See axolotl config

axolotl version: 0.4.0

base_model: microsoft/Phi-3-mini-4k-instruct
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

load_in_8bit: false
load_in_4bit: false
strict: false

datasets:
  - path: last_65000.jsonl
    type: input_output

dataset_prepared_path:
val_set_size: 0.2
output_dir: /mnt/mlblob/Phi-3-mini-4k-instruct-function-calling
hub_model_id: rajdeepV/Phi-3-mini-4k-instruct-function-calling

sequence_len: 4096
sample_packing: true
pad_to_sequence_len: true
trust_remote_code: true

adapter:
lora_model_dir:
lora_r:
lora_alpha:
lora_dropout:
lora_target_linear:
lora_fan_in_fan_out:

wandb_project: phi3-func
wandb_entity: vapi
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 2
micro_batch_size: 2
num_epochs: 2
optimizer: adamw_torch
adam_beta2: 0.95
adam_epsilon: 0.00001
max_grad_norm: 1.0
lr_scheduler: cosine
learning_rate: 0.000003

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

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

warmup_steps: 100
evals_per_epoch: 4
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.1

# resize_token_embeddings_to_32x: true
special_tokens:
  pad_token: "<|endoftext|>"

Phi-3-mini-4k-instruct-function-calling

This model is a fine-tuned version of microsoft/Phi-3-mini-4k-instruct on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5893

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: 3e-06
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 4
• optimizer: Adam with betas=(0.9,0.95) and epsilon=1e-05
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 100
• num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss
1.5244	0.0012	1	0.7526
5.7823	0.2509	216	0.7009
5.4414	0.5017	432	0.6466
4.2933	0.7526	648	0.6110
4.7999	1.0035	864	0.5899
5.2813	1.2544	1080	0.5889
4.9938	1.5052	1296	0.5891
4.0884	1.7561	1512	0.5893

Framework versions

• Transformers 4.40.2
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1