11
+ "dtype": "bfloat16",
12
+ "floor_scale": 8192,
13
+ "for_llm_compressor": false,
14
+ "head_dim": 32,
15
+ "hidden_act": "silu",
16
+ "hidden_size": 32,
17
+ "initializer_range": 0.02,
18
+ "interleave_moe_layer_step": 2,
19
+ "intermediate_size": 64,
20
+ "intermediate_size_mlp": 128,
21
+ "layer_types": [
22
+ "chunked_attention",
23
+ "chunked_attention",
24
+ "chunked_attention",
25
+ "full_attention"
26
+ ],
27
+ "max_position_embeddings": 1048576,
28
+ "model_type": "llama4_text",
29
+ "moe_layers": [
30
+ 1,
31
+ 3
32
+ ],
33
+ "neuron": {
34
+ "_serialized_key": "NxDNeuronConfig",
35
+ "batch_size": 1,
36
+ "capacity_factor": null,
37
+ "checkpoint_id": "tiny-random/llama-4",
38
+ "checkpoint_revision": "9e716f5d4d1ffe0a44a15f46f4a12b840439aba4",
39
+ "continuous_batching": false,
40
+ "ep_degree": 1,
41
+ "fused_qkv": false,
42
+ "glu_mlp": true,
43
+ "local_ranks_size": 2,
44
+ "max_batch_size": 1,
45
+ "max_context_length": 1024,
46
+ "max_topk": 256,
47
+ "n_active_tokens": 1024,
48
+ "neuronxcc_version": "2.21.33363.0+82129205",
49
+ "on_device_sampling": true,
50
+ "optimum_neuron_version": "0.4.6.dev3",
51
+ "output_logits": false,
52
+ "pp_degree": 1,
53
+ "prefill_chunk_size": 0,
54
+ "sequence_length": 1024,
55
+ "speculation_length": 0,
56
+ "start_rank_id": 0,
57
+ "target": "trn1",
58
+ "torch_dtype": "bfloat16",
59
+ "tp_degree": 2
60
+ },
61
+ "no_rope_layers": [
62
+ 1,
63
+ 1,
64
+ 1,
65
+ 0
66
+ ],
67
+ "num_attention_heads": 1,
68
+ "num_experts_per_tok": 1,
69
+ "num_hidden_layers": 4,
70
+ "num_key_value_heads": 1,
71
+ "num_local_experts": 8,
72
+ "output_router_logits": false,
73
+ "rms_norm_eps": 1e-05,
74
+ "rope_scaling": null,
75
+ "rope_theta": 500000.0,
76
+ "router_aux_loss_coef": 0.001,
77
+ "router_jitter_noise": 0.0,
78
+ "tie_word_embeddings": true,
79
+ "use_cache": true,
80
+ "use_qk_norm": true,
81
+ "vocab_size": 202048
82
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/mixtral/dacorvo/Mixtral-tiny/1171cdc1e0668ecced1b.json ADDED
@@ -0,0 +1,59 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "dacorvo/Mixtral-tiny",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "MixtralForCausalLM"
7
+ ],
8
+ "attention_dropout": 0.0,
9
+ "dtype": "float16",
10
+ "head_dim": 32,
11
+ "hidden_act": "silu",
12
+ "hidden_size": 1024,
13
+ "initializer_range": 0.02,
14
+ "intermediate_size": 3584,
15
+ "max_position_embeddings": 1024,
16
+ "model_type": "mixtral",
17
+ "neuron": {
18
+ "_serialized_key": "NxDNeuronConfig",
19
+ "batch_size": 1,
20
+ "capacity_factor": null,
21
+ "checkpoint_id": "dacorvo/Mixtral-tiny",
22
+ "checkpoint_revision": "c557ba205ddff6ea911f4719e0d543d6c08356b6",
23
+ "continuous_batching": false,
24
+ "ep_degree": 1,
25
+ "fused_qkv": false,
26
+ "glu_mlp": true,
27
+ "local_ranks_size": 2,
28
+ "max_batch_size": 1,
29
+ "max_context_length": 1024,
30
+ "max_topk": 256,
31
+ "n_active_tokens": 1024,
32
+ "neuronxcc_version": "2.21.33363.0+82129205",
33
+ "on_device_sampling": false,
34
+ "optimum_neuron_version": "0.4.6.dev3",
35
+ "output_logits": false,
36
