id stringlengths 22 22 | instruction stringlengths 180 263 | input stringclasses 1
value | output stringlengths 460 1.77k | accelerator stringclasses 2
values | category stringclasses 12
values | model stringclasses 11
values | task stringclasses 8
values | dataset stringclasses 8
values | optimizer stringclasses 5
values |
|---|---|---|---|---|---|---|---|---|---|
kaggle_master_v2_00000 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – HuggingFaceH4/ultrachat_200k – optimizer adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 0 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | HuggingFaceH4/ultrachat_200k | adamw_8bit | |
kaggle_master_v2_00001 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – HuggingFaceH4/ultrachat_200k – optimizer paged_adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 1 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | HuggingFaceH4/ultrachat_200k | paged_adamw_8bit | |
kaggle_master_v2_00002 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – HuggingFaceH4/ultrachat_200k – optimizer adamw_torch_fused – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 2 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | HuggingFaceH4/ultrachat_200k | adamw_torch_fused | |
kaggle_master_v2_00003 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – HuggingFaceH4/ultrachat_200k – optimizer lion – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 3 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | HuggingFaceH4/ultrachat_200k | lion | |
kaggle_master_v2_00004 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – HuggingFaceH4/ultrachat_200k – optimizer adafactor – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 4 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | HuggingFaceH4/ultrachat_200k | adafactor | |
kaggle_master_v2_00005 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – tatsu-lab/alpaca – optimizer adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 5 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | tatsu-lab/alpaca | adamw_8bit | |
kaggle_master_v2_00006 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – tatsu-lab/alpaca – optimizer paged_adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 6 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | tatsu-lab/alpaca | paged_adamw_8bit | |
kaggle_master_v2_00007 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – tatsu-lab/alpaca – optimizer adamw_torch_fused – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 7 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | tatsu-lab/alpaca | adamw_torch_fused | |
kaggle_master_v2_00008 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – tatsu-lab/alpaca – optimizer lion – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 8 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | tatsu-lab/alpaca | lion | |
kaggle_master_v2_00009 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – tatsu-lab/alpaca – optimizer adafactor – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 9 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | tatsu-lab/alpaca | adafactor | |
kaggle_master_v2_00010 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – databricks/databricks-dolly-15k – optimizer adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 10 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | databricks/databricks-dolly-15k | adamw_8bit | |
kaggle_master_v2_00011 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – databricks/databricks-dolly-15k – optimizer paged_adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 11 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | databricks/databricks-dolly-15k | paged_adamw_8bit | |
kaggle_master_v2_00012 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – databricks/databricks-dolly-15k – optimizer adamw_torch_fused – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 12 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | databricks/databricks-dolly-15k | adamw_torch_fused | |
kaggle_master_v2_00013 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – databricks/databricks-dolly-15k – optimizer lion – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 13 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | databricks/databricks-dolly-15k | lion | |
kaggle_master_v2_00014 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – databricks/databricks-dolly-15k – optimizer adafactor – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 14 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | databricks/databricks-dolly-15k | adafactor | |
kaggle_master_v2_00015 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – allenai/c4 – optimizer adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 15 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | allenai/c4 | adamw_8bit | |
kaggle_master_v2_00016 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – allenai/c4 – optimizer paged_adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 16 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | allenai/c4 | paged_adamw_8bit | |
kaggle_master_v2_00017 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – allenai/c4 – optimizer adamw_torch_fused – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 17 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | allenai/c4 | adamw_torch_fused | |
kaggle_master_v2_00018 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – allenai/c4 – optimizer lion – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 18 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | allenai/c4 | lion | |
kaggle_master_v2_00019 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – allenai/c4 – optimizer adafactor – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 19 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | allenai/c4 | adafactor | |
