Retire llama3.1 405B and DLRM-DCNv2 references for the v6.1 round. (#901)

Move the retired implementations under retired_benchmarks, add DLRMv4 to the v6.1 README table, and point historical README links at the new paths.

Co-authored-by: Cursor <cursoragent@cursor.com>

Author: ShriyaRishab

README.md

@@ -52,6 +52,7 @@ Some these benchmarks are rather slow or take a long time to run on the referenc
 | gpt_oss_20b | [small_llm_moe_pretraining](https://github.com/mlcommons/training/tree/master/small_llm_moe_pretraining/primus) | Primus | C4 | 20B |
 | deepseekv3 | [llm_moe_pretraining](https://github.com/mlcommons/training/tree/master/llm_moe_pretraining/nemo) | NeMo | C4 | 671B |
 | qwen35_397b_grpo | [llm_post_training](https://github.com/mlcommons/training/tree/master/llm_post_training) | NeMo-RL / NeMo-Gym | SWE tasks | 397B
+| dlrm_v4_hstu | [recommendation](https://github.com/mlcommons/training/tree/master/recommendation) | pytorch | Yambda-5b | HSTU
 
 *Framework here is given for the reference implementation. Submitters are free to use their own frameworks to run the benchmark.
 
@@ -64,8 +65,8 @@ Some these benchmarks are rather slow or take a long time to run on the referenc
 | flux.1 | [text_to_image](https://github.com/mlcommons/training/tree/master/text_to_image) | torchtitan | CC12M subset | 11.9B
 | llama3.1_8b | [small_llm_pretraining](https://github.com/mlcommons/training/tree/master/small_llm_pretraining) | NeMo | C4 | 8b
 | llama2_70b_lora | [llama2_70b_lora](https://github.com/mlcommons/training/tree/master/llama2_70b_lora) | pytorch | SCROLLS GovReport | 70B
-| llama3.1_405b | [large_language_model_pretraining](https://github.com/mlcommons/training/tree/master/large_language_model_pretraining) | NeMo | C4 | 405B
-| dlrm_dcnv2 | [recommendation_v2](https://github.com/mlcommons/training/tree/master/recommendation_v2/torchrec_dlrm) | torchrec | Criteo 3.5TB multi-hot | 167M
+| llama3.1_405b | [large_language_model_pretraining](https://github.com/mlcommons/training/tree/master/retired_benchmarks/large_language_model_pretraining) | NeMo | C4 | 405B
+| dlrm_dcnv2 | [recommendation_v2](https://github.com/mlcommons/training/tree/master/retired_benchmarks/recommendation_v2/torchrec_dlrm) | torchrec | Criteo 3.5TB multi-hot | 167M
 | gpt_oss_20b | [small_llm_moe_pretraining](https://github.com/mlcommons/training/tree/master/small_llm_moe_pretraining/primus) | Primus | C4 | 20B |
 | deepseekv3 | [llm_moe_pretraining](https://github.com/mlcommons/training/tree/master/llm_moe_pretraining/nemo) | NeMo | C4 | 671B |
 
@@ -81,8 +82,8 @@ Some these benchmarks are rather slow or take a long time to run on the referenc
 | flux.1 | [text_to_image](https://github.com/mlcommons/training/tree/master/text_to_image) | torchtitan | CC12M subset | 11.9B
 | llama3.1_8b | [small_llm_pretraining](https://github.com/mlcommons/training/tree/master/small_llm_pretraining) | NeMo | C4 | 8b
 | llama2_70b_lora | [llama2_70b_lora](https://github.com/mlcommons/training/tree/master/llama2_70b_lora) | pytorch | SCROLLS GovReport | 70B
-| llama3.1_405b | [large_language_model_pretraining](https://github.com/mlcommons/training/tree/master/large_language_model_pretraining) | NeMo | C4 | 405B
-| dlrm_dcnv2 | [recommendation_v2](https://github.com/mlcommons/training/tree/master/recommendation_v2/torchrec_dlrm) | torchrec | Criteo 3.5TB multi-hot | 167M
+| llama3.1_405b | [large_language_model_pretraining](https://github.com/mlcommons/training/tree/master/retired_benchmarks/large_language_model_pretraining) | NeMo | C4 | 405B
+| dlrm_dcnv2 | [recommendation_v2](https://github.com/mlcommons/training/tree/master/retired_benchmarks/recommendation_v2/torchrec_dlrm) | torchrec | Criteo 3.5TB multi-hot | 167M
 | rgat | [graph_neural_network](https://github.com/mlcommons/training/tree/master/graph_neural_network) | GLT | IGBH-Full | 25M
 
