Update README.md
Browse files
README.md
CHANGED
|
@@ -8,86 +8,6 @@ model-index:
|
|
| 8 |
results: []
|
| 9 |
---
|
| 10 |
|
| 11 |
-
[<img src="https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/OpenAccess-AI-Collective/axolotl)
|
| 12 |
-
<details>
|
| 13 |
-
<summary>See axolotl config</summary>
|
| 14 |
-
|
| 15 |
-
axolotl version: `0.4.0`
|
| 16 |
-
```yaml
|
| 17 |
-
base_model: alpindale/Mistral-7B-v0.2-hf
|
| 18 |
-
model_type: MistralForCausalLM
|
| 19 |
-
tokenizer_type: LlamaTokenizer
|
| 20 |
-
|
| 21 |
-
load_in_8bit: false
|
| 22 |
-
load_in_4bit: false
|
| 23 |
-
strict: false
|
| 24 |
-
|
| 25 |
-
datasets:
|
| 26 |
-
- path: /data/seongyun/open-instruct-2/augmented_diverse_response/preferences_v1_responses_for_training.jsonl
|
| 27 |
-
type:
|
| 28 |
-
system_prompt: ""
|
| 29 |
-
system_format: "[INST] {system}\n"
|
| 30 |
-
field_system: system
|
| 31 |
-
field_instruction: instruction
|
| 32 |
-
field_output: output
|
| 33 |
-
format: "{instruction} [/INST]"
|
| 34 |
-
no_input_format: "{instruction} [/INST]"
|
| 35 |
-
# conversation: mistral
|
| 36 |
-
|
| 37 |
-
dataset_prepared_path:
|
| 38 |
-
hub_model_id: kaist-ai/mpa-Mistral-7b-v0.2-hf-sft
|
| 39 |
-
hub_strategy: checkpoint
|
| 40 |
-
# val_set_size: 0
|
| 41 |
-
output_dir: /data/suehyun/axolotl/outputs/mpa/mistral-7b-v0.2-hf
|
| 42 |
-
|
| 43 |
-
sequence_len: 8192
|
| 44 |
-
sample_packing: true
|
| 45 |
-
pad_to_sequence_len: true
|
| 46 |
-
eval_sample_packing: false
|
| 47 |
-
|
| 48 |
-
wandb_project: mpa
|
| 49 |
-
wandb_entity: suehyun
|
| 50 |
-
wandb_watch:
|
| 51 |
-
wandb_name: mpa_mistral-7b-v0.2-hf
|
| 52 |
-
wandb_log_model:
|
| 53 |
-
|
| 54 |
-
gradient_accumulation_steps: 4
|
| 55 |
-
micro_batch_size: 2
|
| 56 |
-
num_epochs: 4
|
| 57 |
-
optimizer: adamw_bnb_8bit
|
| 58 |
-
lr_scheduler: cosine
|
| 59 |
-
learning_rate: 0.000005
|
| 60 |
-
|
| 61 |
-
train_on_inputs: false
|
| 62 |
-
group_by_length: false
|
| 63 |
-
bf16: auto
|
| 64 |
-
fp16:
|
| 65 |
-
tf32: false
|
| 66 |
-
|
| 67 |
-
gradient_checkpointing: true
|
| 68 |
-
early_stopping_patience:
|
| 69 |
-
resume_from_checkpoint:
|
| 70 |
-
local_rank:
|
| 71 |
-
logging_steps: 1
|
| 72 |
-
xformers_attention:
|
| 73 |
-
flash_attention: true
|
| 74 |
-
|
| 75 |
-
warmup_steps: 10
|
| 76 |
-
# evals_per_epoch: 4
|
| 77 |
-
eval_table_size:
|
| 78 |
-
# eval_max_new_tokens: 128
|
| 79 |
-
saves_per_epoch: 1
|
| 80 |
-
debug:
|
| 81 |
-
deepspeed:
|
| 82 |
-
weight_decay: 0.0
|
| 83 |
-
fsdp:
|
| 84 |
-
fsdp_config:
|
| 85 |
-
special_tokens:
|
| 86 |
-
|
| 87 |
-
```
|
| 88 |
-
|
| 89 |
-
</details><br>
|
| 90 |
-
|
| 91 |
### Training hyperparameters
|
| 92 |
|
| 93 |
The following hyperparameters were used during training:
|
|
|
|
| 8 |
results: []
|
| 9 |
---
|
| 10 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 11 |
### Training hyperparameters
|
| 12 |
|
| 13 |
The following hyperparameters were used during training:
|