kaggle_master_v2_00020 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – wikitext – optimizer adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 20 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | wikitext | adamw_8bit | |
kaggle_master_v2_00021 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – wikitext – optimizer paged_adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 21 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | wikitext | paged_adamw_8bit | |
kaggle_master_v2_00022 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – wikitext – optimizer adamw_torch_fused – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 22 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | wikitext | adamw_torch_fused | |
kaggle_master_v2_00023 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – wikitext – optimizer lion – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 23 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | wikitext | lion | |
kaggle_master_v2_00024 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – wikitext – optimizer adafactor – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 24 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | wikitext | adafactor | |
kaggle_master_v2_00025 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – lmsys/chatbot_arena_conversations – optimizer adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 25 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | lmsys/chatbot_arena_conversations | adamw_8bit | |
kaggle_master_v2_00026 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – lmsys/chatbot_arena_conversations – optimizer paged_adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 26 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | lmsys/chatbot_arena_conversations | paged_adamw_8bit | |
kaggle_master_v2_00027 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – lmsys/chatbot_arena_conversations – optimizer adamw_torch_fused – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 27 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | lmsys/chatbot_arena_conversations | adamw_torch_fused | |
kaggle_master_v2_00028 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – lmsys/chatbot_arena_conversations – optimizer lion – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 28 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | lmsys/chatbot_arena_conversations | lion | |
kaggle_master_v2_00029 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – lmsys/chatbot_arena_conversations – optimizer adafactor – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 29 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | lmsys/chatbot_arena_conversations | adafactor | |
kaggle_master_v2_00030 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – openassistant/oasst1 – optimizer adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 30 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | openassistant/oasst1 | adamw_8bit | |
kaggle_master_v2_00031 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – openassistant/oasst1 – optimizer paged_adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 31 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | openassistant/oasst1 | paged_adamw_8bit | |
kaggle_master_v2_00032 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – openassistant/oasst1 – optimizer adamw_torch_fused – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 32 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | openassistant/oasst1 | adamw_torch_fused | |
kaggle_master_v2_00033 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – openassistant/oasst1 – optimizer lion – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 33 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | openassistant/oasst1 | lion | |
kaggle_master_v2_00034 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – openassistant/oasst1 – optimizer adafactor – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 34 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | openassistant/oasst1 | adafactor | |
kaggle_master_v2_00035 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – mosaicml/dolly_hhrlhf – optimizer adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 35 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | mosaicml/dolly_hhrlhf | adamw_8bit | |
kaggle_master_v2_00036 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – mosaicml/dolly_hhrlhf – optimizer paged_adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 36 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | mosaicml/dolly_hhrlhf | paged_adamw_8bit | |
kaggle_master_v2_00037 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – mosaicml/dolly_hhrlhf – optimizer adamw_torch_fused – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 37 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | mosaicml/dolly_hhrlhf | adamw_torch_fused | |
kaggle_master_v2_00038 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – mosaicml/dolly_hhrlhf – optimizer lion – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 38 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | mosaicml/dolly_hhrlhf | lion | |
kaggle_master_v2_00039 | Kaggle t4_dual master v2: SFT instruction tuning – meta-llama/Llama-3.1-8B-Instruct – mosaicml/dolly_hhrlhf – optimizer adafactor – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 39 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – SFT instruction tuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# d... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | SFT instruction tuning | mosaicml/dolly_hhrlhf | adafactor | |
