Share
ð³ DeepSeek ã AWS ã§åããïŒ / DeepSeek on AWS (ja)
æ¬ã€ãã³ãã§ã¯ãDeepSeek ã顿ã«ãçæ AI åºç€ã¢ãã«ã»å€§èŠæš¡èšèªã¢ãã« (LLM) ã®ãããã€ã«é¢ããå®è·µçãªã¬ã€ãã³ã¹ãæäŸããŸãã
ãŸããDeepSeek-R1-ZeroãDeepSeek-R1 ããã³ãã®èžçã¢ãã« (DeepSeek-R1-Distill-Qwen, DeepSeek-R1-Distill-Llama) ã®æè¡è§£èª¬ãè¡ããŸããç¶ããŠãAWS ã§ LLM ã®ãããã€ã¡ã³ãã«å©çšå¯èœãªã¢ã¯ã»ã©ã¬ãŒã¿ (NVIDIA GPU, AWS Trainium/Inferentia) ããAWS ãµãŒãã¹ã®éžæè¢ (Amazon Bedrock, Amazon SageMaker AI, Amazon EC2) ã玹ä»ããŸãã
åŸåã§ã¯ãå®éã« DeepSeek-R1 ã®èžçã¢ãã«ãçšãããã³ãºãªã³åœ¢åŒã®ã¯ãŒã¯ã·ã§ãããè¡ããã³ã¹ããšããã©ãŒãã³ã¹ã®æé©åã«ã€ããŠãåãäžããŸããæ¬ã»ãã·ã§ã³ãéããŠãLLM ãããã€ã¡ã³ãæŠç¥ã«åœ¹ç«ã€ç¥èŠãæäŸããŸãã
https://aws.amazon.com/startups/events/deepseek-workshop
Yoshitaka Haribara
Transcript
39 slides
-
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 3 © 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. Yoshitaka Haribara, Ph.D. A W S S T A R T U P L O F T T O K Y O Sr. GenAI/Quantum Startup Solutions Architect AWS ð³ DeepSeek ã AWS ã§åããïŒ -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 4 â¢ èæ¯: Reasoning (è«ççæšè«) ãš Chain-of-Thought (CoT) ⢠DeepSeek-R1 ãšèžçã¢ãã«ã®æŠèŠ â¢ AI ã¢ã¯ã»ã©ã¬ãŒã¿: NVIDIA H200 GPU ãš AWS Trainium, Inferentia ⢠AWS ãµãŒãã¹ã®éžæ: Amazon Bedrock Marketplace, Amazon SageMaker AI, Amazon EC2 ⢠ãã³ãºãªã³: Amazon Bedrock Marketplace, Amazon SageMaker (ml.inf2.xlarge) Agenda -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 5 Generative Pre-trained Transformer (GPT) ã«ä»£è¡šãããããã«ã倧éã®åŠç¿ ããŒã¿ãçšããèªå·±æåž«ããåŠç¿ã«ãããé«ãæ±åæ§èœãæã€èšèªã¢ãã«: LLM äºååŠç¿ (Pre-training) ã¯éèŠãªãã©ãã€ã ã§ãããã¹ã±ãŒãªã³ã°å (ã¢ãã« ãµã€ãºã»ããŒã¿éã»èšç®ãªãœãŒã¹) ãå¢ãããšæ§èœãå調å¢å : Scaling law Supervised Finetuning (SFT) ã Chain-of-Thought (CoT) ã«ä»£è¡šããã In-context learning ãšãã£ãäºåŸåŠç¿ãåŸã ã«èå ãæµŽã³ã æšè«ã«æéããããããšã§é«ãæ§èœãéæãããããã«ãªã£ãŠãã LLM ã®åŠç¿: äºååŠç¿ã»äºåŸåŠç¿ -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 6 課é¡: å ã LLM ã¯ç®æ°ãè«ççæšè«ãäžåŸæ (cf. é圢åŒç æŒç¹¹ç æšè«; Informal deductive reasoning) ãããå æããããã以äžã®ãããªææ³ãç ç©¶ã»ææ¡ãããŠãã (åŸè¿°) ⢠Chain-of-Thought (CoT) ⢠Zero-shot CoT ⢠CoT with Self-consistency (Multi-path CoT) ⢠Tree of thought (ToT) Reasoning (è«ççæšè«) -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 7 LLM ã«äžéçãªæšè«ã¹ããããçæãããããšã§ãè«ççæšè«èœåãåäž ã¢ãã«å ¥å Q: ããžã£ãŒã¯ðŸã5åæã£ãŠããã2猶ã®ðŸãè²·ãè¶³ããã 1猶ã«ã¯3åã®ðŸãå ¥ã£ãŠããä»ã圌ã¯åèšã§äœåã®ðŸãæã£ãŠãïŒ A: ããžã£ãŒã¯ðŸãæå5åæã£ãŠããã3åå ¥ã x 2猶å (6å) ðŸãè²·ãè¶³ããã 5 + 6 = 11ã§ãçãã¯11åã Q: ã«ãã§ããªã¢ã«ðã23åãã£ãããã®ãã¡20åãã©ã³ãã®æºåã«äœ¿ãã æ°ãã«6åè²·ãè¶³ãããä»ãã«ãã§ããªã¢ã«ðã¯äœåïŒ ã¢ãã«åºå A: ã«ãã§ããªã¢ã«ã¯æåðã23åãã©ã³ãã®æºåã«20å䜿ã£ãã ãã£ãŠ23 - 20 = 3åã«ã6åè²·ãè¶³ãã3 + 6 = 9åãçãã¯9åã Chain-of-Thought (CoT) [arXiv:2201.11903] -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 8 âLetâs think step by stepâ (ãã¹ããããã€ã¹ãããã§èãããã) ãšèšãã ã ã¢ãã«å ¥åïŒ Q: ãžã§ã³ã¯16åã®ãã«ãŒããè²·ãããã ãã«ãŒãã®ååã¯ãªã³ãŽã§ããã®ã®ååã¯ðã«ãããðã¯äœåïŒ A: ã¹ããããã€ã¹ãããã§èãããã ã¢ãã«åºåïŒ ãã«ãŒãã¯å šéšã§16åã ãã«ãŒãã®ååããªã³ãŽãã€ãŸããªã³ãŽã¯8åã ãªã³ãŽã®ååãðãã€ãŸãðã¯4åã Zero-shot CoT [arXiv:2205.11916] -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 9 è€æ°çµè·¯ã®æšè«ãè¡ããããã®åçããæçµçãªçµæãå°ã ãžã£ãããã®ðª¿ã¯1æ¥ã«16åã®ð¥ãç£ããåœŒå¥³ã¯æ¯æ3åãæé£ã«é£ã¹ã æ¯æ¥4åã䜿ã£ãŠå人ã®ããã«ããã£ã³ãçŒããæ®ãã®ð¥ã¯1å2ãã«ã§å£²ãã åœŒå¥³ã¯æ¯æ¥ããã皌ããïŒ â¢ åœŒå¥³ã¯16 - 3 - 4 = 9åã®ð¥ãæ®ãã⎠1æ¥ã«2ãã« Ã 9 = 18ãã«çšŒãã â¢ åœŒå¥³ã¯æ®ãã®åµã2ãã« Ã (16 - 4 - 3) = 26ãã«ã§å£²ãããšã«ãªãã â¢ åœŒå¥³ã¯æé£ã«3åé£ã¹ã16 - 3 = 13åãæ®ããæ¬¡ã«ããã£ã³ãçŒãã13 - 4 = 9åã®åµãæ®ãã⎠9åã®åµ à 2ãã« = 18ãã« â çãã¯18ãã« Multi-path CoT with Self-consistency [arXiv:2203.11171] -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 10 æšæ§é ã®ãããªæèããã»ã¹ã«ããè§£ãæ¢çŽ¢ Tree of thought (ToT) [arXiv:2305.08291, arXiv:2305.10601] ToT è«æ [arXiv:2305.10601] ãã -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 13 課é¡: LLM ã§äºåŸåŠç¿ã«åŒ·ååŠç¿ã䜿ããšèšç®éãå€ããªã äºååŠç¿æžã¿ DeepSeek-V3-Base ãããŒã¹ã«ã匷ååŠç¿ãã¬ãŒã ã¯ãŒã¯ãšã㊠Group Relative Policy Optimization (GRPO) [arXiv:2402.03300] ãæ¡çš DeepSeek-R1-Zero [arXiv:2501.12948] DeepSeekMath [arXiv:2402.03300] ã®å³ãæ¹å€ Rule-based -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 14 課é¡: å¯èªæ§ãèšèªã®æ··åš 匷ååŠç¿ããåã«å°éã®ã³ãŒã«ããŒã¹ã¿ãŒãããŒã¿ãšè€æ°ã¹ããŒãžåŠç¿: ⢠DeepSeek-V3-Base ã® fine-tuining ã®ããæ°åã®ã³ãŒã«ãã¹ã¿ãŒãããŒã¿åé ⢠DeepSeek-R1-Zero ã®ããã« Reasoning éèŠã®åŒ·ååŠç¿ ⢠匷ååŠç¿ãåæããããªæç¹ã§æ£åŽãµã³ããªã³ã°ã«ãã SFT ããŒã¿ãåéã DeepSeek-V3 ã®ãã¡ã€ã³æåž«ããŒã¿ (writing, factural QA, self-congnition) ã åããã ⢠DeepSeek-V3-Base ãååŠç¿ (SFT) ⢠åŠç¿åŸãå šãŠã®ã·ããªãªã§åŒ·ååŠç¿ããããã§ãã¯ãã€ã³ãã DeepSeek-R1 DeepSeek-R1 [arXiv:2501.12948] -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 15 DeepSeek-R1 ããèžçããã¢ãã«: Qwen ãš Llama ã®2系統 ⢠DeepSeek-R1-Distill-Qwen (1.5B, 7B, 14B, 32B) ⢠DeepSeek-R1-Distill-Llama (8B, 70B) DeepSeek-R1-Distill (èžçã¢ãã«) [arXiv:2501.12948] https://github.com/deepseek-ai/DeepSeek-R1 -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 16 DeepSeek ãšããã€ãã®èžçã¢ãã«ã®éã¿ãå ¬é ããŒã¹ (V3) ãš R1 ã¢ãã« (671B ãã©ã¡ãŒã¿) ⢠DeepSeek-V3: ããŒã¹ Mixture of Expert (MoE) ã¢ã㫠⢠DeepSeek-R1-Zero: 匷ååŠç¿ã®ã¿ ⢠DeepSeek-R1: ã³ãŒã«ãã¹ã¿ãŒãããŒã¿ããŠåŒ·ååŠç¿ Distilled Models ⢠DeepSeek-R1-Distill-Qwen (1.5B, 7B, 14B, 32B) ⢠DeepSeek-R1-Distill-Llama (8B and 70B) DeepSeek ã¯ è€æ°ã®ã¿ã¹ã¯ã«ãã㊠é«åºŠãªè«ççæšè«èœåãçºæ® -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 17 äž»ãªèœå ⢠è€éãªåé¡è§£æ±ºã®ããã®é«ãè«ççæšè«èœå (æ°åŠãã³ãŒãã£ã³ã°ãžã®å¿çš) ⢠AIME 2024, MATH-500, and SWE-bench ãªã©ã® ãã³ãããŒã¯ã§é«ãæ§èœ ⢠671B ãã©ã¡ãŒã¿ã® Mixture of Experts (MoE) ã¢ãŒããã¯ã㣠⢠37B activation parameter ⢠DeepSeek-R1 ã®æšè«ã«ã¯ FP8 ã§å°ãªããšã 800 GB of HBM ãå¿ èŠ -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 18 Model Options ⢠ã¬ã€ãã³ã·ã»ã³ã¹ããæãã€ã€ ã³ã¢æ©èœãä¿ã£ãèžçã¢ãã« â¢ è€æ°ãµã€ãºãèŠä»¶ã«å¿ããŠéžæ ⢠DeepSeek-R1-Distill-Qwen (1.5B, 7B, 14B, 32B) ⢠DeepSeek-R1-Distill-Llama (8B, 70B) -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 19 AI/ML åã EC2 ã¢ã¯ã»ã©ã¬ãŒã¿ã€ã³ã¹ã¿ã³ã¹ G6 (L4) P5 (H100) DL1 G6e (L40S) P4 (A100) P5e (H200) Inf1 Inf2 P5en (H200) Trn1 GPUs AI/ML accelerators and ASICs Trn2 G5 (A10G) AWS Trainium, Inferentia H100, H200, B200, GB200, A100, L40S, L4, A10G Cloud AI100 Standard Radeon GPU Xilinx accelerator Xilinx FPGA DL2q Gaudi accelerator Announced GB200 B200 -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 20 CPU CPU NSC EBS Host EFA PCIe SSD EFA SSD ⊠Switching layer PCIe PCIe PCIe ML chip interconnect ML chip ML chip ML chip ML chip ⊠Accelerators Accelerated compute architecture -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 23 P5 ã€ã³ã¹ã¿ã³ã¹ Optimized for AI training and inference 900 GB/s NVSwitch for GPU peer-to-peer connections Scale-out with non-blocking interconnect Elastic Fabric Adapter (EFA) Instance GPU GPU memory CPU vCPU Instance memory Networking Local storage P5 8 NVIDIA H100 640 GB AMD Milan 192 2 TB 3200 Gbps EFAv2 30 TB SSD P5e 8 NVIDIA H200 1128 GB AMD Milan 192 2 TB 3200 Gbps EFAv2 30 TB SSD P5en 8 NVIDIA H200 1128 GB Intel SPR 192 2 TB 3200 Gbps EFAv3 30 TB SSD -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 24 G5/G6 ã€ã³ã¹ã¿ã³ã¹ G Compute and graphics optimized GPUs Flexibility with multiple instance sizes Great for single GPU or single node workloads Instance GPU GPU memory CPU vCPU Instance memory Networking Local