Add Sentence Transformers retrieval validation on Trainium2

README.md

@@ -5,6 +5,7 @@ tags:
 - neuron
 - trainium
 - transformers
+- sentence-transformers
 configs:
 - config_name: default
   data_files:
@@ -17,8 +18,9 @@ configs:
 Which Hugging Face Hub models run unmodified on AWS Trainium, through the
 native PyTorch backend
 [`torch-neuronx`](https://github.com/aws-neuron/torch-neuronx), for each
-`neuronx-cc` and library version tested. The only library tested so far is
-[`transformers`](https://github.com/huggingface/transformers).
+`neuronx-cc` and library version tested. Tested libraries are
+[`transformers`](https://github.com/huggingface/transformers) and
+[`sentence-transformers`](https://github.com/huggingface/sentence-transformers).
 
 To look up a model, use the
 [Space](https://huggingface.co/spaces/aws-neuron/torch-neuronx-model-support).
@@ -101,6 +103,68 @@ run 100 models, skipping those whose test script has not changed since their
 last run. Quantized checkpoints and models that need `trust_remote_code` are
 left out. A model that leaves the list keeps the results it already has.
 
+## sentence-transformers
+
+The 2026-11-03 sweep tests the top 10 Hub models tagged `sentence-transformers`,
+sorted by trending score, plus the highest-ranked MultiVectorEncoder at rank 15.
+The selection and checkpoint revisions are frozen in
+`validation/sentence-transformers/2026-11-03/trending.json`. No failed candidate
+was replaced with a passing model.
+
+SentenceTransformer uses `encode_query`, `encode_document`, and `similarity`;
+CrossEncoder uses `predict` on query/document pairs; MultiVectorEncoder uses
+query/document encoding followed by native MaxSim. Saved retrieval prompts are
+applied. All three paths produce four scores on `neuron:0`, with batch size 4:
+
+```python
+query = "Which planet is known as the Red Planet?"
+documents = [
+    "Venus is often called Earth's twin because of its similar size and proximity.",
+    "Mars, known for its reddish appearance, is often referred to as the Red Planet.",
+    "Jupiter, the largest planet in our solar system, has a prominent red spot.",
+    "Saturn, famous for its rings, is sometimes mistaken for the Red Planet.",
+]
+```
+
+`PASS` requires four finite scores, Mars ranked first, and CPU score agreement
+using `atol=rtol=0.001` at the same precision. `FAIL(output)` means the output
+check failed, even when inference completed. Every completed run is repeated;
+all repeats produced identical scores. Scores are in original document order
+and are not comparable between models.
+
+The sweep uses native eager PyTorch with SDPA, without an explicit
+`torch.compile` wrapper, on one logical Neuron device of a Trainium2 instance.
+FP32 is used except for `nvidia/Nemotron-3.5-Rerank-8B-BF16`: allocating its FP32
+weights exhausted device memory. Its BF16 retry ran and ranked Mars first, but
+the maximum CPU score difference was 0.008537, so it is `FAIL(output)` under the
+strict parity criterion. The other ten records pass, including the supplementary
+MultiVectorEncoder.
+
+`google/embeddinggemma-2` uses its documented text-only configuration
+(`vision_config=None`, `audio_config=None`); full multimodal loading with SDPA
+failed on the audio encoder. It required an upgrade from Transformers 5.17.0.
+The MiniLM run used 5.17.0; the other final runs used 5.22.1. Kanana uses its
+required `trust_remote_code=True` at the recorded revision. Perplexity uses the
+reference PyTorch gated-attention implementation without `flash-linear-attention`.
+Sentence Transformers is pinned to commit
+`6be0abf38d00d70bb3c732131df59ad073ff3768` (6.3.0.dev0).
+
+This validates one text retrieval example, not general retrieval quality,
+throughput, training, image/audio inputs, tensor parallelism, or multi-device
+execution. SparseEncoder was not tested. HBM measurements were not collected,
+so the memory columns are null. Wall time includes model loading, CPU reference,
+first Neuron execution/compilation, and the repeat; it is not inference latency.
+Start timestamps are reconstructed from saved log timestamps and measured run
+durations. `runner` names the encoder class; `neuron_cores=1` describes the single
+logical device used, not a tensor-parallel degree.
+
+Full scores, versions, configuration, and earlier unsuccessful attempts are in
+`validation/sentence-transformers/2026-11-03/validation_results.json`, with final
+scores also available as CSV. Each row includes a standalone reproducer derived
+from the validation harness. These scripts write `result.json` and exit nonzero
+if retrieval or CPU parity fails. No compiled cache from this sweep has been
+published.
