metadata
library_name: transformers
tags:
- generated_from_trainer
datasets:
- /content/processed_output/clartts_data.jsonl
model-index:
- name: content/finetuned_model
results: []
See axolotl config
axolotl version: 0.13.0.dev0
base_model: /content/SparkTTS-Finetune/pretrained_models/Spark-TTS-0.5B/LLM
load_in_4bit: false
load_in_8bit: false
trust_remote_code: true
strict: false
datasets:
- path: /content/processed_output/clartts_data.jsonl
type: completion
dataset_prepared_path:
val_set_size: 0.05
output_dir: /content/finetuned_model
sequence_len: 1024
sample_packing: false
eval_sample_packing: false
pad_to_sequence_len: true
wandb_project:
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:
gradient_accumulation_steps: 8
micro_batch_size: 1
num_epochs: 3
optimizer: adamw_torch_fused
lr_scheduler: cosine
learning_rate: 0.0002
train_on_inputs: false
group_by_length: false
bf16: true
fp16: false
tf32: false
gradient_checkpointing: true
gradient_checkpointing_kwargs:
use_reentrant: false
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 50
xformers_attention:
flash_attention: false
warmup_steps: 10
evals_per_epoch: 1
save_steps: 200
debug:
deepspeed:
weight_decay: 0.0
content/finetuned_model
This model was trained from scratch on the /content/processed_output/clartts_data.jsonl dataset. It achieves the following results on the evaluation set:
- Loss: 4.4637
- Memory/max Active (gib): 7.2
- Memory/max Allocated (gib): 7.2
- Memory/device Reserved (gib): 7.62
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: 0.0002
- train_batch_size: 1
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 8
- total_train_batch_size: 8
- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 10
- training_steps: 1016
Training results
| Training Loss | Epoch | Step | Validation Loss | Active (gib) | Allocated (gib) | Reserved (gib) |
|---|---|---|---|---|---|---|
| No log | 0 | 0 | 11.8503 | 3.1 | 3.1 | 3.2 |
| 4.7248 | 1.0 | 339 | 4.6423 | 7.2 | 7.2 | 7.67 |
| 4.3688 | 2.0 | 678 | 4.4637 | 7.2 | 7.2 | 7.62 |
Framework versions
- Transformers 4.57.1
- Pytorch 2.7.1+cu118
- Datasets 4.4.1
- Tokenizers 0.22.1