+ "pp_degree": 1,
37
+ "prefill_chunk_size": 0,
38
+ "sequence_length": 1024,
39
+ "speculation_length": 0,
40
+ "start_rank_id": 0,
41
+ "target": "trn1",
42
+ "torch_dtype": "float16",
43
+ "tp_degree": 2
44
+ },
45
+ "num_attention_heads": 32,
46
+ "num_experts_per_tok": 2,
47
+ "num_hidden_layers": 2,
48
+ "num_key_value_heads": 8,
49
+ "num_local_experts": 8,
50
+ "output_router_logits": false,
51
+ "rms_norm_eps": 1e-05,
52
+ "rope_theta": 10000.0,
53
+ "router_aux_loss_coef": 0.001,
54
+ "router_jitter_noise": 0.0,
55
+ "sliding_window": 4096,
56
+ "tie_word_embeddings": false,
57
+ "use_cache": true,
58
+ "vocab_size": 32000
59
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/phi3/yujiepan/phi-4-tiny-random/6b7aebc14c435ae3db63.json ADDED
@@ -0,0 +1,60 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "yujiepan/phi-4-tiny-random",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "Phi3ForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "auto_map": {},
11
+ "dtype": "bfloat16",
12
+ "embd_pdrop": 0.0,
13
+ "hidden_act": "silu",
14
+ "hidden_size": 16,
15
+ "initializer_range": 0.02,
16
+ "intermediate_size": 32,
17
+ "max_position_embeddings": 16384,
18
+ "model_type": "phi3",
19
+ "neuron": {
20
+ "_serialized_key": "NxDNeuronConfig",
21
+ "batch_size": 1,
22
+ "capacity_factor": null,
23
+ "checkpoint_id": "yujiepan/phi-4-tiny-random",
24
+ "checkpoint_revision": "18a9a1168dc97ac6d128f811925670c275610f5a",
25
+ "continuous_batching": false,
26
+ "ep_degree": 1,
27
+ "fused_qkv": true,
28
+ "glu_mlp": true,
29
+ "local_ranks_size": 2,
30
+ "max_batch_size": 1,
31
+ "max_context_length": 1024,
32
+ "max_topk": 256,
33
+ "n_active_tokens": 1024,
34
+ "neuronxcc_version": "2.21.33363.0+82129205",
35
+ "on_device_sampling": true,
36
+ "optimum_neuron_version": "0.4.6.dev3",
37
+ "output_logits": false,
38
+ "pp_degree": 1,
39
+ "prefill_chunk_size": 0,
40
+ "sequence_length": 1024,
41
+ "speculation_length": 0,
42
+ "start_rank_id": 0,
43
+ "target": "trn1",
44
+ "torch_dtype": "bfloat16",
45
+ "tp_degree": 2
46
+ },
47
+ "num_attention_heads": 2,
48
+ "num_hidden_layers": 2,
49
+ "num_key_value_heads": 1,
50
+ "original_max_position_embeddings": 16384,
51
+ "partial_rotary_factor": 1.0,
52
+ "resid_pdrop": 0.0,
53
+ "rms_norm_eps": 1e-05,
54
+ "rope_scaling": null,
55
+ "rope_theta": 250000,
56
+ "sliding_window": null,
57
+ "tie_word_embeddings": false,
58
+ "use_cache": true,
59
+ "vocab_size": 100352
60
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/qwen2/Qwen/Qwen2.5-0.5B/2c6cc66c4776ff9413f0.json ADDED
@@ -0,0 +1,83 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "Qwen/Qwen2.5-0.5B",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "Qwen2ForCausalLM"
7
+ ],
8
+ "attention_dropout": 0.0,
9
+ "dtype": "bfloat16",
10
+ "hidden_act": "silu",
11
+ "hidden_size": 896,
12
+ "initializer_range": 0.02,
13
+ "intermediate_size": 4864,
14
+ "layer_types": [
15
+ "full_attention",
16
+ "full_attention",
17
+ "full_attention",
18
+ "full_attention",
19
+ "full_attention",
20
+ "full_attention",
21
+ "full_attention",