+
 ## Layout
 
 There is one file per Neuron target (`trn2`, `trn3`) and tested

validation/sentence-transformers/2026-11-03/validation_results.csv

@@ -0,0 +1,12 @@
+rank,model,class,dtype,status,venus,mars,jupiter,saturn,mars_first,max_abs_error,parity_atol_1e-3_rtol_1e-3,repeat_max_abs_error,revision
+1,google/embeddinggemma-2,SentenceTransformer,float32,executed,0.6754712462425232,0.8531948328018188,0.7467760443687439,0.7782547473907471,True,1.3709068298339844e-06,True,0.0,f9972ac22e696a105c482202646979e33e4b5a1e
+2,sentence-transformers/all-MiniLM-L6-v2,SentenceTransformer,float32,executed,0.46468585729599,0.8114601373672485,0.7279221415519714,0.7501919269561768,True,7.748603820800781e-07,True,0.0,1110a243fdf4706b3f48f1d95db1a4f5529b4d41
+3,BAAI/bge-m3,SentenceTransformer,float32,executed,0.4096478521823883,0.7112398743629456,0.648069441318512,0.6503464579582214,True,5.960464477539062e-07,True,0.0,5617a9f61b028005a4858fdac845db406aefb181
+4,kakaocorp/kanana-2-3b-embedding,SentenceTransformer,float32,executed,0.48793739080429077,0.8206104636192322,0.6573861241340637,0.6299965381622314,True,1.2516975402832031e-06,True,0.0,2c85dbe61af8ff033d2c9d46a686c665153069aa
+5,google/embeddinggemma-300m,SentenceTransformer,float32,executed,0.3008213937282562,0.636100172996521,0.49270257353782654,0.48888036608695984,True,5.066394805908203e-07,True,0.0,57c266a740f537b4dc058e1b0cda161fd15afa75
+6,nvidia/Nemotron-3.5-Rerank-8B-BF16,CrossEncoder,bfloat16,executed,-2.700716495513916,-1.0203332901000977,-2.303680419921875,-2.0575268268585205,True,0.008536696434020996,False,0.0,f88f8d9104726747791283ab156d2fcdb29e90a7
+7,Qwen/Qwen3-Embedding-0.6B,SentenceTransformer,float32,executed,0.48112940788269043,0.6901392936706543,0.5837673544883728,0.6643365621566772,True,5.364418029785156e-07,True,0.0,97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3
+8,nvidia/Nemotron-3.5-Embed-1B-BF16,SentenceTransformer,float32,executed,0.24803557991981506,0.6448327302932739,0.4405462443828583,0.4166099727153778,True,5.364418029785156e-07,True,0.0,e24ca1e346bcc53572bd7bdc02b3e8ebf66c197c
+9,BAAI/bge-reranker-v2-m3,CrossEncoder,float32,executed,0.019280483946204185,0.9986975193023682,0.785732090473175,0.98695969581604,True,7.748603820800781e-07,True,0.0,953dc6f6f85a1b2dbfca4c34a2796e7dde08d41e
+10,Qwen/Qwen3-VL-Embedding-2B,SentenceTransformer,float32,executed,0.5317150354385376,0.7854103446006775,0.6537114381790161,0.6054514050483704,True,1.0132789611816406e-06,True,0.0,9f2f7e710d6d81056aa5c0a4f04764fec6bb7bda
+15,perplexity-ai/pplx-embed-v2-late-0.6b,MultiVectorEncoder,float32,executed,8.452167510986328,8.74061393737793,8.631291389465332,8.708816528320312,True,1.9073486328125e-06,True,0.0,7bba8e7cf2075dec747fa8557f41eb90e70361ed

validation/sentence-transformers/2026-11-03/validation_results.json

@@ -0,0 +1,728 @@
+{
+  "validated_at_utc": "2026-11-03T11:26:53.448792+00:00",
+  "hub_listing": "https://huggingface.co/models?library=sentence-transformers",
+  "selection": "Top 10 by trendingScore, plus highest-ranked MultiVectorEncoder at rank 15. No replacement of failing entries.",
+  "repository_commit": "6be0abf38d00d70bb3c732131df59ad073ff3768",
+  "device": "neuron:0 on Trainium2",
+  "torch_neuronx": "2.14.3.0.2115+149b6fca1",
+  "attention": "sdpa",
+  "cpu_parity_tolerance": {
+    "atol": 0.001,
+    "rtol": 0.001
+  },
+  "notes": [
+    "Text-only retrieval with the exact user-provided query and four documents. Scores are in original document order: Venus, Mars, Jupiter, Saturn. Scores are not comparable between models.",
+    "FP32 except the 8B Nemotron reranker, which exceeded device memory in FP32 and was retried in BF16.",
+    "EmbeddingGemma 2 required current Transformers and its documented text-only config: vision_config=None, audio_config=None. Full multimodal SDPA loading failed on its audio component.",
+    "MiniLM used Transformers 5.17.0; other final runs used 5.22.1 in an isolated environment. All other package versions and model revisions are recorded per row.",
+    "Saved retrieval prompts were applied via encode_query/encode_document; Kanana used its required trusted remote code at the recorded revision.",
+    "Perplexity uses the reference PyTorch gated-attention path, without optional flash-linear-attention.",
+    "These are correctness checks for one retrieval example, not retrieval benchmarks or throughput measurements. No sparse models, training, tensor parallelism, image/audio paths, or multi-device inference were tested in this sweep."
+  ],
+  "results": [
+    {
+      "rank": 1,
+      "model": "google/embeddinggemma-2",
+      "revision": "f9972ac22e696a105c482202646979e33e4b5a1e",
+      "class": "SentenceTransformer",
+      "dtype": "float32",
+      "attention": "sdpa",
+      "batch_size": 4,
+      "query": "Which planet is known as the Red Planet?",
+      "documents": [
+        "Venus is often called Earth's twin because of its similar size and proximity.",
+        "Mars, known for its reddish appearance, is often referred to as the Red Planet.",
+        "Jupiter, the largest planet in our solar system, has a prominent red spot.",
+        "Saturn, famous for its rings, is sometimes mistaken for the Red Planet."
+      ],
+      "torch": "2.14.0+cpu",
+      "transformers": "5.22.1",
+      "sentence_transformers": "6.3.0.dev0",
+      "python": "3.14.4",
+      "config_kwargs": {
+        "vision_config": null,
+        "audio_config": null
+      },
+      "prompts": {
+        "query": "task: search result | query: ",
+        "document": "title: none | text: ",
+        "BitextMining": "task: search result | query: ",
+        "Classification": "task: classification | query: ",
+        "Clustering": "task: clustering | query: ",
+        "CodeRetrieval": "task: code retrieval | query: ",
+        "Document": "title: none | text: ",
+        "FactChecking": "task: fact checking | query: ",
+        "InstructionRetrieval": "task: code retrieval | query: ",
+        "MultilabelClassification": "task: classification | query: ",
+        "PairClassification": "task: sentence similarity | query: ",
+        "QuestionAnswering": "task: question answering | query: ",
+        "Reranking": "task: search result | query: ",
+        "Retrieval": "task: search result | query: ",
+        "Retrieval-document": "title: none | text: ",
+        "Retrieval-query": "task: search result | query: ",
+        "STS": "task: sentence similarity | query: ",
+        "SearchQuery": "task: search result | query: ",
+        "SentenceSimilarity": "task: sentence similarity | query: ",
+        "Summarization": "task: sentence similarity | query: "
+      },
+      "load_seconds": 2.764061794616282,
+      "embedding_devices": [
+        "neuron:0",
+        "neuron:0"
+      ],
+      "cpu_seconds": 0.20980793330818415,
+      "cpu_scores": [
+        0.6754717230796814,
+        0.8531962037086487,
+        0.7467769980430603,
+        0.778255820274353
+      ],
+      "model_device": "neuron:0",
+      "score_device": "neuron:0",
+      "neuron_first_seconds": 93.4308730987832,
+      "neuron_scores": [
+        0.6754712462425232,
+        0.8531948328018188,
+        0.7467760443687439,
+        0.7782547473907471
+      ],
+      "max_abs_error": 1.3709068298339844e-06,
+      "cpu_top_document": 2,
+      "neuron_top_document": 2,
+      "mars_first": true,
+      "parity_atol_1e-3_rtol_1e-3": true,
+      "neuron_repeat_seconds": 0.27350239269435406,
+      "repeat_max_abs_error": 0.0,
+      "status": "executed",
+      "total_seconds": 107.6439163023606
+    },
+    {
+      "rank": 2,
+      "model": "sentence-transformers/all-MiniLM-L6-v2",
+      "revision": "1110a243fdf4706b3f48f1d95db1a4f5529b4d41",
+      "class": "SentenceTransformer",
+      "dtype": "float32",
+      "attention": "sdpa",
+      "batch_size": 4,
+      "query": "Which planet is known as the Red Planet?",
+      "documents": [
+        "Venus is often called Earth's twin because of its similar size and proximity.",
+        "Mars, known for its reddish appearance, is often referred to as the Red Planet.",
+        "Jupiter, the largest planet in our solar system, has a prominent red spot.",
+        "Saturn, famous for its rings, is sometimes mistaken for the Red Planet."
