# 아티팩트 생성 시 오류 발생 이슈

**URL:** <https://forums.furiosa.ai/t/topic/370>\
**Category:** Furiosa LLM\
**Created:** [October 27, 2025, 4:33am UTC](https://forums.furiosa.ai/t/topic/370 "2025-10-27T04:33:25Z")\
**Posts on this page:** 11\
**Page:** 1

<div class="post-metadata">

**Author:** ![tobewiseys](https://avatars.discourse-cdn.com/v4/letter/t/a87d85/32.png) [@tobewiseys](https://forums.furiosa.ai/u/tobewiseys)\
**Post date:** [October 27, 2025, 4:33am UTC](https://forums.furiosa.ai/t/topic/370/1 "2025-10-27T04:33:25Z")

</div>

빠른 지원에 감사드리고 있습니다.  
이번에는 다음과 같이 아티팩트 생성과 관련된 내용입니다.

아래와 같은 명령을 이용하여 llama 3.1 8b를 미세조정한 모델을  
양자화한 후, 아티팩트를 만들고자 하였습니다.

> furiosa-llm build ./quantized\_llam318\_komedic   
> ./komedic\_qt8b   
> -tp 8 --max-seq-len-to-capture 4096   
> –num-pipeline-builder-workers 4   
> –num-compile-workers 4

해당 모델은 허깅페이스에 unidocs/llama-3.1-8b-komedic-instruct 로 공개되어 있으며,  
해당 모델을 내려 받아서 8bit로 양자화 시켜서 ./quantized\_llam318\_komedic 폴더에 저장하였습니다.

양자화된 모델로 아티팩트를 만들 때 처리 속도 향상을 위해서 --num-pipeline-builder-workers와 --num-compile-workers의 숫자를 증가시켰습니다. 현재 36core여서 16,16으로 처음에 세팅하였으나, 아티팩트 만드는 시간이 증가하지 않아서 4,4로 조정하였습니다.

위와 같이 수행한 후 process 상태를 확인해 보면 ray::IDLE 상태가 많이 보입니다.

> elicer 12237 12156 0 01:59 pts/2 00:00:20 ray::\_\_build\_fx\_pipelines\_with\_ray  
> elicer 12238 12156 0 01:59 pts/2 00:00:05 ray::IDLE  
> elicer 12239 12156 0 01:59 pts/2 00:00:05 ray::IDLE  
> elicer 12749 5379 0 01:59 pts/3 00:00:00 grep --color=auto ray  
> (llama\_ve310) elicer@f5e3ec3ba50d:~$ source /home/elicer/workspaces/llama\_ve310/bin/activate  
> (llama\_ve310) elicer@f5e3ec3ba50d:~$ ps -ef | grep ray

16,16으로 늘린 경우에는 대부분이 IDLE 상태이고 CPU 점유율이 5%가 되지 않고 있습니다.  
이전에 제가 성공적으로 작업을 완료하였을 때는 50분 경 작업이 종료되었는데 지금은 10시간 가까이 돌려도 아티팩트가 생성되지 않고 있습니다. (SSD에서 HDD로 변경한 것이 관련이 있을지요?)

10시간 정도 지난 후에 아래와 같은 메시지가 나오고 있습니다.