22
+ "full_attention",
23
+ "full_attention",
24
+ "full_attention",
25
+ "full_attention",
26
+ "full_attention",
27
+ "full_attention",
28
+ "full_attention",
29
+ "full_attention",
30
+ "full_attention",
31
+ "full_attention",
32
+ "full_attention",
33
+ "full_attention",
34
+ "full_attention",
35
+ "full_attention",
36
+ "full_attention",
37
+ "full_attention",
38
+ "full_attention"
39
+ ],
40
+ "max_position_embeddings": 32768,
41
+ "max_window_layers": 24,
42
+ "model_type": "qwen2",
43
+ "neuron": {
44
+ "_serialized_key": "NxDNeuronConfig",
45
+ "batch_size": 4,
46
+ "capacity_factor": null,
47
+ "checkpoint_id": "Qwen/Qwen2.5-0.5B",
48
+ "checkpoint_revision": "060db6499f32faf8b98477b0a26969ef7d8b9987",
49
+ "continuous_batching": true,
50
+ "ep_degree": 1,
51
+ "fused_qkv": false,
52
+ "glu_mlp": true,
53
+ "local_ranks_size": 2,
54
+ "max_batch_size": 4,
55
+ "max_context_length": 1024,
56
+ "max_topk": 256,
57
+ "n_active_tokens": 1024,
58
+ "neuronxcc_version": "2.21.33363.0+82129205",
59
+ "on_device_sampling": false,
60
+ "optimum_neuron_version": "0.4.6.dev3",
61
+ "output_logits": false,
62
+ "pp_degree": 1,
63
+ "prefill_chunk_size": 0,
64
+ "sequence_length": 1024,
65
+ "speculation_length": 0,
66
+ "start_rank_id": 0,
67
+ "target": "trn1",
68
+ "torch_dtype": "bfloat16",
69
+ "tp_degree": 2
70
+ },
71
+ "num_attention_heads": 14,
72
+ "num_hidden_layers": 24,
73
+ "num_key_value_heads": 2,
74
+ "rms_norm_eps": 1e-06,
75
+ "rope_scaling": null,
76
+ "rope_theta": 1000000.0,
77
+ "sliding_window": null,
78
+ "tie_word_embeddings": true,
79
+ "use_cache": true,
80
+ "use_mrope": false,
81
+ "use_sliding_window": false,
82
+ "vocab_size": 151936
83
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/qwen2/yujiepan/qwen2.5-128k-tiny-random/7e378b918d8aa52c31f9.json ADDED
@@ -0,0 +1,65 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "yujiepan/qwen2.5-128k-tiny-random",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "Qwen2ForCausalLM"
7
+ ],
8
+ "attention_dropout": 0.0,
9
+ "dtype": "bfloat16",
10
+ "hidden_act": "silu",
11
+ "hidden_size": 8,
12
+ "initializer_range": 0.02,
13
+ "intermediate_size": 16,
14
+ "layer_types": [
15
+ "full_attention",
16
+ "full_attention"
17
+ ],
18
+ "max_position_embeddings": 32768,
19
+ "max_window_layers": 1,
20
+ "model_type": "qwen2",
21
+ "neuron": {
22
+ "_serialized_key": "NxDNeuronConfig",
23
+ "batch_size": 1,
24
+ "capacity_factor": null,
25
+ "checkpoint_id": "yujiepan/qwen2.5-128k-tiny-random",
26
+ "checkpoint_revision": "c8296d4ca3f87782876d2382fbb6481d1beb8ef0",
27
+ "continuous_batching": false,
28
+ "ep_degree": 1,
29
+ "fused_qkv": false,
30
+ "glu_mlp": true,
31
+ "local_ranks_size": 2,
32
+ "max_batch_size": 1,
33
+ "max_context_length": 1024,
34
+ "max_topk": 256,
35
+ "n_active_tokens": 1024,
36
+ "neuronxcc_version": "2.21.33363.0+82129205",
37
+ "on_device_sampling": true,
38
+ "optimum_neuron_version": "0.4.6.dev3",
39
+ "output_logits": false,
40
+ "pp_degree": 1,
41
+ "prefill_chunk_size": 0,
42
+ "sequence_length": 1024,
43