+      ],
+      "torch": "2.14.0+cpu",
+      "transformers": "5.17.0",
+      "sentence_transformers": "6.3.0.dev0",
+      "python": "3.14.4",
+      "prompts": {
+        "query": "",
+        "document": ""
+      },
+      "load_seconds": 3.67690596729517,
+      "embedding_devices": [
+        "neuron:0",
+        "neuron:0"
+      ],
+      "cpu_seconds": 1.0060012973845005,
+      "cpu_scores": [
+        0.4646856188774109,
+        0.8114608526229858,
+        0.7279220223426819,
+        0.7501927018165588
+      ],
+      "model_device": "neuron:0",
+      "score_device": "neuron:0",
+      "neuron_first_seconds": 62.592925063334405,
+      "neuron_scores": [
+        0.46468585729599,
+        0.8114601373672485,
+        0.7279221415519714,
+        0.7501919269561768
+      ],
+      "max_abs_error": 7.748603820800781e-07,
+      "cpu_top_document": 2,
+      "neuron_top_document": 2,
+      "mars_first": true,
+      "parity_atol_1e-3_rtol_1e-3": true,
+      "neuron_repeat_seconds": 0.03206939063966274,
+      "repeat_max_abs_error": 0.0,
+      "status": "executed",
+      "total_seconds": 79.70192950125784
+    },
+    {
+      "rank": 3,
+      "model": "BAAI/bge-m3",
+      "revision": "5617a9f61b028005a4858fdac845db406aefb181",
+      "class": "SentenceTransformer",
+      "dtype": "float32",
+      "attention": "sdpa",
+      "batch_size": 4,
+      "query": "Which planet is known as the Red Planet?",
+      "documents": [
+        "Venus is often called Earth's twin because of its similar size and proximity.",
+        "Mars, known for its reddish appearance, is often referred to as the Red Planet.",
+        "Jupiter, the largest planet in our solar system, has a prominent red spot.",
+        "Saturn, famous for its rings, is sometimes mistaken for the Red Planet."
+      ],
+      "torch": "2.14.0+cpu",
+      "transformers": "5.22.1",
+      "sentence_transformers": "6.3.0.dev0",
+      "python": "3.14.4",
+      "prompts": {
+        "query": "",
+        "document": ""
+      },
+      "load_seconds": 16.82487368490547,
+      "embedding_devices": [
+        "neuron:0",
+        "neuron:0"
+      ],
+      "cpu_seconds": 4.732044198550284,
+      "cpu_scores": [
+        0.4096481204032898,
+        0.7112404704093933,
+        0.6480699777603149,
+        0.6503466963768005
+      ],
+      "model_device": "neuron:0",
+      "score_device": "neuron:0",
+      "neuron_first_seconds": 100.87037728819996,
+      "neuron_scores": [
+        0.4096478521823883,
+        0.7112398743629456,
+        0.648069441318512,
+        0.6503464579582214
+      ],
+      "max_abs_error": 5.960464477539062e-07,
+      "cpu_top_document": 2,
+      "neuron_top_document": 2,
+      "mars_first": true,
+      "parity_atol_1e-3_rtol_1e-3": true,
+      "neuron_repeat_seconds": 0.06637111864984035,
+      "repeat_max_abs_error": 0.0,
+      "status": "executed",
+      "total_seconds": 135.9673376828432
+    },
+    {
+      "rank": 4,
+      "model": "kakaocorp/kanana-2-3b-embedding",
+      "revision": "2c85dbe61af8ff033d2c9d46a686c665153069aa",
+      "class": "SentenceTransformer",
+      "dtype": "float32",
+      "attention": "sdpa",
+      "batch_size": 4,
+      "query": "Which planet is known as the Red Planet?",
+      "documents": [
+        "Venus is often called Earth's twin because of its similar size and proximity.",
+        "Mars, known for its reddish appearance, is often referred to as the Red Planet.",
+        "Jupiter, the largest planet in our solar system, has a prominent red spot.",
+        "Saturn, famous for its rings, is sometimes mistaken for the Red Planet."
+      ],
+      "torch": "2.14.0+cpu",
+      "transformers": "5.22.1",
+      "sentence_transformers": "6.3.0.dev0",
+      "python": "3.14.4",
+      "prompts": {
+        "query": "",
+        "document": ""
+      },
+      "load_seconds": 9.024477198719978,
+      "embedding_devices": [
+        "neuron:0",
+        "neuron:0"
+      ],
+      "cpu_seconds": 2.309540878981352,
+      "cpu_scores": [
+        0.4879377782344818,
+        0.8206117153167725,
+        0.6573871970176697,
+        0.629997193813324
+      ],
+      "model_device": "neuron:0",
+      "score_device": "neuron:0",
+      "neuron_first_seconds": 57.551774497143924,
+      "neuron_scores": [
+        0.48793739080429077,
+        0.8206104636192322,
+        0.6573861241340637,
+        0.6299965381622314
+      ],
+      "max_abs_error": 1.2516975402832031e-06,
+      "cpu_top_document": 2,
+      "neuron_top_document": 2,
+      "mars_first": true,
+      "parity_atol_1e-3_rtol_1e-3": true,
+      "neuron_repeat_seconds": 0.30301052797585726,
+      "repeat_max_abs_error": 0.0,
+      "status": "executed",
+      "total_seconds": 83.3380340654403
+    },
+    {
+      "rank": 5,
+      "model": "google/embeddinggemma-300m",
+      "revision": "57c266a740f537b4dc058e1b0cda161fd15afa75",
+      "class": "SentenceTransformer",
+      "dtype": "float32",
+      "attention": "sdpa",
+      "batch_size": 4,
+      "query": "Which planet is known as the Red Planet?",
+      "documents": [
+        "Venus is often called Earth's twin because of its similar size and proximity.",
+        "Mars, known for its reddish appearance, is often referred to as the Red Planet.",
+        "Jupiter, the largest planet in our solar system, has a prominent red spot.",
+        "Saturn, famous for its rings, is sometimes mistaken for the Red Planet."