> - Embedding  
> tp\_config:  
> inputs:   
> outputs:   
> operators: {}
> 
> (\_\_compile\_supertasks\_with\_ray pid=12237) INFO:2025-10-26 02:00:21+0000 [furiosa-llm] Compiling pipeline Quantized\_furiosa\_llm\_models.llama3.symbolic.mlperf\_submission.LlamaForCausalLM-kv4095-b1-attn4096, supertask 2 for renegade-8pe.  
> (\_\_compile\_supertasks\_with\_ray pid=12237) INFO:2025-10-26 02:00:21+0000 Block type: first  
> (\_\_compile\_supertasks\_with\_ray pid=12237) INFO:2025-10-26 02:00:21+0000 Loading compiler config for CompilerConfigContext(model\_metadata=ModelMetadata(pretrained\_id=‘meta-llama/Llama-3.1-8B’, task\_type=‘text-generation’, llm\_config=LLMConfig(optimization\_config=OptimizationConfig(attention\_type=\<AttentionType.PAGED\_ATTENTION: ‘PAGED\_ATTENTION’\>, optimize\_rope=True, optimize\_packed=True, decompose\_layernorm=False, optimize\_furiosa=False, use\_unsplit\_packed=False, compact\_causal\_mask=False, use\_rngd\_gelu=False, causal\_mask\_free\_decoding=True, kv\_cache\_sharing\_across\_beams=False, inbound\_beamsearch\_softmax=False, calculate\_logit\_only\_for\_last\_token=False, optimized\_for\_speculative\_decoding=False, use\_2d\_masks=False, merged\_kv\_indices=False), quantization\_config=QuantizationConfig(weight=\<QDtype.FP8: ‘fp8’\>, activation=\<QDtype.FP8: ‘fp8’\>, kv\_cache=\<QDtype.FP8: ‘fp8’\>, use\_mcp=True)), hf\_configs={‘vocab\_size’: 128256, ‘max\_position\_embeddings’: 131072, ‘hidden\_size’: 4096, ‘intermediate\_size’: 14336, ‘num\_hidden\_layers’: 32, ‘num\_attention\_heads’: 32, ‘num\_key\_value\_heads’: 8, ‘hidden\_act’: ‘silu’, ‘initializer\_range’: 0.02, ‘rms\_norm\_eps’: 1e-05, ‘pretraining\_tp’: 1, ‘use\_cache’: False, ‘rope\_theta’: 500000.0, ‘rope\_scaling’: {‘factor’: 8.0, ‘high\_freq\_factor’: 4.0, ‘low\_freq\_factor’: 1.0, ‘original\_max\_position\_embeddings’: 8192, ‘rope\_type’: ‘llama3’}, ‘attention\_bias’: False, ‘attention\_dropout’: 0.0, ‘mlp\_bias’: False, ‘head\_dim’: 128, ‘torch\_dtype’: ‘bfloat16’, ‘tie\_word\_embeddings’: False, ‘architectures’: [‘LlamaForCausalLM’], ‘bos\_token\_id’: 128000, ‘eos\_token\_id’: [128001, 128008, 128009], ‘\_name\_or\_path’: ‘./quantized\_llam318\_komedic’, ‘transformers\_version’: ‘4.48.1’, ‘model\_type’: ‘llama’}, model\_weight\_path=None, trust\_remote\_code=False, allow\_bfloat16\_cast\_with\_mcp=True, auto\_bfloat16\_cast=None), num\_pe\_per\_chip=8, num\_chip=1, block\_type=\<BlockType.FIRST: ‘first’\>, num\_blocks\_per\_graph=1, embedding\_as\_single\_block=False, bucket=Bucket(batch\_size=1, attention\_size=4096, kv\_cache\_size=4095), phase=\<PipelineMode.LLM\_DECODE: ‘decode’\>, beam\_size=None, compiler\_config\_overrides=None, enable\_bf16\_partial\_sum\_for\_split=True)  
> (\_\_compile\_supertasks\_with\_ray pid=12237) INFO:2025-10-26 02:00:21+0000 Using compiler config {‘progress\_mode’: ‘ProgressBar’, ‘propagate\_sparse\_axis\_from\_op’: ‘Gather’, ‘cast\_64bit\_types’: True, ‘support\_i64\_index’: False, ‘implicit\_type\_casting’: False, ‘implicit\_type\_casting\_all\_activation’: False, ‘reshape\_remover\_ignores\_slice\_concat’: False, ‘ignore\_casts\_for\_precision\_gain’: True, ‘arithmetic\_qk\_masking’: False, ‘separate\_vector\_ops\_from\_dpe’: False, ‘separate\_vrf\_from\_interleaving’: True, ‘no\_interleaving\_at\_fusion\_stage’: False, ‘experimental\_demote\_concat’: True, ‘activation\_shape\_guide’: None, ‘normalize\_sine\_cosine\_operator\_domain’: False, ‘max\_operation\_memory\_ratio’: 1.0, ‘tensor\_size\_in\_page\_after\_split’: 2, ‘tensor\_unit\_bridge\_threshold\_in\_page’: 12, ‘dma\_bridge\_threshold\_in\_bytes’: 4294967296, ‘use\_dma\_bridges\_only’: False, ‘use\_split\_operation\_in\_tk’: True, ‘allow\_unlowered\_operators’: False, ‘remove\_lower’: False, ‘remove\_unlower’: False, ‘lowering\_mode’: ‘Optimal’, ‘max\_num\_partitioning\_axes’: 2, ‘max\_num\_unique\_shapes’: 500, ‘padding\_policy’: {‘Small’: 1.1}, ‘allow\_unlimited\_padding’: True, ‘tactic\_hint’: ‘ForLlmModelDecode’, ‘tactic\_context\_config’: ‘None’, ‘local\_population\_threshold’: 0, ‘use\_sparse\_bridge\_population’: True, ‘reshape\_einsum\_mode’: {‘Reshape’: {‘permute’: False}}, ‘use\_efficiently\_broadcasted\_tactic’: False, ‘populate\_dma\_optimized\_einsum\_tactics’: True, ‘use\_exhaustive\_search\_in\_binary\_lowering’: False, ‘dma\_segment\_mode’: ‘Outer’, ‘populate\_irregular\_indices\_shape’: True, ‘allow\_hp\_lowering’: True, ‘use\_attention\_kernel’: True, ‘attention\_kernel\_hint\_mode’: ‘Decode’, ‘tactic\_sorting\_policy’: ‘ByEstimation’, ‘dont\_care\_bridge\_cost’: False, ‘dma\_preference’: 1.0, ‘num\_transaction\_simulation\_per\_pe’: 1024, ‘tactic\_tail\_shape\_alignment’: 8, ‘enable\_vrf\_half\_mode’: True, ‘vrf\_reuse\_optimization\_level’: 2, ‘reduce\_by\_ve\_allow\_partitioning\_tail\_split’: True, ‘enable\_tactic\_pruning’: True, ‘use\_aligned\_repartition2’: True, ‘bf16\_partial\_sum\_policy’: ‘EnableForSplitAndChipAndCluster’, ‘coalesce\_tensors\_by\_common\_split’: False, ‘use\_block\_compile’: False, ‘skip\_clustering\_by\_scc’: False, ‘allow\_external\_operators\_in\_lir’: False, ‘no\_dram\_reuse’: False, ‘scheduler\_beam\_search’: True, ‘subgraph\_scheduling’: False, ‘attention\_mask\_reuse\_traverse’: True, ‘expected\_total\_beam\_states’: 100000, ‘focused\_ops\_coverage\_limit’: None, ‘dma\_grace\_period’: 10000, ‘sync\_grace\_period\_per\_chip’: 10000, ‘estimate\_dma\_command\_gather\_with\_concrete\_samples’: False, ‘mimic\_sync\_io’: False, ‘profile\_sync’: False, ‘dump\_in\_nvp’: False, ‘instruction\_mem\_budget’: 720896, ‘instruction\_chunk\_size’: 2000, ‘enable\_tuc\_profile’: False, ‘dedup\_task\_commands’: True, ‘insert\_wait\_by\_estimation’: False, ‘resolve\_noc\_timeout’: True, ‘c\_compile\_debug\_mode’: False, ‘dma\_throughput\_per\_pe’: 180, ‘remove\_dtoh\_htod’: False, ‘use\_einsum\_by\_dpe\_for\_interleaved\_mul’: False, ‘use\_dma\_stos\_for\_concat\_paste’: True, ‘use\_custom\_broadcast\_for\_concat\_paste’: False, ‘profile\_exact\_command\_cycle’: False, ‘duplicate\_arm\_binary’: False, ‘allow\_reduce\_by\_ve\_cluster\_chip\_reduce’: True, ‘allow\_multiple\_consumer\_rf’: False, ‘allow\_reduce\_by\_ve\_cluster\_chip\_reduce\_base\_population’: False, ‘all\_reduce\_as\_dma\_reduce\_broadcast\_tactic’: True}  
> (\_\_compile\_supertasks\_with\_ray pid=12237) INFO:2025-10-26 02:00:21+0000 [furiosa-llm] Generated graph metadata: —  
> (\_\_compile\_supertasks\_with\_ray pid=12237) valid\_length: ~  
> (\_\_compile\_supertasks\_with\_ray pid=12237) graph\_io\_category:  
> (\_\_compile\_supertasks\_with\_ray pid=12237) input\_category:  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - ModelInput  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - ModelInput  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - ModelInput  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> INFO:2025-10-26 02:00:21+0000 hash for the graph: 015314bf4f0084e296cc44b2da7571f3  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - ModelInput  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - ModelInput  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - ModelInput  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - ModelInput  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - ModelInput  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - ModelInput  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Weight  
> (\_\_compile\_supertasks\_with\_ray pid=12237) output\_category:  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Intermediate:  
> (\_\_compile\_supertasks\_with\_ray pid=12237) named\_axes:  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Batch  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Sequence  
> (\_\_compile\_supertasks\_with\_ray pid=12237) - Embedding  
> (\_\_compile\_supertasks\_with\_ray pid=12237) tp\_config:  
> (\_\_compile\_supertasks\_with\_ray pid=12237) inputs:   
> (\_\_compile\_supertasks\_with\_ray pid=12237) outputs:   
> (\_\_compile\_supertasks\_with\_ray pid=12237) operators: {}  
> (\_\_compile\_supertasks\_with\_ray pid=12237)  
> (\_\_compile\_supertasks\_with\_ray pid=12237) INFO:2025-10-26 02:00:21+0000 hash for the graph: d3809467d55a7b74d967ae5279a35b1f  
> source /home/elicer/workspaces/llama\_ve310/bin/activate  
> ERROR: invalid npu id: renegade-8pe  
> Encountered exception!  