+ "speculation_length": 0,
44
+ "start_rank_id": 0,
45
+ "target": "trn1",
46
+ "torch_dtype": "bfloat16",
47
+ "tp_degree": 2
48
+ },
49
+ "num_attention_heads": 4,
50
+ "num_hidden_layers": 2,
51
+ "num_key_value_heads": 2,
52
+ "rms_norm_eps": 1e-06,
53
+ "rope_scaling": {
54
+ "factor": 4.0,
55
+ "original_max_position_embeddings": 32768,
56
+ "rope_type": "yarn",
57
+ "type": "yarn"
58
+ },
59
+ "rope_theta": 1000000.0,
60
+ "sliding_window": null,
61
+ "tie_word_embeddings": false,
62
+ "use_cache": true,
63
+ "use_sliding_window": false,
64
+ "vocab_size": 152064
65
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/qwen3/Qwen/Qwen3-0.6B/c3b960b90a5e3c592c04.json ADDED
@@ -0,0 +1,88 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "Qwen/Qwen3-0.6B",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "Qwen3ForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "dtype": "bfloat16",
11
+ "head_dim": 128,
12
+ "hidden_act": "silu",
13
+ "hidden_size": 1024,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 3072,
16
+ "layer_types": [
17
+ "full_attention",
18
+ "full_attention",
19
+ "full_attention",
20
+ "full_attention",
21
+ "full_attention",
22
+ "full_attention",
23
+ "full_attention",
24
+ "full_attention",
25
+ "full_attention",
26
+ "full_attention",
27
+ "full_attention",
28
+ "full_attention",
29
+ "full_attention",
30
+ "full_attention",
31
+ "full_attention",
32
+ "full_attention",
33
+ "full_attention",
34
+ "full_attention",
35
+ "full_attention",
36
+ "full_attention",
37
+ "full_attention",
38
+ "full_attention",
39
+ "full_attention",
40
+ "full_attention",
41
+ "full_attention",
42
+ "full_attention",
43
+ "full_attention",
44
+ "full_attention"
45
+ ],
46
+ "max_position_embeddings": 40960,
47
+ "max_window_layers": 28,
48
+ "model_type": "qwen3",
49
+ "neuron": {
50
+ "_serialized_key": "NxDNeuronConfig",
51
+ "batch_size": 4,
52
+ "capacity_factor": null,
53
+ "checkpoint_id": "Qwen/Qwen3-0.6B",
54
+ "checkpoint_revision": "c1899de289a04d12100db370d81485cdf75e47ca",
55
+ "continuous_batching": true,
56
+ "ep_degree": 1,
57
+ "fused_qkv": true,
58
+ "glu_mlp": true,
59
+ "local_ranks_size": 2,
60
+ "max_batch_size": 4,
61
+ "max_context_length": 1024,
62
+ "max_topk": 256,
63
+ "n_active_tokens": 1024,
64
+ "neuronxcc_version": "2.21.33363.0+82129205",
65
+ "on_device_sampling": false,
66
+ "optimum_neuron_version": "0.4.6.dev3",
67
+ "output_logits": false,
68
+ "pp_degree": 1,
69
+ "prefill_chunk_size": 0,
70
+ "sequence_length": 1024,
71
+ "speculation_length": 0,
72
+ "start_rank_id": 0,
73
+ "target": "trn1",
74
+ "torch_dtype": "bfloat16",
75
+ "tp_degree": 2
76
+ },
77
+ "num_attention_heads": 16,
78
+ "num_hidden_layers": 28,
79
+ "num_key_value_heads": 8,
80
+ "rms_norm_eps": 1e-06,
81
+ "rope_scaling": null,
82
+ "rope_theta": 1000000,
83
+ "sliding_window": null,
84
+ "tie_word_embeddings": true,
85
+ "use_cache": true,
86
+ "use_sliding_window": false,
87
+ "vocab_size": 151936
88
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/qwen3/Qwen/Qwen3-Embedding-0.6B/43be67542ceab3db3486.json ADDED