+      ],
+      "torch": "2.14.0+cpu",
+      "transformers": "5.22.1",
+      "sentence_transformers": "6.3.0.dev0",
+      "python": "3.14.4",
+      "prompts": {
+        "query": "task: search result | query: ",
+        "document": "title: none | text: ",
+        "BitextMining": "task: search result | query: ",
+        "Clustering": "task: clustering | query: ",
+        "Classification": "task: classification | query: ",
+        "InstructionRetrieval": "task: code retrieval | query: ",
+        "MultilabelClassification": "task: classification | query: ",
+        "PairClassification": "task: sentence similarity | query: ",
+        "Reranking": "task: search result | query: ",
+        "Retrieval": "task: search result | query: ",
+        "Retrieval-query": "task: search result | query: ",
+        "Retrieval-document": "title: none | text: ",
+        "STS": "task: sentence similarity | query: ",
+        "Summarization": "task: summarization | query: "
+      },
+      "load_seconds": 2.9100955119356513,
+      "embedding_devices": [
+        "neuron:0",
+        "neuron:0"
+      ],
+      "cpu_seconds": 0.3329894030466676,
+      "cpu_scores": [
+        0.3008216619491577,
+        0.6361004710197449,
+        0.49270308017730713,
+        0.48888057470321655
+      ],
+      "model_device": "neuron:0",
+      "score_device": "neuron:0",
+      "neuron_first_seconds": 73.40440620388836,
+      "neuron_scores": [
+        0.3008213937282562,
+        0.636100172996521,
+        0.49270257353782654,
+        0.48888036608695984
+      ],
+      "max_abs_error": 5.066394805908203e-07,
+      "cpu_top_document": 2,
+      "neuron_top_document": 2,
+      "mars_first": true,
+      "parity_atol_1e-3_rtol_1e-3": true,
+      "neuron_repeat_seconds": 0.2501273164525628,
+      "repeat_max_abs_error": 0.0,
+      "status": "executed",
+      "total_seconds": 88.49807658046484
+    },
+    {
+      "rank": 6,
+      "model": "nvidia/Nemotron-3.5-Rerank-8B-BF16",
+      "revision": "f88f8d9104726747791283ab156d2fcdb29e90a7",
+      "class": "CrossEncoder",
+      "dtype": "bfloat16",
+      "attention": "sdpa",
+      "batch_size": 4,
+      "query": "Which planet is known as the Red Planet?",
+      "documents": [
+        "Venus is often called Earth's twin because of its similar size and proximity.",
+        "Mars, known for its reddish appearance, is often referred to as the Red Planet.",
+        "Jupiter, the largest planet in our solar system, has a prominent red spot.",
+        "Saturn, famous for its rings, is sometimes mistaken for the Red Planet."
+      ],
+      "torch": "2.14.0+cpu",
+      "transformers": "5.22.1",
+      "sentence_transformers": "6.3.0.dev0",
+      "python": "3.14.4",
+      "prompts": {},
+      "load_seconds": 1.2430504048243165,
+      "cpu_seconds": 1.7110898373648524,
+      "cpu_scores": [
+        -2.699960470199585,
+        -1.0288699865341187,
+        -2.2972545623779297,
+        -2.0584821701049805
+      ],
+      "model_device": "neuron:0",
+      "score_device": "neuron:0",
+      "neuron_first_seconds": 23.328446314670146,
+      "neuron_scores": [
+        -2.700716495513916,
+        -1.0203332901000977,
+        -2.303680419921875,
+        -2.0575268268585205
+      ],
+      "max_abs_error": 0.008536696434020996,
+      "cpu_top_document": 2,
+      "neuron_top_document": 2,
+      "mars_first": true,
+      "parity_atol_1e-3_rtol_1e-3": false,
+      "neuron_repeat_seconds": 0.16023080982267857,
+      "repeat_max_abs_error": 0.0,
+      "status": "executed",
+      "total_seconds": 40.27650658786297
+    },
+    {
+      "rank": 7,
+      "model": "Qwen/Qwen3-Embedding-0.6B",
+      "revision": "97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3",
+      "class": "SentenceTransformer",
+      "dtype": "float32",
+      "attention": "sdpa",
+      "batch_size": 4,
+      "query": "Which planet is known as the Red Planet?",
+      "documents": [
+        "Venus is often called Earth's twin because of its similar size and proximity.",
+        "Mars, known for its reddish appearance, is often referred to as the Red Planet.",
+        "Jupiter, the largest planet in our solar system, has a prominent red spot.",
+        "Saturn, famous for its rings, is sometimes mistaken for the Red Planet."
+      ],
+      "torch": "2.14.0+cpu",
+      "transformers": "5.22.1",
+      "sentence_transformers": "6.3.0.dev0",
+      "python": "3.14.4",
+      "prompts": {
+        "query": "Instruct: Given a web search query, retrieve relevant passages that answer the query\nQuery:",
+        "document": ""
+      },
+      "load_seconds": 1.1713379649445415,
+      "embedding_devices": [
+        "neuron:0",
+        "neuron:0"
+      ],
+      "cpu_seconds": 0.4523088913410902,
+      "cpu_scores": [
+        0.4811299443244934,
+        0.6901398301124573,
+        0.5837672352790833,
+        0.6643370985984802
+      ],
+      "model_device": "neuron:0",
+      "score_device": "neuron:0",
+      "neuron_first_seconds": 59.35582494363189,
+      "neuron_scores": [
+        0.48112940788269043,
+        0.6901392936706543,
+        0.5837673544883728,
+        0.6643365621566772
+      ],
+      "max_abs_error": 5.364418029785156e-07,
+      "cpu_top_document": 2,
+      "neuron_top_document": 2,
+      "mars_first": true,
+      "parity_atol_1e-3_rtol_1e-3": true,
+      "neuron_repeat_seconds": 0.2696740850806236,
+      "repeat_max_abs_error": 0.0,
+      "status": "executed",
+      "total_seconds": 72.76618314534426
+    },
+    {
+      "rank": 8,
+      "model": "nvidia/Nemotron-3.5-Embed-1B-BF16",
+      "revision": "e24ca1e346bcc53572bd7bdc02b3e8ebf66c197c",
+      "class": "SentenceTransformer",
+      "dtype": "float32",
+      "attention": "sdpa",
+      "batch_size": 4,
+      "query": "Which planet is known as the Red Planet?",
+      "documents": [
+        "Venus is often called Earth's twin because of its similar size and proximity.",
+        "Mars, known for its reddish appearance, is often referred to as the Red Planet.",
+        "Jupiter, the largest planet in our solar system, has a prominent red spot.",
+        "Saturn, famous for its rings, is sometimes mistaken for the Red Planet."