> non\_shared\_configs(w/o past\_kv): [NonSharedPipelineBuildConfig(args\_data=(), kwargs\_data={‘input\_ids’: TensorGenInfo(shape=torch.Size([1, 4096]), dtype=torch.int32), ‘position\_ids’: TensorGenInfo(shape=torch.Size([1, 4096]), dtype=torch.int32), ‘new\_key\_location’: TensorGenInfo(shape=torch.Size([1, 4096]), dtype=torch.int32), ‘new\_value\_location’: TensorGenInfo(shape=torch.Size([1, 4096]), dtype=torch.int32), ‘bucket\_size’: 4096, ‘use\_cache’: False, ‘is\_prefill’: True, ‘causal\_mask’: TensorGenInfo(shape=torch.Size([1, 4096, 4096]), dtype=torch.bool)}, pipeline\_name=‘Quantized\_furiosa\_llm\_models.llama3.symbolic.mlperf\_submission.LlamaForCausalLM-kv0-b1-attn4096’, compile\_config=CompilerConfigContext(model\_metadata=ModelMetadata(pretrained\_id=‘meta-llama/Llama-3.1-8B’, task\_type=‘text-generation’, llm\_config=LLMConfig(optimization\_config=OptimizationConfig(attention\_type=\<AttentionType.PAGED\_ATTENTION: ‘PAGED\_ATTENTION’\>, optimize\_rope=True, optimize\_packed=True, decompose\_layernorm=False, optimize\_furiosa=False, use\_unsplit\_packed=False, compact\_causal\_mask=False, use\_rngd\_gelu=False, causal\_mask\_free\_decoding=True, kv\_cache\_sharing\_across\_beams=False, inbound\_beamsearch\_softmax=False, calculate\_logit\_only\_for\_last\_token=False, optimized\_for\_speculative\_decoding=False, use\_2d\_masks=False, merged\_kv\_indices=False), quantization\_config=QuantizationConfig(weight=\<QDtype.FP8: ‘fp8’\>, activation=\<QDtype.FP8: ‘fp8’\>, kv\_cache=\<QDtype.FP8: ‘fp8’\>, use\_mcp=True)), hf\_configs={‘vocab\_size’: 128256, ‘max\_position\_embeddings’: 131072, ‘hidden\_size’: 4096, ‘intermediate\_size’: 14336, ‘num\_hidden\_layers’: 32, ‘num\_attention\_heads’: 32, ‘num\_key\_value\_heads’: 8, ‘hidden\_act’: ‘silu’, ‘initializer\_range’: 0.02, ‘rms\_norm\_eps’: 1e-05, ‘pretraining\_tp’: 1, ‘use\_cache’: False, ‘rope\_theta’: 500000.0, ‘rope\_scaling’: {‘factor’: 8.0, ‘high\_freq\_factor’: 4.0, ‘low\_freq\_factor’: 1.0, ‘original\_max\_position\_embeddings’: 8192, ‘rope\_type’: ‘llama3’}, ‘attention\_bias’: False, ‘attention\_dropout’: 0.0, ‘mlp\_bias’: False, ‘head\_dim’: 128, ‘torch\_dtype’: ‘bfloat16’, ‘tie\_word\_embeddings’: False, ‘architectures’: [‘LlamaForCausalLM’], ‘bos\_token\_id’: 128000, ‘eos\_token\_id’: [128001, 128008, 128009], ‘\_name\_or\_path’: ‘./quantized\_llam318\_komedic’, ‘transformers\_version’: ‘4.48.1’, ‘model\_type’: ‘llama’}, model\_weight\_path=None, trust\_remote\_code=False, allow\_bfloat16\_cast\_with\_mcp=True, auto\_bfloat16\_cast=None), num\_pe\_per\_chip=None, num\_chip=None, block\_type=None, num\_blocks\_per\_graph=1, embedding\_as\_single\_block=False, bucket=Bucket(batch\_size=1, attention\_size=4096, kv\_cache\_size=0), phase=\<PipelineMode.LLM\_PREFILL: ‘prefill’\>, beam\_size=None, compiler\_config\_overrides=None, enable\_bf16\_partial\_sum\_for\_split=True), logits\_slice\_config=LogitsSliceConfig(slice\_direction=‘left’, slice\_size=1), num\_blocks\_per\_supertask=1), NonSharedPipelineBuildConfig(args\_data=(), kwargs\_data={‘input\_ids’: TensorGenInfo(shape=torch.Size([1, 1]), dtype=torch.int32), ‘attention\_mask’: TensorGenInfo(shape=torch.Size([1, 4096]), dtype=torch.bool), ‘position\_ids’: TensorGenInfo(shape=torch.Size([1, 1]), dtype=torch.int32), ‘new\_key\_location’: TensorGenInfo(shape=torch.Size([1, 1]), dtype=torch.int32), ‘new\_value\_location’: TensorGenInfo(shape=torch.Size([1, 1]), dtype=torch.int32), ‘bucket\_size’: 4096, ‘use\_cache’: False, ‘is\_prefill’: False, ‘past\_valid\_key\_indices’: TensorGenInfo(shape=torch.Size([4095]), dtype=torch.int32), ‘past\_valid\_value\_indices’: TensorGenInfo(shape=torch.Size([4095]), dtype=torch.int32)}, pipeline\_name=‘Quantized\_furiosa\_llm\_models.llama3.symbolic.mlperf\_submission.LlamaForCausalLM-kv4095-b1-attn4096’, compile\_config=CompilerConfigContext(model\_metadata=ModelMetadata(pretrained\_id=‘meta-llama/Llama-3.1-8B’, task\_type=‘text-generation’, llm\_config=LLMConfig(optimization\_config=OptimizationConfig(attention\_type=\<AttentionType.PAGED\_ATTENTION: ‘PAGED\_ATTENTION’\>, optimize\_rope=True, optimize\_packed=True, decompose\_layernorm=False, optimize\_furiosa=False, use\_unsplit\_packed=False, compact\_causal\_mask=False, use\_rngd\_gelu=False, causal\_mask\_free\_decoding=True, kv\_cache\_sharing\_across\_beams=False, inbound\_beamsearch\_softmax=False, calculate\_logit\_only\_for\_last\_token=False, optimized\_for\_speculative\_decoding=False, use\_2d\_masks=False, merged\_kv\_indices=False), quantization\_config=QuantizationConfig(weight=\<QDtype.FP8: ‘fp8’\>, activation=\<QDtype.FP8: ‘fp8’\>, kv\_cache=\<QDtype.FP8: ‘fp8’\>, use\_mcp=True)), hf\_configs={‘vocab\_size’: 128256, ‘max\_position\_embeddings’: 131072, ‘hidden\_size’: 4096, ‘intermediate\_size’: 14336, ‘num\_hidden\_layers’: 32, ‘num\_attention\_heads’: 32, ‘num\_key\_value\_heads’: 8, ‘hidden\_act’: ‘silu’, ‘initializer\_range’: 0.02, ‘rms\_norm\_eps’: 1e-05, ‘pretraining\_tp’: 1, ‘use\_cache’: False, ‘rope\_theta’: 500000.0, ‘rope\_scaling’: {‘factor’: 8.0, ‘high\_freq\_factor’: 4.0, ‘low\_freq\_factor’: 1.0, ‘original\_max\_position\_embeddings’: 8192, ‘rope\_type’: ‘llama3’}, ‘attention\_bias’: False, ‘attention\_dropout’: 0.0, ‘mlp\_bias’: False, ‘head\_dim’: 128, ‘torch\_dtype’: ‘bfloat16’, ‘tie\_word\_embeddings’: False, ‘architectures’: [‘LlamaForCausalLM’], ‘bos\_token\_id’: 128000, ‘eos\_token\_id’: [128001, 128008, 128009], ‘\_name\_or\_path’: ‘./quantized\_llam318\_komedic’, ‘transformers\_version’: ‘4.48.1’, ‘model\_type’: ‘llama’}, model\_weight\_path=None, trust\_remote\_code=False, allow\_bfloat16\_cast\_with\_mcp=True, auto\_bfloat16\_cast=None), num\_pe\_per\_chip=None, num\_chip=None, block\_type=None, num\_blocks\_per\_graph=1, embedding\_as\_single\_block=False, bucket=Bucket(batch\_size=1, attention\_size=4096, kv\_cache\_size=4095), phase=\<PipelineMode.LLM\_DECODE: ‘decode’\>, beam\_size=None, compiler\_config\_overrides=None, enable\_bf16\_partial\_sum\_for\_split=True), logits\_slice\_config=None, num\_blocks\_per\_supertask=1)]
> 
> Traceback (most recent call last):  
> File “/home/elicer/workspaces/llama\_ve310/lib/python3.10/site-packages/furiosa\_llm/parallelize/pipeline/builder/converter.py”, line 819, in compile\_gm\_and\_get\_preprocessed\_gm\_hash  
> compiled = compile(  
> RuntimeError: fail to compile: Invalid NPU ID
> 
> During handling of the above exception, another exception occurred:
> 
> Traceback (most recent call last):  
> File “/home/elicer/workspaces/llama\_ve310/bin/furiosa-llm”, line 7, in   
> sys.exit(main())  
> File “/home/elicer/workspaces/llama\_ve310/lib/python3.10/site-packages/furiosa\_llm/cli/main.py”, line 20, in main  
> args.dispatch\_function(args)  
> File “/home/elicer/workspaces/llama\_ve310/lib/python3.10/site-packages/furiosa\_llm/cli/convert.py”, line 246, in convert  
> builder.build(  
> File “/home/elicer/workspaces/llama\_ve310/lib/python3.10/site-packages/furiosa\_llm/artifact/builder.py”, line 687, in build  
> target\_model\_artifact, target\_model\_pipelines = self.\_build\_model\_artifact(  
> File “/home/elicer/workspaces/llama\_ve310/lib/python3.10/site-packages/furiosa\_llm/artifact/builder.py”, line 590, in \_build\_model\_artifact  
> pipelines\_with\_metadata = build\_pipelines(  
> File “/home/elicer/workspaces/llama\_ve310/lib/python3.10/site-packages/furiosa\_llm/artifact/helper.py”, line 285, in build\_pipelines  
> pipelines = pipeline\_builder.build\_pipelines(  
> File “/home/elicer/workspaces/llama\_ve310/lib/python3.10/site-packages/furiosa\_llm/parallelize/pipeline/builder/api.py”, line 844, in build\_pipelines  
> return PipelineBuilder.compile\_supertasks\_in\_parallel(  
> File “/home/elicer/workspaces/llama\_ve310/lib/python3.10/site-packages/furiosa\_llm/parallelize/pipeline/builder/api.py”, line 1064, in compile\_supertasks\_in\_parallel  
> local\_pipelines = PipelineBuilder.\_\_compile\_supertasks\_aux(  
> File “/home/elicer/workspaces/llama\_ve310/lib/python3.10/site-packages/furiosa\_llm/parallelize/pipeline/builder/api.py”, line 1108, in \_\_compile\_supertasks\_aux  
> \_compile\_supertasks\_in\_pipeline(  
> File “/home/elicer/workspaces/llama\_ve310/lib/python3.10/site-packages/furiosa\_llm/parallelize/pipeline/builder/api.py”, line 250, in \_compile\_supertasks\_in\_pipeline  
> compile\_result, hash\_val = GraphModuleConverter.compile\_gm\_and\_get\_preprocessed\_gm\_hash(  
> File “/home/elicer/workspaces/llama\_ve310/lib/python3.10/site-packages/furiosa\_llm/parallelize/pipeline/builder/converter.py”, line 837, in compile\_gm\_and\_get\_preprocessed\_gm\_hash  
> raise RuntimeError(f"Compilation failed with error {e}")  
> RuntimeError: Compilation failed with error fail to compile: Invalid NPU ID  
> (llama\_ve310) elicer@f5e3ec3ba50d:~$