kaggle_master_v2_00040 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – HuggingFaceH4/ultrachat_200k – optimizer adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 40 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | HuggingFaceH4/ultrachat_200k | adamw_8bit | |
kaggle_master_v2_00041 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – HuggingFaceH4/ultrachat_200k – optimizer paged_adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 41 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | HuggingFaceH4/ultrachat_200k | paged_adamw_8bit | |
kaggle_master_v2_00042 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – HuggingFaceH4/ultrachat_200k – optimizer adamw_torch_fused – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 42 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | HuggingFaceH4/ultrachat_200k | adamw_torch_fused | |
kaggle_master_v2_00043 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – HuggingFaceH4/ultrachat_200k – optimizer lion – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 43 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | HuggingFaceH4/ultrachat_200k | lion | |
kaggle_master_v2_00044 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – HuggingFaceH4/ultrachat_200k – optimizer adafactor – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 44 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | HuggingFaceH4/ultrachat_200k | adafactor | |
kaggle_master_v2_00045 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – tatsu-lab/alpaca – optimizer adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 45 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | tatsu-lab/alpaca | adamw_8bit | |
kaggle_master_v2_00046 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – tatsu-lab/alpaca – optimizer paged_adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 46 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | tatsu-lab/alpaca | paged_adamw_8bit | |
kaggle_master_v2_00047 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – tatsu-lab/alpaca – optimizer adamw_torch_fused – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 47 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | tatsu-lab/alpaca | adamw_torch_fused | |
kaggle_master_v2_00048 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – tatsu-lab/alpaca – optimizer lion – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 48 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | tatsu-lab/alpaca | lion | |
kaggle_master_v2_00049 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – tatsu-lab/alpaca – optimizer adafactor – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 49 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | tatsu-lab/alpaca | adafactor | |
kaggle_master_v2_00050 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – databricks/databricks-dolly-15k – optimizer adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 50 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | databricks/databricks-dolly-15k | adamw_8bit | |
kaggle_master_v2_00051 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – databricks/databricks-dolly-15k – optimizer paged_adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 51 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | databricks/databricks-dolly-15k | paged_adamw_8bit | |
kaggle_master_v2_00052 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – databricks/databricks-dolly-15k – optimizer adamw_torch_fused – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 52 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | databricks/databricks-dolly-15k | adamw_torch_fused | |
kaggle_master_v2_00053 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – databricks/databricks-dolly-15k – optimizer lion – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 53 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | databricks/databricks-dolly-15k | lion | |
kaggle_master_v2_00054 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – databricks/databricks-dolly-15k – optimizer adafactor – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 54 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | databricks/databricks-dolly-15k | adafactor | |
kaggle_master_v2_00055 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – allenai/c4 – optimizer adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 55 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | allenai/c4 | adamw_8bit | |
kaggle_master_v2_00056 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – allenai/c4 – optimizer paged_adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 56 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | allenai/c4 | paged_adamw_8bit | |
kaggle_master_v2_00057 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – allenai/c4 – optimizer adamw_torch_fused – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 57 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | allenai/c4 | adamw_torch_fused | |
kaggle_master_v2_00058 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – allenai/c4 – optimizer lion – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 58 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | allenai/c4 | lion | |
kaggle_master_v2_00059 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – allenai/c4 – optimizer adafactor – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 59 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | allenai/c4 | adafactor | |
kaggle_master_v2_00060 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – wikitext – optimizer adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 60 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | wikitext | adamw_8bit | |