@@ -0,0 +1,88 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "Qwen/Qwen3-Embedding-0.6B",
4
+ "_task": "feature-extraction",
5
+ "architectures": [
6
+ "Qwen3ForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "dtype": "bfloat16",
11
+ "head_dim": 128,
12
+ "hidden_act": "silu",
13
+ "hidden_size": 1024,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 3072,
16
+ "layer_types": [
17
+ "full_attention",
18
+ "full_attention",
19
+ "full_attention",
20
+ "full_attention",
21
+ "full_attention",
22
+ "full_attention",
23
+ "full_attention",
24
+ "full_attention",
25
+ "full_attention",
26
+ "full_attention",
27
+ "full_attention",
28
+ "full_attention",
29
+ "full_attention",
30
+ "full_attention",
31
+ "full_attention",
32
+ "full_attention",
33
+ "full_attention",
34
+ "full_attention",
35
+ "full_attention",
36
+ "full_attention",
37
+ "full_attention",
38
+ "full_attention",
39
+ "full_attention",
40
+ "full_attention",
41
+ "full_attention",
42
+ "full_attention",
43
+ "full_attention",
44
+ "full_attention"
45
+ ],
46
+ "max_position_embeddings": 32768,
47
+ "max_window_layers": 28,
48
+ "model_type": "qwen3",
49
+ "neuron": {
50
+ "_serialized_key": "NxDNeuronConfig",
51
+ "batch_size": 4,
52
+ "capacity_factor": null,
53
+ "checkpoint_id": "Qwen/Qwen3-Embedding-0.6B",
54
+ "checkpoint_revision": "c54f2e6e80b2d7b7de06f51cec4959f6b3e03418",
55
+ "continuous_batching": false,
56
+ "ep_degree": 1,
57
+ "fused_qkv": true,
58
+ "glu_mlp": true,
59
+ "local_ranks_size": 2,
60
+ "max_batch_size": 4,
61
+ "max_context_length": 8192,
62
+ "max_topk": 256,
63
+ "n_active_tokens": 8192,
64
+ "neuronxcc_version": "2.21.33363.0+82129205",
65
+ "on_device_sampling": false,
66
+ "optimum_neuron_version": "0.4.6.dev3",
67
+ "output_logits": false,
68
+ "pp_degree": 1,
69
+ "prefill_chunk_size": 0,
70
+ "sequence_length": 8192,
71
+ "speculation_length": 0,
72
+ "start_rank_id": 0,
73
+ "target": "trn1",
74
+ "torch_dtype": "bfloat16",
75
+ "tp_degree": 2
76
+ },
77
+ "num_attention_heads": 16,
78
+ "num_hidden_layers": 28,
79
+ "num_key_value_heads": 8,
80
+ "rms_norm_eps": 1e-06,
81
+ "rope_scaling": null,
82
+ "rope_theta": 1000000,
83
+ "sliding_window": null,
84
+ "tie_word_embeddings": true,
85
+ "use_cache": true,
86
+ "use_sliding_window": false,
87
+ "vocab_size": 151669
88
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/qwen3_moe/optimum-internal-testing/tiny-random-qwen3_moe/ad0f2997db25e16e085b.json ADDED
@@ -0,0 +1,66 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "optimum-internal-testing/tiny-random-qwen3_moe",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "Qwen3MoeForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "decoder_sparse_step": 2,
11
+ "dtype": "float32",
12
+ "head_dim": 32,
13
+ "hidden_act": "silu",
14
+ "hidden_size": 64,
15
+ "initializer_range": 0.02,
16
+ "intermediate_size": 128,
17
+ "max_position_embeddings": 40960,
18
+ "max_window_layers": 1,
19
+ "mlp_only_layers": [],
20
+ "model_type": "qwen3_moe",
21
+ "moe_intermediate_size": 128,
22
+ "neuron": {
23