Invalid NPU ID오류가 발생하는데 무엇이 잘못된 것일까요?

> (llama\_ve310) elicer@f5e3ec3ba50d:~/workspaces$ furiosa-smi info  
> ±------±-----±-------±-----------------±-----------------±--------±--------±-------------+  
> | Index | Arch | Device | Firmware | PERT | Temp. | Power | PCI-BDF |  
> ±------±-----±-------±-----------------±-----------------±--------±--------±-------------+  
> | 0 | rngd | npu5 | 2025.3.0+c097ea0 | 2025.3.0+52e5705 | 40.04°C | 34.56 W | 0000:bc:00.0 |  
> ±------±-----±-------±-----------------±-----------------±--------±--------±-------------+  
> | 1 | rngd | npu6 | 2025.3.0+c097ea0 | 2025.3.1+52e5705 | 30.37°C | 34.56 W | 0000:bd:00.0 |  
> ±------±-----±-------±-----------------±-----------------±--------±--------±-------------+

furiosa-llm 버전 2025.3.3 버전으로 아티팩트만들 때 시간이 너무 오래 걸려  
2025.3.1 버전으로 다운그레이드 시켜서 테스트를 한 상태입니다.

> (llama\_ve310) elicer@f5e3ec3ba50d:~/workspaces$ pip list | grep furiosa  
> furiosa-llm 2025.3.1  
> furiosa-llm-models 2025.3.0  
> furiosa-model-compressor 2025.3.0  
> furiosa-model-compressor-impl 2025.3.0  
> furiosa-models-lang 2025.3.0  
> furiosa-native-compiler 2025.3.1  
> furiosa-native-llm-common 2025.3.1  
> furiosa-native-runtime 2025.3.2  
> furiosa-smi-py 2025.3.0  
> furiosa-torch-ext 2025.3.1

---

<div class="post-metadata">

**Author:** ![hyunsik](https://yyz1.discourse-cdn.com/flex031/user_avatar/forums.furiosa.ai/hyunsik/32/6_2.png) [@hyunsik](https://forums.furiosa.ai/u/hyunsik)\
**Post date:** [October 29, 2025, 1:50am UTC](https://forums.furiosa.ai/t/topic/370/2 "2025-10-29T01:50:00Z")

</div>

안녕하세요? 불편드려 죄송합니다. Invalid NPU ID는 `-tp8` 옵션을 주셨으면 발생하면 안되는 에러이기는 합니다. 한번 공유 주신 버전으로 제가 재현해보고 내일 까지 업데이트 드리도록 하겠습니다.

---

<div class="post-metadata">

**Author:** ![hyunsik](https://yyz1.discourse-cdn.com/flex031/user_avatar/forums.furiosa.ai/hyunsik/32/6_2.png) [@hyunsik](https://forums.furiosa.ai/u/hyunsik)\
**Post date:** [October 29, 2025, 1:51am UTC](https://forums.furiosa.ai/t/topic/370/3 "2025-10-29T01:51:39Z")

</div>

혹시 `apt list --installed | grep furiosa` 커맨드의 실행 결과도 공유 부탁드려도 될까요?

예를 들면, 다음과 같이 출력됩니다.

```auto
apt list --installed | grep furiosa 

WARNING: apt does not have a stable CLI interface. Use with caution in scripts.

furiosa-compiler/jammy-rc,now 2025.3.0-3 amd64 [installed]
furiosa-driver-rngd/jammy-rc,now 1.11.0-3 all [installed,upgradable to: 2025.3.1-3]
furiosa-firmware-image-rngd/jammy-rc,now 1.7.0 all [installed,upgradable to: 2025.3.1]
furiosa-firmware-tools-rngd/jammy-rc,now 2025.3.1-3 amd64 [installed]
furiosa-libhal-nvp/jammy-nightly,now 0.11.0-3+nightly-250714+9f4c23a amd64 [installed,upgradable to: 0.11.0-3+nightly-251023+99925dc]
furiosa-libsmi/jammy-rc,now 2025.3.0-3 amd64 [installed]
furiosa-mlperf-resources/jammy-rc,now 5.0.0 amd64 [installed]
furiosa-pert-rngd/jammy-nightly,now 0.1.0-3+nightly-250929+f9abfb9 amd64 [installed,upgradable to: 2025.3.1-3]
furiosa-smi/jammy-rc,now 2025.3.0-3 amd64 [installed]

```

---

<div class="post-metadata">

**Author:** ![tobewiseys](https://avatars.discourse-cdn.com/v4/letter/t/a87d85/32.png) [@tobewiseys](https://forums.furiosa.ai/u/tobewiseys)\
**Post date:** [October 29, 2025, 1:53am UTC](https://forums.furiosa.ai/t/topic/370/4 "2025-10-29T01:53:33Z")

</div>

요청하신 내용은 다음과 같습니다.

