Commit
•
f1cd101
1
Parent(s):
cb20077
Update README.md
Browse files
README.md
CHANGED
|
@@ -8,99 +8,8 @@ model-index:
|
|
| 8 |
results: []
|
| 9 |
---
|
| 10 |
|
| 11 |
-
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 12 |
-
should probably proofread and complete it, then remove this comment. -->
|
| 13 |
|
| 14 |
-
|
| 15 |
-
<details><summary>See axolotl config</summary>
|
| 16 |
-
|
| 17 |
-
axolotl version: `0.3.0`
|
| 18 |
-
```yaml
|
| 19 |
-
base_model: mhenrichsen/danskgpt-tiny
|
| 20 |
-
|
| 21 |
-
model_type: LlamaForCausalLM
|
| 22 |
-
tokenizer_type: LlamaTokenizer
|
| 23 |
-
is_llama_derived_model: true
|
| 24 |
-
|
| 25 |
-
load_in_8bit: false
|
| 26 |
-
load_in_4bit: false
|
| 27 |
-
strict: false
|
| 28 |
-
|
| 29 |
-
#pretraining_dataset: mhenrichsen/terra
|
| 30 |
-
|
| 31 |
-
datasets:
|
| 32 |
-
- path: mhenrichsen/rag-qa-sharegpt
|
| 33 |
-
type: sharegpt
|
| 34 |
-
conversation: chatml
|
| 35 |
-
- path: mhenrichsen/creator
|
| 36 |
-
type: sharegpt
|
| 37 |
-
conversation: chatml
|
| 38 |
-
- path: mhenrichsen/puffin-sharegpt-fix
|
| 39 |
-
type: sharegpt
|
| 40 |
-
conversation: chatml
|
| 41 |
-
- path: mhenrichsen/orcaslim-sharegpt-fix
|
| 42 |
-
type: sharegpt
|
| 43 |
-
conversation: chatml
|
| 44 |
-
- path: mhenrichsen/dansk-tekst-sharegpt
|
| 45 |
-
type: sharegpt
|
| 46 |
-
conversation: chatml
|
| 47 |
-
|
| 48 |
-
chat_template: chatml
|
| 49 |
-
|
| 50 |
-
dataset_prepared_path:
|
| 51 |
-
val_set_size: 0.001
|
| 52 |
-
output_dir: ./tiny-chat
|
| 53 |
-
|
| 54 |
-
sequence_len: 2048
|
| 55 |
-
sample_packing: true
|
| 56 |
-
pad_to_sequence_len: true
|
| 57 |
-
|
| 58 |
-
wandb_project: tiny-danskgpt-chat
|
| 59 |
-
wandb_entity:
|
| 60 |
-
wandb_watch:
|
| 61 |
-
wandb_name:
|
| 62 |
-
wandb_log_model:
|
| 63 |
-
|
| 64 |
-
gradient_accumulation_steps: 4
|
| 65 |
-
micro_batch_size: 16
|
| 66 |
-
num_epochs: 3
|
| 67 |
-
optimizer: adamw_bnb_8bit
|
| 68 |
-
lr_scheduler: cosine
|
| 69 |
-
learning_rate: 0.00005
|
| 70 |
-
|
| 71 |
-
train_on_inputs: false
|
| 72 |
-
group_by_length: false
|
| 73 |
-
bf16: true
|
| 74 |
-
fp16: false
|
| 75 |
-
tf32: false
|
| 76 |
-
|
| 77 |
-
gradient_checkpointing: true
|
| 78 |
-
early_stopping_patience:
|
| 79 |
-
resume_from_checkpoint:
|
| 80 |
-
local_rank:
|
| 81 |
-
logging_steps: 1
|
| 82 |
-
xformers_attention:
|
| 83 |
-
flash_attention: true
|
| 84 |
-
|
| 85 |
-
warmup_steps: 10
|
| 86 |
-
evals_per_epoch: 4
|
| 87 |
-
eval_table_size:
|
| 88 |
-
saves_per_epoch: 2
|
| 89 |
-
debug:
|
| 90 |
-
deepspeed: deepspeed/zero2.json
|
| 91 |
-
weight_decay: 0.1
|
| 92 |
-
fsdp:
|
| 93 |
-
fsdp_config:
|
| 94 |
-
special_tokens:
|
| 95 |
-
eos_token: "<|im_end|>"
|
| 96 |
-
tokens:
|
| 97 |
-
- "<|im_start|>"
|
| 98 |
-
|
| 99 |
-
```
|
| 100 |
-
|
| 101 |
-
</details><br>
|
| 102 |
-
|
| 103 |
-
# tiny-chat
|
| 104 |
|
| 105 |
This model is a fine-tuned version of [mhenrichsen/danskgpt-tiny](https://huggingface.co/mhenrichsen/danskgpt-tiny) on the None dataset.
|
| 106 |
It achieves the following results on the evaluation set:
|
|
|
|
| 8 |
results: []
|
| 9 |
---
|
| 10 |
|
|
|
|
|
|
|
| 11 |
|
| 12 |
+
# DanskGPT-chat-tiny
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 13 |
|
| 14 |
This model is a fine-tuned version of [mhenrichsen/danskgpt-tiny](https://huggingface.co/mhenrichsen/danskgpt-tiny) on the None dataset.
|
| 15 |
It achieves the following results on the evaluation set:
|