+ "_serialized_key": "NxDNeuronConfig",
24
+ "batch_size": 1,
25
+ "capacity_factor": null,
26
+ "checkpoint_id": "optimum-internal-testing/tiny-random-qwen3_moe",
27
+ "checkpoint_revision": "e0230be2839556b44b7400a233c73c74b4abb7af",
28
+ "continuous_batching": false,
29
+ "ep_degree": 1,
30
+ "fused_qkv": false,
31
+ "glu_mlp": true,
32
+ "local_ranks_size": 2,
33
+ "max_batch_size": 1,
34
+ "max_context_length": 1024,
35
+ "max_topk": 256,
36
+ "n_active_tokens": 1024,
37
+ "neuronxcc_version": "2.21.33363.0+82129205",
38
+ "on_device_sampling": true,
39
+ "optimum_neuron_version": "0.4.6.dev3",
40
+ "output_logits": false,
41
+ "pp_degree": 1,
42
+ "prefill_chunk_size": 0,
43
+ "sequence_length": 1024,
44
+ "speculation_length": 0,
45
+ "start_rank_id": 0,
46
+ "target": "trn1",
47
+ "torch_dtype": "float32",
48
+ "tp_degree": 2
49
+ },
50
+ "norm_topk_prob": true,
51
+ "num_attention_heads": 2,
52
+ "num_experts": 8,
53
+ "num_experts_per_tok": 2,
54
+ "num_hidden_layers": 2,
55
+ "num_key_value_heads": 1,
56
+ "output_router_logits": false,
57
+ "rms_norm_eps": 1e-06,
58
+ "rope_scaling": null,
59
+ "rope_theta": 1000000.0,
60
+ "router_aux_loss_coef": 0.001,
61
+ "sliding_window": null,
62
+ "tie_word_embeddings": true,
63
+ "use_cache": true,
64
+ "use_sliding_window": false,
65
+ "vocab_size": 151936
66
+ }
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev3/smollm3/HuggingFaceTB/SmolLM3-3B/b2c32dabac1b1c034891.json ADDED
@@ -0,0 +1,135 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "HuggingFaceTB/SmolLM3-3B",
4
+ "_task": "text-generation",
5
+ "architectures": [
6
+ "SmolLM3ForCausalLM"
7
+ ],
8
+ "attention_bias": false,
9
+ "attention_dropout": 0.0,
10
+ "dtype": "bfloat16",
11
+ "hidden_act": "silu",
12
+ "hidden_size": 2048,
13
+ "initializer_range": 0.02,
14
+ "intermediate_size": 11008,
15
+ "layer_types": [
16
+ "full_attention",
17
+ "full_attention",
18
+ "full_attention",
19
+ "full_attention",
20
+ "full_attention",
21
+ "full_attention",
22
+ "full_attention",
23
+ "full_attention",
24
+ "full_attention",
25
+ "full_attention",
26
+ "full_attention",
27
+ "full_attention",
28
+ "full_attention",
29
+ "full_attention",
30
+ "full_attention",
31
+ "full_attention",
32
+ "full_attention",
33
+ "full_attention",
34
+ "full_attention",
35
+ "full_attention",
36
+ "full_attention",
37
+ "full_attention",
38
+ "full_attention",
39
+ "full_attention",
40
+ "full_attention",
41
+ "full_attention",
42
+ "full_attention",
43
+ "full_attention",
44
+ "full_attention",
45
+ "full_attention",
46
+ "full_attention",
47
+ "full_attention",
48
+ "full_attention",
49
+ "full_attention",
50
+ "full_attention",
51
+ "full_attention"
52
+ ],
53
+ "max_position_embeddings": 65536,
54
+ "max_window_layers": 28,
55
+ "mlp_bias": false,
56
+ "model_type": "smollm3",
57
+ "neuron": {
58
+ "_serialized_key": "NxDNeuronConfig",
59
+ "batch_size": 4,
60
+ "capacity_factor": null,
61
+ "checkpoint_id": "HuggingFaceTB/SmolLM3-3B",
62
+ "checkpoint_revision": "a07cc9a04f16550a088caea529712d1d335b0ac1",
63
+ "continuous_batching": true,
64
