RichardErkhov/habanoz_-_TinyLlama-1.1B-step-2T-lr-5-5ep-oasst1-top1-instruct-V1-gguf

Quantization made by Richard Erkhov.

Github

Discord

Request more models

TinyLlama-1.1B-step-2T-lr-5-5ep-oasst1-top1-instruct-V1 - GGUF

• —Model creator: https://huggingface.co/habanoz/
• —Original model: https://huggingface.co/habanoz/TinyLlama-1.1B-step-2T-lr-5-5ep-oasst1-top1-instruct-V1/

Original model description: --- language:

• —en license: apache-2.0 datasets:
• —OpenAssistant/oassttop12023-08-25 pipelinetag: text-generation basemodel: TinyLlama/TinyLlama-1.1B-intermediate-step-955k-token-2T model-index:
• —name: TinyLlama-1.1B-step-2T-lr-5-5ep-oasst1-top1-instruct-V1 results:
• —task: type: text-generation name: Text Generation dataset: name: AI2 Reasoning Challenge (25-Shot) type: ai2arc config: ARC-Challenge split: test args: numfew_shot: 25 metrics:
• —type: accnorm value: 31.06 name: normalized accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/openllm_leaderboard?query=habanoz/TinyLlama-1.1B-step-2T-lr-5-5ep-oasst1-top1-instruct-V1 name: Open LLM Leaderboard
• —task: type: text-generation name: Text Generation dataset: name: HellaSwag (10-Shot) type: hellaswag split: validation args: numfewshot: 10 metrics:
• —type: accnorm value: 55.02 name: normalized accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/openllm_leaderboard?query=habanoz/TinyLlama-1.1B-step-2T-lr-5-5ep-oasst1-top1-instruct-V1 name: Open LLM Leaderboard
• —task: type: text-generation name: Text Generation dataset: name: MMLU (5-Shot) type: cais/mmlu config: all split: test args: numfewshot: 5 metrics:
• —type: acc value: 26.41 name: accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/openllmleaderboard?query=habanoz/TinyLlama-1.1B-step-2T-lr-5-5ep-oasst1-top1-instruct-V1 name: Open LLM Leaderboard
• —task: type: text-generation name: Text Generation dataset: name: TruthfulQA (0-shot) type: truthfulqa config: multiplechoice split: validation args: numfewshot: 0 metrics:
• —type: mc2 value: 35.08 source: url: https://huggingface.co/spaces/HuggingFaceH4/openllmleaderboard?query=habanoz/TinyLlama-1.1B-step-2T-lr-5-5ep-oasst1-top1-instruct-V1 name: Open LLM Leaderboard
• —task: type: text-generation name: Text Generation dataset: name: Winogrande (5-shot) type: winogrande config: winograndexl split: validation args: numfew_shot: 5 metrics:
• —type: acc value: 58.01 name: accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/openllmleaderboard?query=habanoz/TinyLlama-1.1B-step-2T-lr-5-5ep-oasst1-top1-instruct-V1 name: Open LLM Leaderboard
• —task: type: text-generation name: Text Generation dataset: name: GSM8k (5-shot) type: gsm8k config: main split: test args: numfewshot: 5 metrics:
• —type: acc value: 1.59 name: accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/openllmleaderboard?query=habanoz/TinyLlama-1.1B-step-2T-lr-5-5ep-oasst1-top1-instruct-V1 name: Open LLM Leaderboard ---

TinyLlama/TinyLlama-1.1B-intermediate-step-955k-token-2T finetuned using OpenAssistant/oassttop12023-08-25 dataset.

Trained for 5 epochs using Qlora. Adapter is merged.

SFT code: https://github.com/habanoz/qlora.git

Command used:

accelerate launch $BASE_DIR/qlora/train.py \
  --model_name_or_path $BASE_MODEL \
  --working_dir $BASE_DIR/$OUTPUT_NAME-checkpoints \
  --output_dir $BASE_DIR/$OUTPUT_NAME-peft \
  --merged_output_dir $BASE_DIR/$OUTPUT_NAME \
  --final_output_dir $BASE_DIR/$OUTPUT_NAME-final \
  --num_train_epochs 5 \
  --logging_steps 1 \
  --save_strategy steps \
  --save_steps 75 \
  --save_total_limit 2 \
  --data_seed 11422 \
  --evaluation_strategy steps \
  --per_device_eval_batch_size 4 \
  --eval_dataset_size 0.01 \
  --eval_steps 75 \
  --max_new_tokens 1024 \
  --dataloader_num_workers 3 \
  --logging_strategy steps \
  --do_train \
  --do_eval \
  --lora_r 64 \
  --lora_alpha 16 \
  --lora_modules all \
  --bits 4 \
  --double_quant \
  --quant_type nf4 \
  --lr_scheduler_type constant \
  --dataset oasst1-top1 \
  --dataset_format oasst1 \
  --model_max_len 1024 \
  --per_device_train_batch_size 4 \
  --gradient_accumulation_steps 4 \
  --learning_rate 1e-5 \
  --adam_beta2 0.999 \
  --max_grad_norm 0.3 \
  --lora_dropout 0.0 \
  --weight_decay 0.0 \
  --seed 11422 \
  --gradient_checkpointing \
  --use_flash_attention_2 \
  --ddp_find_unused_parameters False

Open LLM Leaderboard Evaluation Results

Detailed results can be found here