kaggle_master_v2_00061 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – wikitext – optimizer paged_adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 61 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | wikitext | paged_adamw_8bit | |
kaggle_master_v2_00062 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – wikitext – optimizer adamw_torch_fused – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 62 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | wikitext | adamw_torch_fused | |
kaggle_master_v2_00063 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – wikitext – optimizer lion – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 63 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | wikitext | lion | |
kaggle_master_v2_00064 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – wikitext – optimizer adafactor – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 64 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | wikitext | adafactor | |
kaggle_master_v2_00065 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – lmsys/chatbot_arena_conversations – optimizer adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 65 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | lmsys/chatbot_arena_conversations | adamw_8bit | |
kaggle_master_v2_00066 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – lmsys/chatbot_arena_conversations – optimizer paged_adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 66 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | lmsys/chatbot_arena_conversations | paged_adamw_8bit | |
kaggle_master_v2_00067 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – lmsys/chatbot_arena_conversations – optimizer adamw_torch_fused – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 67 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | lmsys/chatbot_arena_conversations | adamw_torch_fused | |
kaggle_master_v2_00068 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – lmsys/chatbot_arena_conversations – optimizer lion – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 68 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | lmsys/chatbot_arena_conversations | lion | |
kaggle_master_v2_00069 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – lmsys/chatbot_arena_conversations – optimizer adafactor – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 69 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | lmsys/chatbot_arena_conversations | adafactor | |
kaggle_master_v2_00070 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – openassistant/oasst1 – optimizer adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 70 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | openassistant/oasst1 | adamw_8bit | |
kaggle_master_v2_00071 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – openassistant/oasst1 – optimizer paged_adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 71 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | openassistant/oasst1 | paged_adamw_8bit | |
kaggle_master_v2_00072 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – openassistant/oasst1 – optimizer adamw_torch_fused – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 72 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | openassistant/oasst1 | adamw_torch_fused | |
kaggle_master_v2_00073 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – openassistant/oasst1 – optimizer lion – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 73 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | openassistant/oasst1 | lion | |
kaggle_master_v2_00074 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – openassistant/oasst1 – optimizer adafactor – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 74 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | openassistant/oasst1 | adafactor | |
kaggle_master_v2_00075 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – mosaicml/dolly_hhrlhf – optimizer adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 75 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | mosaicml/dolly_hhrlhf | adamw_8bit | |
kaggle_master_v2_00076 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – mosaicml/dolly_hhrlhf – optimizer paged_adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 76 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | mosaicml/dolly_hhrlhf | paged_adamw_8bit | |
kaggle_master_v2_00077 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – mosaicml/dolly_hhrlhf – optimizer adamw_torch_fused – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 77 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | mosaicml/dolly_hhrlhf | adamw_torch_fused | |
kaggle_master_v2_00078 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – mosaicml/dolly_hhrlhf – optimizer lion – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 78 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | mosaicml/dolly_hhrlhf | lion | |
kaggle_master_v2_00079 | Kaggle t4_dual master v2: QLoRA 4bit finetuning – meta-llama/Llama-3.1-8B-Instruct – mosaicml/dolly_hhrlhf – optimizer adafactor – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 79 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – QLoRA 4bit finetuning
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# di... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | QLoRA 4bit finetuning | mosaicml/dolly_hhrlhf | adafactor | |
kaggle_master_v2_00080 | Kaggle t4_dual master v2: DPO alignment – meta-llama/Llama-3.1-8B-Instruct – HuggingFaceH4/ultrachat_200k – optimizer adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 80 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – DPO alignment