+ "ep_degree": 1,
65
+ "fused_qkv": true,
66
+ "glu_mlp": true,
67
+ "local_ranks_size": 2,
68
+ "max_batch_size": 4,
69
+ "max_context_length": 1024,
70
+ "max_topk": 256,
71
+ "n_active_tokens": 1024,
72
+ "neuronxcc_version": "2.21.33363.0+82129205",
73
+ "on_device_sampling": true,
74
+ "optimum_neuron_version": "0.4.6.dev3",
75
+ "output_logits": false,
76
+ "pp_degree": 1,
77
+ "prefill_chunk_size": 0,
78
+ "sequence_length": 1024,
79
+ "speculation_length": 0,
80
+ "start_rank_id": 0,
81
+ "target": "trn1",
82
+ "torch_dtype": "bfloat16",
83
+ "tp_degree": 2
84
+ },
85
+ "no_rope_layer_interval": 4,
86
+ "no_rope_layers": [
87
+ 1,
88
+ 1,
89
+ 1,
90
+ 0,
91
+ 1,
92
+ 1,
93
+ 1,
94
+ 0,
95
+ 1,
96
+ 1,
97
+ 1,
98
+ 0,
99
+ 1,
100
+ 1,
101
+ 1,
102
+ 0,
103
+ 1,
104
+ 1,
105
+ 1,
106
+ 0,
107
+ 1,
108
+ 1,
109
+ 1,
110
+ 0,
111
+ 1,
112
+ 1,
113
+ 1,
114
+ 0,
115
+ 1,
116
+ 1,
117
+ 1,
118
+ 0,
119
+ 1,
120
+ 1,
121
+ 1,
122
+ 0
123
+ ],
124
+ "num_attention_heads": 16,
125
+ "num_hidden_layers": 36,
126
+ "num_key_value_heads": 4,
127
+ "pretraining_tp": 2,
128
+ "rms_norm_eps": 1e-06,
129
+ "rope_scaling": null,
130
+ "rope_theta": 5000000.0,
131
+ "sliding_window": null,
132
+ "use_cache": false,
133
+ "use_sliding_window": false,
134
+ "vocab_size": 128256
135
+ }
neuronxcc-2.21.33363.0+82129205/MODULE_10244305442015770634+e30acd3a/model.neff CHANGED
Binary files a/neuronxcc-2.21.33363.0+82129205/MODULE_10244305442015770634+e30acd3a/model.neff and b/neuronxcc-2.21.33363.0+82129205/MODULE_10244305442015770634+e30acd3a/model.neff differ
 
neuronxcc-2.21.33363.0+82129205/MODULE_10606948783918825529+e30acd3a/model.neff CHANGED
Binary files a/neuronxcc-2.21.33363.0+82129205/MODULE_10606948783918825529+e30acd3a/model.neff and b/neuronxcc-2.21.33363.0+82129205/MODULE_10606948783918825529+e30acd3a/model.neff differ
 
neuronxcc-2.21.33363.0+82129205/MODULE_10645643398657092095+e30acd3a/model.neff CHANGED
Binary files a/neuronxcc-2.21.33363.0+82129205/MODULE_10645643398657092095+e30acd3a/model.neff and b/neuronxcc-2.21.33363.0+82129205/MODULE_10645643398657092095+e30acd3a/model.neff differ
 
neuronxcc-2.21.33363.0+82129205/MODULE_10661660426924300837+e30acd3a/model.neff CHANGED
Binary files a/neuronxcc-2.21.33363.0+82129205/MODULE_10661660426924300837+e30acd3a/model.neff and b/neuronxcc-2.21.33363.0+82129205/MODULE_10661660426924300837+e30acd3a/model.neff differ
 
neuronxcc-2.21.33363.0+82129205/MODULE_10746122569655005679+e30acd3a/model.neff CHANGED
Binary files a/neuronxcc-2.21.33363.0+82129205/MODULE_10746122569655005679+e30acd3a/model.neff and b/neuronxcc-2.21.33363.0+82129205/MODULE_10746122569655005679+e30acd3a/model.neff differ
 
neuronxcc-2.21.33363.0+82129205/MODULE_11086318750207148626+e30acd3a/model.neff CHANGED
Binary files a/neuronxcc-2.21.33363.0+82129205/MODULE_11086318750207148626+e30acd3a/model.neff and b/neuronxcc-2.21.33363.0+82129205/MODULE_11086318750207148626+e30acd3a/model.neff differ
 
neuronxcc-2.21.33363.0+82129205/MODULE_11301370888592189606+e30acd3a/compile_flags.json ADDED
@@ -0,0 +1 @@
 
 
1
+ ["--target=trn1"]