storage G5 Up to 8 NVIDIA A10G Up to 192 GB AMD Rome Up to 192 Up to 768 GB Up to 100 Gbps Up to 7.6 TB SSD G6 Up to 8 NVIDIA L4 Up to 192 GB GDDR6 AMD Milan Up to 192 Up to 768 GB Up to 100 Gbps Up to 7.6 TB SSD G6e Up to 8 NVIDIA L40S Up to 384 GB GDDR6 AMD Milan Up to 192 Up to 1.536 TB Up to 400 Gbps Up to 7.6 TB SSD -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 25 Bedrock Marketplace implementation ⢠Bedrock Marketplace ã§ DeepSeek-R1 ãšèžçã¢ãã«ãå©çšå¯èœ ⢠簡åããã〠⢠Bedrock ã®ã»ãã¥ãªãã£ã»ã¢ãã¿ãªã³ã°æ©èœã å©çšã§ãã -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 26 Bedrock Marketplace delivers 100+ models from 30+ providers EVOLUTIONARY SCALE WIDN CAMB.AI GRETEL ARCEE AI PREFERRED NETWORKS WRITER UPSTAGE NCSOFT STOCKMARK KARAKURI JOHN SNOW LABS LIQUID DATABRICKS CYBERAGENT HUGGING FACE STABILITY AI LG AI RESEARCH M I S T R A L AI SNOWFLAKE N V I D I A DEEPSEEK -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 27 äºåæºå: ãããã€åã«äžéç·©å (R1: ml.p5e.48xlarge) -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 28 Step1: Model catalog ã§ DeepSeek-R1 ã¢ãã«ãèŠã€ãã -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 29 Step2: ããã〠-
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 30 Step3: Playground ã InvokeModel API ã§éã¶ -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 32 ãã³ãºãªã³1: Amazon Bedrock Marketplace ⢠Amazon Bedrock ãéã ⢠Model catalog ãéã ⢠DeepSeek ãšæã¡èŸŒãã§ãã£ã«ã¿ãŒ ⢠DeepSeek-R1-Distill-Llama-8B ãªã©ãéžãã§ããã〠⢠Marketplace deployments ã«è¡šç€ºãããŠãã¢ãã«ãéžã³ Playground ⢠https://gist.github.com/hariby/c6b4d1f7ceee8e976b8b752c388d7ae5 -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 36 Tips: é©å㪠chat template (model tokenizer) ãå©çš 36 <ïœbeginâofâsentenceïœ><ïœUserïœ>A man has 53 socks in his drawer: 21 identical blue, 15 identical black and 17 identical red. The lights are out, and he is completely in the dark. How many socks must he take out to make 100 percent certain he has at least one pair of black socks?<ïœAssistantïœ> Bedrock Playground ã§äœ¿ãéã«ã¯ãé©å㪠chat template ã¿ã°ãã€ããå¿ èŠããã: InvokeModel API ã䜿ãå Žåãé©å㪠tokenizer ã䜿ãå¿ èŠããã: tokenizer = AutoTokenizer.from_pretrained(hf_model_id) messages = [{"role": "user", "content": test_prompt}] prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=not continuation) Bad quality output Good quality output -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 37 DeepSeek-R1 