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# dist.init_... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | DPO alignment | HuggingFaceH4/ultrachat_200k | adamw_8bit | |
kaggle_master_v2_00081 | Kaggle t4_dual master v2: DPO alignment – meta-llama/Llama-3.1-8B-Instruct – HuggingFaceH4/ultrachat_200k – optimizer paged_adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 81 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – DPO alignment
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# dist.init_... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | DPO alignment | HuggingFaceH4/ultrachat_200k | paged_adamw_8bit | |
kaggle_master_v2_00082 | Kaggle t4_dual master v2: DPO alignment – meta-llama/Llama-3.1-8B-Instruct – HuggingFaceH4/ultrachat_200k – optimizer adamw_torch_fused – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 82 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – DPO alignment
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# dist.init_... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | DPO alignment | HuggingFaceH4/ultrachat_200k | adamw_torch_fused | |
kaggle_master_v2_00083 | Kaggle t4_dual master v2: DPO alignment – meta-llama/Llama-3.1-8B-Instruct – HuggingFaceH4/ultrachat_200k – optimizer lion – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 83 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – DPO alignment
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# dist.init_... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | DPO alignment | HuggingFaceH4/ultrachat_200k | lion | |
kaggle_master_v2_00084 | Kaggle t4_dual master v2: DPO alignment – meta-llama/Llama-3.1-8B-Instruct – HuggingFaceH4/ultrachat_200k – optimizer adafactor – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 84 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – DPO alignment
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# dist.init_... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | DPO alignment | HuggingFaceH4/ultrachat_200k | adafactor | |
kaggle_master_v2_00085 | Kaggle t4_dual master v2: DPO alignment – meta-llama/Llama-3.1-8B-Instruct – tatsu-lab/alpaca – optimizer adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 85 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – DPO alignment
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# dist.init_... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | DPO alignment | tatsu-lab/alpaca | adamw_8bit | |
kaggle_master_v2_00086 | Kaggle t4_dual master v2: DPO alignment – meta-llama/Llama-3.1-8B-Instruct – tatsu-lab/alpaca – optimizer paged_adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 86 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – DPO alignment
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# dist.init_... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | DPO alignment | tatsu-lab/alpaca | paged_adamw_8bit | |
kaggle_master_v2_00087 | Kaggle t4_dual master v2: DPO alignment – meta-llama/Llama-3.1-8B-Instruct – tatsu-lab/alpaca – optimizer adamw_torch_fused – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 87 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – DPO alignment
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# dist.init_... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | DPO alignment | tatsu-lab/alpaca | adamw_torch_fused | |
kaggle_master_v2_00088 | Kaggle t4_dual master v2: DPO alignment – meta-llama/Llama-3.1-8B-Instruct – tatsu-lab/alpaca – optimizer lion – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 88 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – DPO alignment
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# dist.init_... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | DPO alignment | tatsu-lab/alpaca | lion | |
kaggle_master_v2_00089 | Kaggle t4_dual master v2: DPO alignment – meta-llama/Llama-3.1-8B-Instruct – tatsu-lab/alpaca – optimizer adafactor – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 89 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – DPO alignment
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# dist.init_... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | DPO alignment | tatsu-lab/alpaca | adafactor | |
kaggle_master_v2_00090 | Kaggle t4_dual master v2: DPO alignment – meta-llama/Llama-3.1-8B-Instruct – databricks/databricks-dolly-15k – optimizer adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 90 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – DPO alignment
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# dist.init_... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | DPO alignment | databricks/databricks-dolly-15k | adamw_8bit | |
kaggle_master_v2_00091 | Kaggle t4_dual master v2: DPO alignment – meta-llama/Llama-3.1-8B-Instruct – databricks/databricks-dolly-15k – optimizer paged_adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 91 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – DPO alignment
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# dist.init_... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | DPO alignment | databricks/databricks-dolly-15k | paged_adamw_8bit | |
kaggle_master_v2_00092 | Kaggle t4_dual master v2: DPO alignment – meta-llama/Llama-3.1-8B-Instruct – databricks/databricks-dolly-15k – optimizer adamw_torch_fused – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 92 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – DPO alignment
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# dist.init_... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | DPO alignment | databricks/databricks-dolly-15k | adamw_torch_fused | |