+      ],
+      "torch": "2.14.0+cpu",
+      "transformers": "5.22.1",
+      "sentence_transformers": "6.3.0.dev0",
+      "python": "3.14.4",
+      "prompts": {
+        "query": "query: ",
+        "document": "passage: "
+      },
+      "load_seconds": 5.307680940255523,
+      "embedding_devices": [
+        "neuron:0",
+        "neuron:0"
+      ],
+      "cpu_seconds": 0.7383365361019969,
+      "cpu_scores": [
+        0.24803577363491058,
+        0.6448332667350769,
+        0.44054660201072693,
+        0.4166104197502136
+      ],
+      "model_device": "neuron:0",
+      "score_device": "neuron:0",
+      "neuron_first_seconds": 55.88916517701,
+      "neuron_scores": [
+        0.24803557991981506,
+        0.6448327302932739,
+        0.4405462443828583,
+        0.4166099727153778
+      ],
+      "max_abs_error": 5.364418029785156e-07,
+      "cpu_top_document": 2,
+      "neuron_top_document": 2,
+      "mars_first": true,
+      "parity_atol_1e-3_rtol_1e-3": true,
+      "neuron_repeat_seconds": 0.1378226550295949,
+      "repeat_max_abs_error": 0.0,
+      "status": "executed",
+      "total_seconds": 74.03900135494769
+    },
+    {
+      "rank": 9,
+      "model": "BAAI/bge-reranker-v2-m3",
+      "revision": "953dc6f6f85a1b2dbfca4c34a2796e7dde08d41e",
+      "class": "CrossEncoder",
+      "dtype": "float32",
+      "attention": "sdpa",
+      "batch_size": 4,
+      "query": "Which planet is known as the Red Planet?",
+      "documents": [
+        "Venus is often called Earth's twin because of its similar size and proximity.",
+        "Mars, known for its reddish appearance, is often referred to as the Red Planet.",
+        "Jupiter, the largest planet in our solar system, has a prominent red spot.",
+        "Saturn, famous for its rings, is sometimes mistaken for the Red Planet."
+      ],
+      "torch": "2.14.0+cpu",
+      "transformers": "5.22.1",
+      "sentence_transformers": "6.3.0.dev0",
+      "python": "3.14.4",
+      "prompts": {},
+      "load_seconds": 3.388627864420414,
+      "cpu_seconds": 2.5045143608003855,
+      "cpu_scores": [
+        0.019280454143881798,
+        0.9986975193023682,
+        0.785731315612793,
+        0.9869596362113953
+      ],
+      "model_device": "neuron:0",
+      "score_device": "neuron:0",
+      "neuron_first_seconds": 29.88394799735397,
+      "neuron_scores": [
+        0.019280483946204185,
+        0.9986975193023682,
+        0.785732090473175,
+        0.98695969581604
+      ],
+      "max_abs_error": 7.748603820800781e-07,
+      "cpu_top_document": 2,
+      "neuron_top_document": 2,
+      "mars_first": true,
+      "parity_atol_1e-3_rtol_1e-3": true,
+      "neuron_repeat_seconds": 0.036954318173229694,
+      "repeat_max_abs_error": 0.0,
+      "status": "executed",
+      "total_seconds": 47.27732175774872
+    },
+    {
+      "rank": 10,
+      "model": "Qwen/Qwen3-VL-Embedding-2B",
+      "revision": "9f2f7e710d6d81056aa5c0a4f04764fec6bb7bda",
+      "class": "SentenceTransformer",
+      "dtype": "float32",
+      "attention": "sdpa",
+      "batch_size": 4,
+      "query": "Which planet is known as the Red Planet?",
+      "documents": [
+        "Venus is often called Earth's twin because of its similar size and proximity.",
+        "Mars, known for its reddish appearance, is often referred to as the Red Planet.",
+        "Jupiter, the largest planet in our solar system, has a prominent red spot.",
+        "Saturn, famous for its rings, is sometimes mistaken for the Red Planet."
+      ],
+      "torch": "2.14.0+cpu",
+      "transformers": "5.22.1",
+      "sentence_transformers": "6.3.0.dev0",
+      "python": "3.14.4",
+      "prompts": {
+        "query": "",
+        "document": "",
+        "default": "Represent the user's input."
+      },
+      "load_seconds": 6.661330422386527,
+      "embedding_devices": [
+        "neuron:0",
+        "neuron:0"
+      ],
+      "cpu_seconds": 1.6249698027968407,
+      "cpu_scores": [
+        0.531715452671051,
+        0.7854113578796387,
+        0.6537123918533325,
+        0.6054521799087524
+      ],
+      "model_device": "neuron:0",
+      "score_device": "neuron:0",
+      "neuron_first_seconds": 49.71057551726699,
+      "neuron_scores": [
+        0.5317150354385376,
+        0.7854103446006775,
+        0.6537114381790161,
+        0.6054514050483704
+      ],
+      "max_abs_error": 1.0132789611816406e-06,
+      "cpu_top_document": 2,
+      "neuron_top_document": 2,
+      "mars_first": true,
+      "parity_atol_1e-3_rtol_1e-3": true,
+      "neuron_repeat_seconds": 0.310243034735322,
+      "repeat_max_abs_error": 0.0,
+      "status": "executed",
+      "total_seconds": 71.60194528196007
+    },
+    {
+      "rank": 15,
+      "model": "perplexity-ai/pplx-embed-v2-late-0.6b",
+      "revision": "7bba8e7cf2075dec747fa8557f41eb90e70361ed",
+      "class": "MultiVectorEncoder",
+      "dtype": "float32",
+      "attention": "sdpa",
+      "batch_size": 4,
+      "query": "Which planet is known as the Red Planet?",
+      "documents": [
+        "Venus is often called Earth's twin because of its similar size and proximity.",
+        "Mars, known for its reddish appearance, is often referred to as the Red Planet.",
+        "Jupiter, the largest planet in our solar system, has a prominent red spot.",
+        "Saturn, famous for its rings, is sometimes mistaken for the Red Planet."