> (llama\_ve310) elicer@f5e3ec3ba50d:~$ apt list --installed | grep furiosa
> 
> WARNING: apt does not have a stable CLI interface. Use with caution in scripts.
> 
> furiosa-bench/now 0.10.3-3 amd64 [installed,local]  
> furiosa-compiler/now 0.10.3-3 amd64 [installed,upgradable to: 2025.3.0-3]  
> furiosa-libcompiler/now 0.10.1-3 amd64 [installed,local]  
> furiosa-libhal-warboy/now 0.12.0-3 amd64 [installed,local]  
> furiosa-libnux/now 0.10.1-3 amd64 [installed,local]  
> furiosa-libsmi/jammy,now 2025.3.0-3 amd64 [installed,automatic]  
> furiosa-smi/jammy,now 2025.3.0-3 amd64 [installed]  
> furiosa-toolkit/now 0.11.0-3 amd64 [installed,local]

결과가 많이 다른데요… 버전에 맞게 설치를 해야 하는 것일까요?  
제가 CSP에서 인스턴스를 만들어서 사용중이라 서버 재기동은 별도로 요청해야 합니다.

---

<div class="post-metadata">

**Author:** ![tobewiseys](https://avatars.discourse-cdn.com/v4/letter/t/a87d85/32.png) [@tobewiseys](https://forums.furiosa.ai/u/tobewiseys)\
**Post date:** [October 29, 2025, 2:35am UTC](https://forums.furiosa.ai/t/topic/370/5 "2025-10-29T02:35:41Z")

</div>

참고로 말씀드리면 이전에 만들어진 아티팩트는 허깅페이스에 공개한 상태입니다.  
해당 아티팩트는 잘 작동이 됩니다.

> **[unidocs/llama-3.1-8b-komedic-inst-furiosa-rngd · Hugging Face](https://huggingface.co/unidocs/llama-3.1-8b-komedic-inst-furiosa-rngd)**
>
> We’re on a journey to advance and democratize artificial intelligence through open source and open science.

새로운 인스턴스에서 새로 개발환경을 만들어서 양자화를 수행하고  
아티팩트를 생성하려고 할 때 발생한 오류입니다.  
참고하시기 바랍니다.

원래는 양자화를 안 하고 바로 아티팩트를 만들어서 성능을 비교하려고 했었는데  
아티팩트 만드는 시간이 너무 오래 걸려서 양자화를 한 모델로 성능을 평가하였습니다.

---

<div class="post-metadata">

**Author:** ![hyunsik](https://yyz1.discourse-cdn.com/flex031/user_avatar/forums.furiosa.ai/hyunsik/32/6_2.png) [@hyunsik](https://forums.furiosa.ai/u/hyunsik)\
**Post date:** [October 29, 2025, 3:32pm UTC](https://forums.furiosa.ai/t/topic/370/6 "2025-10-29T15:32:48Z")

</div>

안녕하세요?

설치된 패키지들을 보면 1세대 제품인 Warboy용 패키지들이네요. 그렇다면, Invalid NPU id 오류가 설명이 되네요.

인스턴스에 미리 설치된 패키지였다면, Warboy, RNGD용 인스턴스가 구분되어 있을 것 같네요.

만약에 설치 문서를 보고 직접 설치하신거라면, [Welcome to Furiosa Docs — FuriosaAI Developer Center 2025.3.1 documentation](https://developer.furiosa.ai/latest/en/index.html) 를 참고하시면 됩니다.

---

<div class="post-metadata">

**Author:** ![tobewiseys](https://avatars.discourse-cdn.com/v4/letter/t/a87d85/32.png) [@tobewiseys](https://forums.furiosa.ai/u/tobewiseys)\
**Post date:** [October 30, 2025, 8:20am UTC](https://forums.furiosa.ai/t/topic/370/7 "2025-10-30T08:20:13Z")

</div>

제가 일일이 손으로 맞추어 보았지만 아티팩트를 생성하는 시간이 오래 걸리는 것은 동일했습니다.

수작업으로 아래와 같은 상태까지 만들었지만…  
firmware를 새로 설치하여 콜드 부팅이필요하고  
furiosa-libhal-nvp는 설치가 되지 않아서  
아래와 같은 상태로 양자화는 되지만 아티팩트는 만들어지지 않습니다.  
계속 대기 상태인데요…

> furiosa-compiler/now 0.10.3-3 amd64 [installed,upgradable to: 2025.3.0-3]  
> furiosa-driver-rngd/jammy,now 2025.3.1-3 all [installed]  
> furiosa-firmware-image-rngd/jammy,now 2025.3.1 all [installed]  
> furiosa-firmware-tools-rngd/jammy,now 2025.3.1-3 amd64 [installed]  
> furiosa-libsmi/jammy,now 2025.3.0-3 amd64 [installed,automatic]  
> furiosa-pert-rngd/jammy,now 2025.3.1-3 amd64 [installed]  
> furiosa-smi/jammy,now 2025.3.0-3 amd64 [installed]  
> furiosa-toolkit/now 0.11.0-3 amd64 [installed,local]

그래서 인스턴스를 삭제하고 새로 생성하니  
아래와 같이 warboy 상태로 설정되어 있습니다.