kaggle_master_v2_00093 | Kaggle t4_dual master v2: DPO alignment – meta-llama/Llama-3.1-8B-Instruct – databricks/databricks-dolly-15k – optimizer lion – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 93 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – DPO alignment
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# dist.init_... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | DPO alignment | databricks/databricks-dolly-15k | lion | |
kaggle_master_v2_00094 | Kaggle t4_dual master v2: DPO alignment – meta-llama/Llama-3.1-8B-Instruct – databricks/databricks-dolly-15k – optimizer adafactor – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 94 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – DPO alignment
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# dist.init_... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | DPO alignment | databricks/databricks-dolly-15k | adafactor | |
kaggle_master_v2_00095 | Kaggle t4_dual master v2: DPO alignment – meta-llama/Llama-3.1-8B-Instruct – allenai/c4 – optimizer adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 95 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – DPO alignment
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# dist.init_... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | DPO alignment | allenai/c4 | adamw_8bit | |
kaggle_master_v2_00096 | Kaggle t4_dual master v2: DPO alignment – meta-llama/Llama-3.1-8B-Instruct – allenai/c4 – optimizer paged_adamw_8bit – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 96 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – DPO alignment
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# dist.init_... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | DPO alignment | allenai/c4 | paged_adamw_8bit | |
kaggle_master_v2_00097 | Kaggle t4_dual master v2: DPO alignment – meta-llama/Llama-3.1-8B-Instruct – allenai/c4 – optimizer adamw_torch_fused – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 97 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – DPO alignment
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# dist.init_... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | DPO alignment | allenai/c4 | adamw_torch_fused | |
kaggle_master_v2_00098 | Kaggle t4_dual master v2: DPO alignment – meta-llama/Llama-3.1-8B-Instruct – allenai/c4 – optimizer lion – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 98 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – DPO alignment
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# dist.init_... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | DPO alignment | allenai/c4 | lion | |
kaggle_master_v2_00099 | Kaggle t4_dual master v2: DPO alignment – meta-llama/Llama-3.1-8B-Instruct – allenai/c4 – optimizer adafactor – bf16 – dual-T4 DDP with NCCL_P2P_DISABLE, Unsloth QLoRA r=64, CLI agentic kaggle kernels push – run 99 | # Kaggle T4x2 DDP – meta-llama/Llama-3.1-8B-Instruct – DPO alignment
import os, torch, torch.distributed as dist
os.environ["NCCL_P2P_DISABLE"]="1"
os.environ["NCCL_SHM_DISABLE"]="1"
os.environ["TOKENIZERS_PARALLELISM"]="false"
os.environ["CUDA_VISIBLE_DEVICES"]="0,1"
# torchrun --nproc_per_node=2 train.py
# dist.init_... | t4_dual | t4_ddp_master | meta-llama/Llama-3.1-8B-Instruct | DPO alignment | allenai/c4 | adafactor |
End of preview. Expand in Data Studio
YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
Single split, 25,000 examples. Recommended random split for SFT:
- train: 23,500
- validation: 1,500
Stratify by category if you want balanced accelerator coverage.
Category breakdown: Total: 20,778 T4 dual / 4,222 TPU v3-8
Models covered: Llama-3.1-8B, Mistral-7B-v0.3, Qwen2.5-7B, Gemma-2-9b, Phi-3-medium, Hermes-3, DeepSeek-LLM-7B, Yi-1.5-9B, Flan-T5-XXL, Pythia-2.8B
Tasks: SFT, QLoRA 4bit, DPO, ORPO, continued pretraining, classification, embedding training, vision-text finetuning
Usage
Load with datasets
SFT training – QLoRA on Kaggle T4x2
Training a 7B base on this dataset with QLoRA r=64 fits comfortably in a single Kaggle T4 (16GB). For faster training, use both T4s with Accelerate: accelerate launch --multi_gpu --num_processes=2 train.py
Prompt format
Alpaca-style:
Proven formulas encoded in every example
T4 dual bring-up
Unsloth QLoRA
TPU v3-8
Kaggle API agentic
Memory – 7B QLoRA on T4
∼4.4GB base (NF4) + 0.2GB adapters + 2GB activations ≈ 6.7GB → fits 1x T4
Effective batch
global_batch = per_device_batch × grad_accum_steps × world_size
Limitations
- Synthetic expert-curated. All 25,000 examples are programmatically generated from 100+ competition-proven templates, grounded in public Kaggle docs / kernels (Dec 2025 – May 2026). Not scraped from private notebooks. No PII, no secrets.
- Kaggle-specific. Paths, quotas, NCCL env vars, and API commands target Kaggle Notebooks (July 2026). Adapt for other platforms.
- Code is illustrative. Always review before running in production. Check
transformers,peft,trl,unsloth,torch_xlaversions – Kaggle images change. - English only.
Intended use: SFT / instruction-tuning LLMs to generate correct Kaggle T4 / TPU training code, debug DDP/TPU jobs, and drive the Kaggle API CLI agentically.
Citation
License
Apache-2.0 – commercial use allowed.
Acknowledgments
Built from public sources:
- Kaggle Docs – TPUs / GPUs – https://www.kaggle.com/docs/tpu
- PyTorch Distributed Data Parallel – Kaggle T4x2 guide – LearnOpenCV
- Unsloth – Fine-Tuning Qwen VL on a Single T4 – Towards AI, May 2026
- Kaggle API –
kaggle kernels push --accelerator NvidiaTeslaT4– https://github.com/Kaggle/kaggle-api - TRL – SFTTrainer with UnslothVisionDataCollator
Thanks to the Kaggle community for publishing competition kernels that made the ground-truth formulas possible.
- Downloads last month
- -