+      ],
+      "torch": "2.14.0+cpu",
+      "transformers": "5.22.1",
+      "sentence_transformers": "6.3.0.dev0",
+      "python": "3.14.4",
+      "prompts": {
+        "query": "[Q] ",
+        "document": "[D] "
+      },
+      "load_seconds": 3.263010887429118,
+      "embedding_devices": [
+        "neuron:0",
+        "neuron:0"
+      ],
+      "cpu_seconds": 0.4739787392318249,
+      "cpu_scores": [
+        8.452167510986328,
+        8.74061393737793,
+        8.631292343139648,
+        8.70881462097168
+      ],
+      "model_device": "neuron:0",
+      "score_device": "neuron:0",
+      "neuron_first_seconds": 116.29114723391831,
+      "neuron_scores": [
+        8.452167510986328,
+        8.74061393737793,
+        8.631291389465332,
+        8.708816528320312
+      ],
+      "max_abs_error": 1.9073486328125e-06,
+      "cpu_top_document": 2,
+      "neuron_top_document": 2,
+      "mars_first": true,
+      "parity_atol_1e-3_rtol_1e-3": true,
+      "neuron_repeat_seconds": 0.3539726445451379,
+      "repeat_max_abs_error": 0.0,
+      "status": "executed",
+      "total_seconds": 133.92629071511328
+    }
+  ],
+  "prior_attempts": [
+    {
+      "rank": 1,
+      "model": "google/embeddinggemma-2",
+      "revision": "f9972ac22e696a105c482202646979e33e4b5a1e",
+      "class": "SentenceTransformer",
+      "dtype": "float32",
+      "attention": "sdpa",
+      "batch_size": 4,
+      "query": "Which planet is known as the Red Planet?",
+      "documents": [
+        "Venus is often called Earth's twin because of its similar size and proximity.",
+        "Mars, known for its reddish appearance, is often referred to as the Red Planet.",
+        "Jupiter, the largest planet in our solar system, has a prominent red spot.",
+        "Saturn, famous for its rings, is sometimes mistaken for the Red Planet."
+      ],
+      "torch": "2.14.0+cpu",
+      "transformers": "5.17.0",
+      "sentence_transformers": "6.3.0.dev0",
+      "python": "3.14.4",
+      "status": "error",
+      "stage": "load_cpu",
+      "error_type": "ValueError",
+      "error": "The checkpoint you are trying to load has model type `embedding_gemma2` but Transformers does not recognize this architecture. This could be because of an issue with the checkpoint, or because your version of Transformers is out of date.\n\nYou can update Transformers with the command `pip install --upgrade transformers`. If this does not work, and the checkpoint is very new, then there may not be a release version that supports this model yet. In this case, you can get the most up-to-date code by installing Transformers from source with the command `pip install git+https://github.com/huggingface/transformers.git`",
+      "traceback": "Traceback (most recent call last):\n  File \"/home/jingyahuang/.venv-oct09/lib/python3.14/site-packages/transformers/models/auto/configuration_auto.py\", line 419, in from_pretrained\n    config_class = CONFIG_MAPPING[config_dict[\"model_type\"]]\n                   ~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^\n  File \"/home/jingyahuang/.venv-oct09/lib/python3.14/site-packages/transformers/models/auto/configuration_auto.py\", line 107, in __getitem__\n    raise KeyError(key)\nKeyError: 'embedding_gemma2'\n\nDuring handling of the above exception, another exception occurred:\n\nTraceback (most recent call last):\n  File \"/tmp/neuron-top10-sweep/run_model.py\", line 19, in <module>\n    model=cls(name,revision=meta['sha'],device='cpu',trust_remote_code=name.startswith('kakaocorp/'),model_kwargs={'torch_dtype':torch.float32,'attn_implementation':'sdpa'})\n  File \"/home/jingyahuang/sentence-transformers/sentence_transformers/util/decorators.py\", line 46, in wrapper\n    return func(*args, **kwargs)\n  File \"/home/jingyahuang/sentence-transformers/sentence_transformers/sentence_transformer/model.py\", line 193, in __init__\n    super().__init__(\n    ~~~~~~~~~~~~~~~~^\n        model_name_or_path=model_name_or_path,\n        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n    ...<13 lines>...\n        default_prompt_name=default_prompt_name,\n        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n    )\n    ^\n  File \"/home/jingyahuang/sentence-transformers/sentence_transformers/base/model.py\", line 256, in __init__\n    modules, self.module_kwargs = self._load_modules(\n                                  ~~~~~~~~~~~~~~~~~~^\n        model_name_or_path,\n        ^^^^^^^^^^^^^^^^^^^\n    ...<7 lines>...\n        config_kwargs=config_kwargs,\n        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n    )\n    ^\n  File \"/home/jingyahuang/sentence-transformers/sentence_transformers/base/model.py\", line 1211, in _load_modules\n    return self._load_config_modules(model_name_or_path, **load_kwargs)\n           ~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n  File \"/home/jingyahuang/sentence-transformers/sentence_transformers/base/model.py\", line 1431, in _load_config_modules\n    module = module_class.load(\n        model_name_or_path,\n    ...<12 lines>...\n        **extra_load_kwargs,\n    )\n  File \"/home/jingyahuang/sentence-transformers/sentence_transformers/base/modules/transformer.py\", line 2641, in load\n    return cls(model_name_or_path=model_name_or_path, **init_kwargs)\n  File \"/home/jingyahuang/sentence-transformers/sentence_transformers/util/decorators.py\", line 92, in wrapper\n    return func(*args, **kwargs)\n  File \"/home/jingyahuang/sentence-transformers/sentence_transformers/base/modules/transformer.py\", line 839, in __init__\n    config, is_peft_model = self._load_config(model_name_or_path, backend, config_kwargs)\n                            ~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n  File \"/home/jingyahuang/sentence-transformers/sentence_transformers/base/modules/transformer.py\", line 2259, in _load_config\n    return AutoConfig.from_pretrained(model_name_or_path, **config_kwargs), False\n           ~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n  File \"/home/jingyahuang/.venv-oct09/lib/python3.14/site-packages/transformers/models/auto/configuration_auto.py\", line 421, in from_pretrained\n    raise ValueError(\n    ...<8 lines>...\n    )\nValueError: The checkpoint you are trying to load has model type `embedding_gemma2` but Transformers does not recognize this architecture. This could be because of an issue with the checkpoint, or because your version of Transformers is out of date.\n\nYou can update Transformers with the command `pip install --upgrade transformers`. If this does not work, and the checkpoint is very new, then there may not be a release version that supports this model yet. In this case, you can get the most up-to-date code by installing Transformers from source with the command `pip install git+https://github.com/huggingface/transformers.git`\n",
+      "total_seconds": 0.93162942212075
+    },
+    {
+      "rank": 1,
+      "model": "google/embeddinggemma-2",
+      "revision": "f9972ac22e696a105c482202646979e33e4b5a1e",
+      "class": "SentenceTransformer",
+      "dtype": "float32",
+      "attention": "sdpa",
+      "batch_size": 4,
+      "query": "Which planet is known as the Red Planet?",
+      "documents": [
+        "Venus is often called Earth's twin because of its similar size and proximity.",
+        "Mars, known for its reddish appearance, is often referred to as the Red Planet.",
+        "Jupiter, the largest planet in our solar system, has a prominent red spot.",
+        "Saturn, famous for its rings, is sometimes mistaken for the Red Planet."
