Update README.md
Browse files
README.md
CHANGED
|
@@ -14,120 +14,7 @@ model-index:
|
|
| 14 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 15 |
should probably proofread and complete it, then remove this comment. -->
|
| 16 |
|
| 17 |
-
[
|
| 18 |
-
<details><summary>See axolotl config</summary>
|
| 19 |
|
| 20 |
-
|
| 21 |
-
```yaml
|
| 22 |
-
base_model: Qwen/Qwen2.5-3B
|
| 23 |
-
load_in_8bit: false
|
| 24 |
-
load_in_4bit: false
|
| 25 |
-
strict: false
|
| 26 |
-
|
| 27 |
-
datasets:
|
| 28 |
-
- path: arcee-ai/eval_tome
|
| 29 |
-
type: sharegpt
|
| 30 |
-
conversation: chatml
|
| 31 |
-
- path: arcee-ai/math_code_5k_claude
|
| 32 |
-
type: sharegpt
|
| 33 |
-
conversation: chatml
|
| 34 |
-
split: validation
|
| 35 |
-
- path: Undi95/Capybara-ShareGPT
|
| 36 |
-
type: sharegpt
|
| 37 |
-
conversation: chatml
|
| 38 |
-
dataset_prepared_path:
|
| 39 |
-
val_set_size: 0.0
|
| 40 |
-
|
| 41 |
-
sequence_len: 8192
|
| 42 |
-
sample_packing: true
|
| 43 |
-
|
| 44 |
-
lora_fan_in_fan_out:
|
| 45 |
-
wandb_project: qwen2.5-3b-gelato
|
| 46 |
-
wandb_entity:
|
| 47 |
-
wandb_watch:
|
| 48 |
-
wandb_name:
|
| 49 |
-
wandb_log_model:
|
| 50 |
-
output_dir: ./outputs/gelato-3b
|
| 51 |
-
gradient_accumulation_steps: 8
|
| 52 |
-
micro_batch_size: 2
|
| 53 |
-
num_epochs: 4
|
| 54 |
-
optimizer: adamw_bnb_8bit
|
| 55 |
-
torchdistx_path:
|
| 56 |
-
lr_scheduler: cosine
|
| 57 |
-
learning_rate: 0.0002
|
| 58 |
-
train_on_inputs: false
|
| 59 |
-
group_by_length: false
|
| 60 |
-
bf16: auto
|
| 61 |
-
fp16:
|
| 62 |
-
tf32: true
|
| 63 |
-
gradient_checkpointing: true
|
| 64 |
-
early_stopping_patience:
|
| 65 |
-
resume_from_checkpoint:
|
| 66 |
-
local_rank:
|
| 67 |
-
logging_steps: 1
|
| 68 |
-
xformers_attention:
|
| 69 |
-
flash_attention: true
|
| 70 |
-
gptq_groupsize:
|
| 71 |
-
s2_attention:
|
| 72 |
-
gptq_model_v1:
|
| 73 |
-
warmup_steps: 50
|
| 74 |
-
evals_per_epoch:
|
| 75 |
-
saves_per_epoch: 1
|
| 76 |
-
debug:
|
| 77 |
-
deepspeed: deepspeed_configs/zero3_bf16_cpuoffload_params.json
|
| 78 |
-
weight_decay: 0.1
|
| 79 |
-
fsdp:
|
| 80 |
-
fsdp_config:
|
| 81 |
-
special_tokens:
|
| 82 |
-
eos_token: "<|im_end|>"
|
| 83 |
-
bos_token: "<|im_start|>"
|
| 84 |
-
|
| 85 |
-
```
|
| 86 |
-
|
| 87 |
-
</details><br>
|
| 88 |
-
|
| 89 |
-
# outputs/gelato-3b
|
| 90 |
-
|
| 91 |
-
This model is a fine-tuned version of [Qwen/Qwen2.5-3B](https://huggingface.co/Qwen/Qwen2.5-3B) on the None dataset.
|
| 92 |
-
|
| 93 |
-
## Model description
|
| 94 |
-
|
| 95 |
-
More information needed
|
| 96 |
-
|
| 97 |
-
## Intended uses & limitations
|
| 98 |
-
|
| 99 |
-
More information needed
|
| 100 |
-
|
| 101 |
-
## Training and evaluation data
|
| 102 |
-
|
| 103 |
-
More information needed
|
| 104 |
-
|
| 105 |
-
## Training procedure
|
| 106 |
-
|
| 107 |
-
### Training hyperparameters
|
| 108 |
-
|
| 109 |
-
The following hyperparameters were used during training:
|
| 110 |
-
- learning_rate: 0.0002
|
| 111 |
-
- train_batch_size: 2
|
| 112 |
-
- eval_batch_size: 2
|
| 113 |
-
- seed: 42
|
| 114 |
-
- distributed_type: multi-GPU
|
| 115 |
-
- num_devices: 4
|
| 116 |
-
- gradient_accumulation_steps: 8
|
| 117 |
-
- total_train_batch_size: 64
|
| 118 |
-
- total_eval_batch_size: 8
|
| 119 |
-
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 120 |
-
- lr_scheduler_type: cosine
|
| 121 |
-
- lr_scheduler_warmup_steps: 50
|
| 122 |
-
- num_epochs: 4
|
| 123 |
-
|
| 124 |
-
### Training results
|
| 125 |
-
|
| 126 |
-
|
| 127 |
-
|
| 128 |
-
### Framework versions
|
| 129 |
|
| 130 |
-
- Transformers 4.45.1
|
| 131 |
-
- Pytorch 2.3.1+cu121
|
| 132 |
-
- Datasets 2.21.0
|
| 133 |
-
- Tokenizers 0.20.0
|
|
|
|
| 14 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 15 |
should probably proofread and complete it, then remove this comment. -->
|
| 16 |
|
| 17 |
+
|
|
|
|
| 18 |
|
| 19 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 20 |
|
|
|
|
|
|
|
|
|
|
|
|