å©çšã«ãããã責任ãã AIã 37 (through the ApplyGuardrail API) can provide an extra layer of security and responsible AI measures -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 38 Enterprise Protection ⢠Enterprise-grade security features built-in ⢠Complete data privacy when using AWS services ⢠No data sharing with model providers ⢠End-to-end encryption for all operations ⢠Access controls and governance features ⢠Compliance with AWS security standards -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 39 Critical Concerns ⢠Models hosted by AWS without any communication with DeepSeek servers or APIs ⢠No customer data used to improve base models ⢠Enterprise data protection capabilities ⢠Privacy control through AWS services -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 43 Custom Model Import implementation ⢠Bedrock Custom Model Import enables DeepSeek deployment ⢠Support for Llama 8B and 70B distilled DeepSeek R1 variants ⢠Complete code samples and step-by-step deployment guides provided for quick implementation ⢠Standard Bedrock security and monitoring features ⢠Pricing is on-demand in 5-minute window from first successful invocation ⢠There is a cold-start and scaling up/down time -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 49 Customizing & Deploying models with Amazon SageMakerAI Select Evaluate Customize Deploy Amazon Build, train, and deploy ML modelsâincluding FMsâfor any use case with fully managed infrastructure, tools, and workflows Data Scientists and ML Engineers Bring Your Own Model Bring Your Own Container Choice of hundreds of models from SageMaker JumpStart Automated & Human evaluation of models Customize models to your use cases by pre-training, fine-tuning, model distillation, etc. Optimize and deploy for inference -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 61 Trn1/Trn2 ã€ã³ã¹ã¿ã³ã¹ Powered by AWS Trainium/Trainium2 custom ML chips Optimized for large-scale training distributed workloads Trn2 Ultraservers with extended NeuronLink for trillion-parameter AI Neuron Kernel Interface (NKI) for custom operators Instance Accelerators Accelerator memory vCPU Instance memory Networking trn1.32xlarge 16 512 GB 128 512 GB 800 Gbps EFAv2 trn1n.32xlarge 16 512 GB 128 512 GB 1600 Gbps EFAv2 trn2.48xlarge 16 1.5 TB 192 2 TB 3.2 Tbps EFAv3 -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 62 AWS Trainium/Inferentia ã¢ãŒããã¯ã㣠⢠Tensor engine are based