 *Framework here is given for the reference implementation. Submitters are free to use their own frameworks to run the benchmark.
@@ -99,9 +100,9 @@ Some these benchmarks are rather slow or take a long time to run on the referenc
 | retinanet | [single_stage_detector](https://github.com/mlcommons/training/tree/master/single_stage_detector) | pytorch | OpenImages | 37M
 | Stable Diffusionv2 | [image generation](https://github.com/mlcommons/training/tree/master/stable_diffusion) | pytorch | LAION-400M-filtered | 865M
 | bert | [language_model](https://github.com/mlcommons/training/tree/master/language_model/tensorflow/bert) | tensorflow | Wikipedia 2020/01/01 | 340M
-| llama3.1_405b | [large_language_model_pretraining](https://github.com/mlcommons/training/tree/master/large_language_model_pretraining) | NeMo | C4 | 405B
+| llama3.1_405b | [large_language_model_pretraining](https://github.com/mlcommons/training/tree/master/retired_benchmarks/large_language_model_pretraining) | NeMo | C4 | 405B
 | llama2_70b_lora | [llama2_70b_lora](https://github.com/mlcommons/training/tree/master/llama2_70b_lora) | pytorch | SCROLLS GovReport | 70B
-| dlrm_dcnv2 | [recommendation_v2](https://github.com/mlcommons/training/tree/master/recommendation_v2/torchrec_dlrm) | torchrec | Criteo 3.5TB multi-hot | 167M
+| dlrm_dcnv2 | [recommendation_v2](https://github.com/mlcommons/training/tree/master/retired_benchmarks/recommendation_v2/torchrec_dlrm) | torchrec | Criteo 3.5TB multi-hot | 167M
 | rgat | [graph_neural_network](https://github.com/mlcommons/training/tree/master/graph_neural_network) | GLT | IGBH-Full | 25M
 
 # MLPerf Training v4.1 (Submission Deadline Oct 11, 2024)
@@ -114,7 +115,7 @@ Some these benchmarks are rather slow or take a long time to run on the referenc
 | BERT-large | [language/nlp](https://github.com/mlcommons/training/tree/master/language_model/tensorflow/bert) | tensorflow | Wikipedia 2020/01/01 | 340M
 | GPT3 | [language/llm](https://github.com/mlcommons/training/tree/master/large_language_model) | paxml,megatron-lm | C4 | 175B
 | LLama2 70B-LoRA | [language/LLM fine-tuning](https://github.com/mlcommons/training/tree/master/llama2_70b_lora) | pytorch | SCROLLS GovReport | 70B
-| DLRMv2 | [recommendation](https://github.com/mlcommons/training/tree/master/recommendation_v2/torchrec_dlrm) | torchrec | Criteo 3.5TB multi-hot | 167M
+| DLRMv2 | [recommendation](https://github.com/mlcommons/training/tree/master/retired_benchmarks/recommendation_v2/torchrec_dlrm) | torchrec | Criteo 3.5TB multi-hot | 167M
 | RGAT | [GNN](https://github.com/mlcommons/training/tree/master/graph_neural_network) | pytorch | IGBH-Full | 25M 
 
 *Note model parameter count is not the same as active parameter that are being trained in the benchmark. 
@@ -131,7 +132,7 @@ Some these benchmarks are rather slow or take a long time to run on the referenc
 | BERT-large | [language/nlp](https://github.com/mlcommons/training/tree/master/language_model/tensorflow/bert) | tensorflow | Wikipedia 2020/01/01
 | GPT3 | [language/llm](https://github.com/mlcommons/training/tree/master/large_language_model) | paxml,megatron-lm | C4
 | LLama2 70B-LoRA | [language/LLM fine-tuning](https://github.com/mlcommons/training/tree/master/llama2_70b_lora) | pytorch | SCROLLS GovReport
-| DLRMv2 | [recommendation](https://github.com/mlcommons/training/tree/master/recommendation_v2/torchrec_dlrm) | torchrec | Criteo 3.5TB multi-hot
+| DLRMv2 | [recommendation](https://github.com/mlcommons/training/tree/master/retired_benchmarks/recommendation_v2/torchrec_dlrm) | torchrec | Criteo 3.5TB multi-hot
 | RGAT | [GNN](https://github.com/mlcommons/training/tree/master/graph_neural_network) | pytorch | IGBFull