metadata
base_model: microsoft/Phi-3-mini-4k-instruct
library_name: peft
license: mit
tags:
- generated_from_trainer
model-index:
- name: outputs/phi-sft-out
results: []
See axolotl config
axolotl version: 0.4.1
base_model: microsoft/Phi-3-mini-4k-instruct
trust_remote_code: true
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer
load_in_8bit: false
load_in_4bit: true
strict: false
datasets:
- path: ptoro/honkers-phi
type: alpaca
dataset_prepared_path:
val_set_size: 0.05
output_dir: ./outputs/phi-sft-out
sequence_len: 4096
sample_packing: true
pad_to_sequence_len: true
adapter: qlora
lora_model_dir:
lora_r: 64
lora_alpha: 32
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:
wandb_project: axolotl-june
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:
gradient_accumulation_steps: 1
micro_batch_size: 1
num_epochs: 3
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: true
gradient_checkpointing: true
gradient_checkpointing_kwargs:
use_reentrant: True
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
fsdp:
fsdp_config:
resize_token_embeddings_to_32x: true
special_tokens:
pad_token: "<|endoftext|>"
outputs/phi-sft-out
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: 4.8947
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: 1
- eval_batch_size: 1
- seed: 42
- optimizer: Adam with betas=(0.9,0.95) and epsilon=1e-05
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 100
- num_epochs: 3
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 7.126 | 0.0093 | 1 | 5.2723 |
| 6.503 | 0.25 | 27 | 5.2703 |
| 5.9853 | 0.5 | 54 | 5.2576 |
| 5.7324 | 0.75 | 81 | 5.2320 |
| 6.5292 | 1.0 | 108 | 5.1854 |
| 5.6106 | 1.2222 | 135 | 5.1238 |
| 6.3981 | 1.4722 | 162 | 5.0544 |
| 5.602 | 1.7222 | 189 | 4.9929 |
| 5.3998 | 1.9722 | 216 | 4.9468 |
| 5.1841 | 2.1944 | 243 | 4.9171 |
| 6.0764 | 2.4444 | 270 | 4.9009 |
| 5.2345 | 2.6944 | 297 | 4.8961 |
| 5.4896 | 2.9444 | 324 | 4.8947 |
Framework versions
- PEFT 0.11.2.dev0
- Transformers 4.41.1
- Pytorch 2.1.2+cu118
- Datasets 2.19.1
- Tokenizers 0.19.1