on power-optimized systolic array ⢠AWS Neuron SDK supports typical architecutres such as Llama -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 63 ãã³ãºãªã³2: Amazon SageMaker AI ⢠SageMaker AI ã®ããŒãžãéã ⢠Notebook ã€ã³ã¹ã¿ã³ã¹ãäœã (CPU ã€ã³ã¹ã¿ã³ã¹ã§ ok) ⢠Hugging Face ã§äœ¿ãããã¢ãã«ãéžã³ã Deploy > Amazon SageMaker ãã Python ã¹ã¯ãªãããã³ã㌠https://huggingface.co/collections/deepseek-ai/deepseek-r1 ⢠deepseek-ai/DeepSeek-R1-Distill-Llama-8B ⢠Noteobok ã€ã³ã¹ã¿ã³ã¹äžãã SageMaker Python SDK ã§ ãšã³ããã€ã³ãã« DeepSeek-Distill ã¢ãã«ããã〠-
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 64 ãã³ãºãªã³2: Amazon SageMaker AI ⢠SageMaker Notebook ã€ã³ã¹ã¿ã³ã¹ã® Jupyter Notebook ã« Python ã¹ã¯ãªããã貌ãä»ãã ⢠huggingface_model.deploy: ãããã€, predictor.predict: æšè« -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 66 DeepSeek-R1 on AWS ãŸãšã 1. Amazon Bedrock Marketplace (Amazon SageMaker JumpStart) ã§ DeepSeek-R1/Distill ã¢ãã«ã®ããã〠2. Amazon SageMaker AI Inf2 ã€ã³ã¹ã¿ã³ã¹ãž DeepSeek-R1-Distill ã¢ãã«ã®ããã〠3. Amazon Bedrock Custom Model Import ã§ DeepSeek-R1-Distill ã¢ãã«ã®ããã〠DeepSeek on AWS ããã°ããããŸã â https://aws.amazon.com/jp/blogs/news/deepseek-r1-models-now-available-on-aws/ -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 67 AWS ã®ããã®âœ£æ AI ã¢ããªæ§ç¯å®è·µã¬ã€ã 察象èªè ïžâœ£æ AI ã®æ¬æ ŒæŽ»âœ€ãæ€èšããŠããæè¡è ✣æ AI ã¢ããªãæ§ç¯ããããã®åºç€æŠå¿µã解説 (ããã³ãããšã³ãžãã¢ãªã³ã°ãRAGããšãŒãžã§ã³ã) åºç€æŠå¿µã®å¿âœ€ãããããã®ããå®è·µçãªãã³ãºãªã³ (RAGããšãŒãžã§ã³ã) æ¬çªå°âŒããããã®ãã€ã³ãã解説 (責任ãã AI, Working Backwards, etc.) 67 2025幎æ¥é çºå£²äºå®ïžAmazon ã§äºçŽã§ããŸãïž https://www.amazon.co.jp/dp/4296205234 -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 69 Thank you! © 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. Yoshitaka Haribara, Ph.D. X: @_hariby -
© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. 70 Further reading ⢠DeepSeek ⢠Anthropic CEO Dario Blog ⢠https://darioamodei.com/on-deepseek-and-export-controls ⢠Startup Customer Case Studies on AWS ⢠Sakana AI ⢠https://aws.amazon.com/startups/learn/letting-nature-lead-how-sakana-ai-is- transforming-model-building?lang=en-US ⢠ELYZA (Llama2 Speculative Decoding on AWS Inferentia2 chip) ⢠https://aws.amazon.com/jp/blogs/startup/tech-interview-elyza-2024/ ⢠LLM Development on Trn1 ⢠https://aws.amazon.com/jp/blogs/machine-learning/unlocking-japanese-llms- with-aws-trainium-innovators-showcase-from-the-aws-llm-development-support- program/