+      ],
+      "torch": "2.14.0+cpu",
+      "transformers": "5.22.1",
+      "sentence_transformers": "6.3.0.dev0",
+      "python": "3.14.4",
+      "status": "error",
+      "stage": "load_cpu",
+      "error_type": "ValueError",
+      "error": "Gemma4AudioModel does not support an attention implementation through torch.nn.functional.scaled_dot_product_attention yet. Please request the support for this architecture: https://github.com/huggingface/transformers/issues/28005. If you believe this error is a bug, please open an issue in Transformers GitHub repository and load your model with the argument `attn_implementation=\"eager\"` meanwhile. Example: `model = AutoModel.from_pretrained(\"openai/whisper-tiny\", attn_implementation=\"eager\")`",
+      "traceback": "Traceback (most recent call last):\n  File \"/tmp/neuron-top10-sweep/run_model.py\", line 20, in <module>\n    model=cls(name,revision=meta['sha'],device='cpu',trust_remote_code=name.startswith('kakaocorp/'),model_kwargs={'torch_dtype':dtype,'attn_implementation':'sdpa'})\n  File \"/home/jingyahuang/sentence-transformers/sentence_transformers/util/decorators.py\", line 46, in wrapper\n    return func(*args, **kwargs)\n  File \"/home/jingyahuang/sentence-transformers/sentence_transformers/sentence_transformer/model.py\", line 193, in __init__\n    super().__init__(\n    ~~~~~~~~~~~~~~~~^\n        model_name_or_path=model_name_or_path,\n        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n    ...<13 lines>...\n        default_prompt_name=default_prompt_name,\n        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n    )\n    ^\n  File \"/home/jingyahuang/sentence-transformers/sentence_transformers/base/model.py\", line 256, in __init__\n    modules, self.module_kwargs = self._load_modules(\n                                  ~~~~~~~~~~~~~~~~~~^\n        model_name_or_path,\n        ^^^^^^^^^^^^^^^^^^^\n    ...<7 lines>...\n        config_kwargs=config_kwargs,\n        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n    )\n    ^\n  File \"/home/jingyahuang/sentence-transformers/sentence_transformers/base/model.py\", line 1211, in _load_modules\n    return self._load_config_modules(model_name_or_path, **load_kwargs)\n           ~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n  File \"/home/jingyahuang/sentence-transformers/sentence_transformers/base/model.py\", line 1431, in _load_config_modules\n    module = module_class.load(\n        model_name_or_path,\n    ...<12 lines>...\n        **extra_load_kwargs,\n    )\n  File \"/home/jingyahuang/sentence-transformers/sentence_transformers/base/modules/transformer.py\", line 2641, in load\n    return cls(model_name_or_path=model_name_or_path, **init_kwargs)\n  File \"/home/jingyahuang/sentence-transformers/sentence_transformers/util/decorators.py\", line 92, in wrapper\n    return func(*args, **kwargs)\n  File \"/home/jingyahuang/sentence-transformers/sentence_transformers/base/modules/transformer.py\", line 855, in __init__\n    self.model = self._load_model(\n                 ~~~~~~~~~~~~~~~~^\n        model_name_or_path, transformer_task, config, backend, is_peft_model, **model_kwargs\n        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n    )\n    ^\n  File \"/home/jingyahuang/sentence-transformers/sentence_transformers/base/modules/transformer.py\", line 2343, in _load_model\n    return model_cls.from_pretrained(model_name_or_path, config=config, **model_kwargs)\n           ~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n  File \"/tmp/st-neuron-sweep/lib/python3.14/site-packages/transformers/models/auto/auto_factory.py\", line 406, in from_pretrained\n    return model_class.from_pretrained(\n           ~~~~~~~~~~~~~~~~~~~~~~~~~~~^\n        pretrained_model_name_or_path, *model_args, config=config, **hub_kwargs, **kwargs\n        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n    )\n    ^\n  File \"/tmp/st-neuron-sweep/lib/python3.14/site-packages/transformers/modeling_utils.py\", line 4300, in from_pretrained\n    model = cls(config, *model_args, **model_kwargs)\n  File \"/tmp/st-neuron-sweep/lib/python3.14/site-packages/transformers/models/embedding_gemma2/modeling_embedding_gemma2.py\", line 649, in __init__\n    self.audio_tower = AutoModel.from_config(config.audio_config) if config.audio_config is not None else None\n                       ~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^\n  File \"/tmp/st-neuron-sweep/lib/python3.14/site-packages/transformers/models/auto/auto_factory.py\", line 248, in from_config\n    return model_class._from_config(config, **kwargs)\n           ~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^\n  File \"/tmp/st-neuron-sweep/lib/python3.14/site-packages/transformers/modeling_utils.py\", line 1452, in _from_config\n    model = cls(config, **kwargs)\n  File \"/tmp/st-neuron-sweep/lib/python3.14/site-packages/transformers/models/gemma4/modeling_gemma4.py\", line 1907, in __init__\n    super().__init__(config)\n    ~~~~~~~~~~~~~~~~^^^^^^^^\n  File \"/tmp/st-neuron-sweep/lib/python3.14/site-packages/transformers/modeling_utils.py\", line 1258, in __init__\n    self.config._attn_implementation_internal = self._check_and_adjust_attn_implementation(\n                                                ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^\n        self.config._attn_implementation,\n        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n    ...<2 lines>...