> elicer@a9c6e8a77490:~$ apt list --installed | grep furiosa
> 
> WARNING: apt does not have a stable CLI interface. Use with caution in scripts.
> 
> furiosa-bench/now 0.10.3-3 amd64 [installed,local]  
> furiosa-compiler/now 0.10.3-3 amd64 [installed,local]  
> furiosa-libcompiler/now 0.10.1-3 amd64 [installed,local]  
> furiosa-libhal-warboy/now 0.12.0-3 amd64 [installed,local]  
> furiosa-libnux/now 0.10.1-3 amd64 [installed,local]  
> furiosa-toolkit/now 0.11.0-3 amd64 [installed,local]

인스턴스를 생성할 때 warboy나 renegade를 위한 환경을 특별히 설정하는 것이 없는데요  
이런 경우. csp에게 renegade를 위한 환경설정 작업을 요청해야 하는 것일까요?  
매번 인스턴스 생성할 때 마다 요청을 해야 하는 것인지 모르겠지만  
csp에게 어떤 식으로 요청을 하면 될지 문의드립니다.

---

<div class="post-metadata">

**Author:** ![jongwook.kim](https://avatars.discourse-cdn.com/v4/letter/j/b5a626/32.png) [@jongwook.kim](https://forums.furiosa.ai/u/jongwook.kim)\
**Post date:** [October 30, 2025, 9:13am UTC](https://forums.furiosa.ai/t/topic/370/8 "2025-10-30T09:13:50Z")

</div>

> [@tobewiseys](#):
>
> furiosa-compiler/now 0.10.3-3 amd64 [installed,local]

안녕하세요, 퓨리오사에이아이 김종욱입니다.

환경에 있는 `/etc/apt/sources.list.d/furiosa.list` 파일과 `/etc/apt/auth.conf.d/furiosa.conf` 파일을 삭제하신 후에, [여기 페이지](https://developer.furiosa.ai/latest/en/get_started/prerequisites.html#setting-up-apt) 에서 Setting up APT 항목만 따라하시고, 아래와 같이 comiler를 RNGD 버전에 맞춰서 설치를 하실 수 있을까요?

```auto
$ sudo apt update
$ apt list -a | grep furiosa-compiler # 여기서 2025.3.0 기준의 버전을 확인한 다음
$ sudo apt install furiosa-compiler=2025.3.0-3 # 해당 버전을 지정한 후 설치

```

---

<div class="post-metadata">

**Author:** ![tobewiseys](https://avatars.discourse-cdn.com/v4/letter/t/a87d85/32.png) [@tobewiseys](https://forums.furiosa.ai/u/tobewiseys)\
**Post date:** [October 30, 2025, 9:51am UTC](https://forums.furiosa.ai/t/topic/370/9 "2025-10-30T09:51:21Z")

</div>

수행한 결과는 아래와 같습니다.

> elicer@a9c6e8a77490:/etc/apt/trusted.gpg.d$ apt list --installed | grep furiosa
> 
> WARNING: apt does not have a stable CLI interface. Use with caution in scripts.  
> furiosa-bench/now 0.10.3-3 amd64 [installed,local]  
> furiosa-compiler/jammy,now 2025.3.0-3 amd64 [installed]  
> furiosa-libcompiler/now 0.10.1-3 amd64 [installed,local]  
> furiosa-libhal-warboy/now 0.12.0-3 amd64 [installed,local]  
> furiosa-libnux/now 0.10.1-3 amd64 [installed,local]  
> furiosa-toolkit/now 0.11.0-3 amd64 [installed,local]

---

<div class="post-metadata">

**Author:** ![jongwook.kim](https://avatars.discourse-cdn.com/v4/letter/j/b5a626/32.png) [@jongwook.kim](https://forums.furiosa.ai/u/jongwook.kim)\
**Post date:** [October 30, 2025, 11:51am UTC](https://forums.furiosa.ai/t/topic/370/10 "2025-10-30T11:51:42Z")

</div>

compiler는 정상적으로 설치가된 것으로 보이며, 아티팩트 자체는 만들 수 있을 것으로 보입니다. 한번 시도를 해볼 수 있으실까요? 다만, 실행을 위해서는 RNGD Instance 환경을 따로 받아야할 것 같습니다.

---

<div class="post-metadata">

**Author:** ![tobewiseys](https://avatars.discourse-cdn.com/v4/letter/t/a87d85/32.png) [@tobewiseys](https://forums.furiosa.ai/u/tobewiseys)\
**Post date:** [November 3, 2025, 2:16am UTC](https://forums.furiosa.ai/t/topic/370/11 "2025-11-03T02:16:48Z")

</div>

아티팩트 생성이 기존에서는 1시간 안에 처리되었는데 3시간이 넘어도 처리가 되지 않아서  
일단 중지시킨 상태입니다.

관련하여 csp에 레너게이드를 위한 기본 패키지 변경을 금요일에 요청한 상태이고 대기중입니다.