See axolotl config

axolotl version: 0.6.0

base_model: mistralai/Mistral-7B-v0.1
# optionally might have model_type or tokenizer_type
model_type: MistralForCausalLM
tokenizer_type: LlamaTokenizer
# Automatically upload checkpoint and final model to HF
hub_model_id: AiAF/Mistral-QLoRA-Pretraining-Test-v1.2

load_in_8bit: false
load_in_4bit: true
strict: false

datasets:
  - path: AiAF/pretraining.jsonl
    type: completion

dataset_prepared_path: last_run_prepared
val_set_size: 0.1
output_dir: /workspace/axolotl/outputs/qlora-out/Mistral-QLoRA-Pretraining-Test-V1.2

adapter: qlora
lora_model_dir:

sequence_len: 8192
sample_packing: false
pad_to_sequence_len: true

lora_r: 64
lora_alpha: 128
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:
lora_target_modules:
  - gate_proj
  - down_proj
  - up_proj
  - q_proj
  - v_proj
  - k_proj
  - o_proj

wandb_project: "LLM_QLoRA-Pretraining-Practice"
wandb_entity:
wandb_watch: "all"
wandb_name: "Mistral-QLoRA-Pretraining-Test-V1.2"
wandb_log_model: "false"

gradient_accumulation_steps: 32
micro_batch_size: 2
num_epochs: 1
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.000005

train_on_inputs: false
group_by_length: false
bf16: auto
fp16:
tf32: false

gradient_checkpointing: true
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true

loss_watchdog_threshold: 5.0
loss_watchdog_patience: 3

warmup_steps: 1
evals_per_epoch: 4
eval_table_size:
eval_max_new_tokens: 128
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:

Mistral-QLoRA-Pretraining-Test-v1.2

This model is a fine-tuned version of mistralai/Mistral-7B-v0.1 on the AiAF/pretraining.jsonl dataset. It achieves the following results on the evaluation set:

• Loss: 1.8778

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-06
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 32
• total_train_batch_size: 64
• optimizer: Use OptimizerNames.ADAMW_BNB with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 2
• num_epochs: 1.0

Training results

Training Loss	Epoch	Step	Validation Loss
1.9625	0.9143	1	1.8778

Framework versions

• PEFT 0.14.0
• Transformers 4.48.3
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0