\n        allow_all_kernels=hub_kernels.ALLOW_ALL_KERNELS,\n        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n    )\n    ^\n  File \"/tmp/st-neuron-sweep/lib/python3.14/site-packages/transformers/modeling_utils.py\", line 1821, in _check_and_adjust_attn_implementation\n    applicable_attn_implementation = self.get_correct_attn_implementation(\n        applicable_attn_implementation, is_init_check\n    )\n  File \"/tmp/st-neuron-sweep/lib/python3.14/site-packages/transformers/modeling_utils.py\", line 1877, in get_correct_attn_implementation\n    raise e\n  File \"/tmp/st-neuron-sweep/lib/python3.14/site-packages/transformers/modeling_utils.py\", line 1874, in get_correct_attn_implementation\n    self._sdpa_can_dispatch(is_init_check)\n    ~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^\n  File \"/tmp/st-neuron-sweep/lib/python3.14/site-packages/transformers/modeling_utils.py\", line 1665, in _sdpa_can_dispatch\n    raise ValueError(\n    ...<3 lines>...\n    )\nValueError: Gemma4AudioModel does not support an attention implementation through torch.nn.functional.scaled_dot_product_attention yet. Please request the support for this architecture: https://github.com/huggingface/transformers/issues/28005. If you believe this error is a bug, please open an issue in Transformers GitHub repository and load your model with the argument `attn_implementation=\"eager\"` meanwhile. Example: `model = AutoModel.from_pretrained(\"openai/whisper-tiny\", attn_implementation=\"eager\")`\n",
+      "total_seconds": 0.16130424942821264
+    },
+    {
+      "rank": 6,
+      "model": "nvidia/Nemotron-3.5-Rerank-8B-BF16",
+      "revision": "f88f8d9104726747791283ab156d2fcdb29e90a7",
+      "class": "CrossEncoder",
+      "dtype": "float32",
+      "attention": "sdpa",
+      "batch_size": 4,
+      "query": "Which planet is known as the Red Planet?",
+      "documents": [
+        "Venus is often called Earth's twin because of its similar size and proximity.",
+        "Mars, known for its reddish appearance, is often referred to as the Red Planet.",
+        "Jupiter, the largest planet in our solar system, has a prominent red spot.",
+        "Saturn, famous for its rings, is sometimes mistaken for the Red Planet."
+      ],
+      "torch": "2.14.0+cpu",
+      "transformers": "5.22.1",
+      "sentence_transformers": "6.3.0.dev0",
+      "python": "3.14.4",
+      "prompts": {},
+      "load_seconds": 12.531818921677768,
+      "cpu_seconds": 5.40267545171082,
+      "cpu_scores": [
+        -2.697294235229492,
+        -1.0245951414108276,
+        -2.3062751293182373,
+        -2.0577659606933594
+      ],
+      "status": "error",
+      "stage": "move_to_neuron",
+      "error_type": "RuntimeError",
+      "error": "NRT Execution error occurred on Neuron for operation=neuron::alloc::lazy\nerror message=\"NRT EXECUTION FAILED: lazy::AllocBind: NRT_RESOURCE; no pending ops on stream to wait for (cannot defer), Failed to allocate resource\"\nNote: Neuron errors are reported asynchronously. The Python traceback shows where the\nerror was detected (synchronization point), not necessarily where the failing operation\nwas dispatched. For synchronous error reporting, set NEURON_LAUNCH_BLOCKING=1.",
+      "traceback": "Traceback (most recent call last):\n  File \"/tmp/neuron-top10-sweep/run_model.py\", line 35, in <module>\n    stage='move_to_neuron'; model.to('neuron:0');r['model_device']=str(model.device)\n                            ~~~~~~~~^^^^^^^^^^^^\n  File \"/home/jingyahuang/sentence-transformers/sentence_transformers/base/model.py\", line 1720, in to\n    result = super().to(*args, **kwargs)\n  File \"/home/jingyahuang/.venv-oct09/lib/python3.14/site-packages/torch/nn/modules/module.py\", line 1388, in to\n    return self._apply(convert)\n           ~~~~~~~~~~~^^^^^^^^^\n  File \"/home/jingyahuang/sentence-transformers/sentence_transformers/base/model.py\", line 1748, in _apply\n    return super()._apply(*args, **kwargs)\n           ~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^\n  File \"/home/jingyahuang/.venv-oct09/lib/python3.14/site-packages/torch/nn/modules/module.py\", line 933, in _apply\n    module._apply(fn)\n    ~~~~~~~~~~~~~^^^^\n  File \"/home/jingyahuang/.venv-oct09/lib/python3.14/site-packages/torch/nn/modules/module.py\", line 933, in _apply\n    module._apply(fn)\n    ~~~~~~~~~~~~~^^^^\n  File \"/home/jingyahuang/.venv-oct09/lib/python3.14/site-packages/torch/nn/modules/module.py\", line 933, in _apply\n    module._apply(fn)\n    ~~~~~~~~~~~~~^^^^\n  [Previous line repeated 4 more times]\n  File \"/home/jingyahuang/.venv-oct09/lib/python3.14/site-packages/torch/nn/modules/module.py\", line 966, in _apply\n    param_applied = fn(param)\n  File \"/home/jingyahuang/.venv-oct09/lib/python3.14/site-packages/torch/nn/modules/module.py\", line 1374, in convert\n    return t.to(\n           ~~~~^\n        device,\n        ^^^^^^^\n        dtype if t.is_floating_point() or t.is_complex() else None,\n        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n        non_blocking,\n        ^^^^^^^^^^^^^\n    )\n    ^\nRuntimeError: NRT Execution error occurred on Neuron for operation=neuron::alloc::lazy\nerror message=\"NRT EXECUTION FAILED: lazy::AllocBind: NRT_RESOURCE; no pending ops on stream to wait for (cannot defer), Failed to allocate resource\"\nNote: Neuron errors are reported asynchronously. The Python traceback shows where the\nerror was detected (synchronization point), not necessarily where the failing operation\nwas dispatched. For synchronous error reporting, set NEURON_LAUNCH_BLOCKING=1.\n",
+      "total_seconds": 44.26883608661592
+    }
+  ]
+}