Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,67 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language:
|
3 |
+
- en
|
4 |
+
license: apache-2.0
|
5 |
+
tags:
|
6 |
+
- text-generation-inference
|
7 |
+
- transformers
|
8 |
+
- unsloth
|
9 |
+
- llama
|
10 |
+
- trl
|
11 |
+
base_model: unsloth/llama-3-70b-bnb-4bit
|
12 |
+
datasets:
|
13 |
+
- lightblue/tagengo-gpt4
|
14 |
+
---
|
15 |
+
|
16 |
+
# Uploaded model
|
17 |
+
|
18 |
+
- **Developed by:** ruslandev
|
19 |
+
- **License:** apache-2.0
|
20 |
+
- **Finetuned from model :** unsloth/llama-3-70b-bnb-4bit
|
21 |
+
|
22 |
+
This model is finetuned on the Tagengo dataset.
|
23 |
+
Please note - this model has been created for educational purposes and it needs further training/fine tuning.
|
24 |
+
|
25 |
+
# How to use
|
26 |
+
|
27 |
+
The easiest way to use this model on your own computer is to use the GGUF version of this model ([ruslandev/llama-3-70b-tagengo-GGUF](https://huggingface.co/ruslandev/llama-3-70b-tagengo-GGUF)) using a program such as [llama.cpp](https://github.com/ggerganov/llama.cpp).
|
28 |
+
If you want to use this model directly with the Huggingface Transformers stack, I recommend using my framework [gptchain](https://github.com/RuslanPeresy/gptchain).
|
29 |
+
|
30 |
+
```
|
31 |
+
git clone https://github.com/RuslanPeresy/gptchain.git
|
32 |
+
cd gptchain
|
33 |
+
pip install -r requirements-train.txt
|
34 |
+
python gptchain.py chat -m ruslandev/llama-3-70b-tagengo \
|
35 |
+
--chatml true \
|
36 |
+
-q '[{"from": "human", "value": "Из чего состоит нейронная сеть?"}]'
|
37 |
+
```
|
38 |
+
|
39 |
+
# Training
|
40 |
+
[gptchain](https://github.com/RuslanPeresy/gptchain) framework has been used for training.
|
41 |
+
|
42 |
+
```
|
43 |
+
python gptchain.py train -m unsloth/llama-3-70b-bnb-4bit \
|
44 |
+
-dn tagengo_gpt4 \
|
45 |
+
-sp checkpoints/llama-3-70b-tagengo \
|
46 |
+
-hf llama-3-70b-tagengo \
|
47 |
+
--max-steps 2400
|
48 |
+
```
|
49 |
+
|
50 |
+
# Training hyperparameters
|
51 |
+
|
52 |
+
- learning_rate: 2e-4
|
53 |
+
- seed: 3407
|
54 |
+
- gradient_accumulation_steps: 4
|
55 |
+
- per_device_train_batch_size: 2
|
56 |
+
- optimizer: adamw_8bit
|
57 |
+
- lr_scheduler_type: linear
|
58 |
+
- warmup_steps: 5
|
59 |
+
- max_steps: 2400
|
60 |
+
- weight_decay: 0.01
|
61 |
+
|
62 |
+
# Training results
|
63 |
+
[wandb report](https://api.wandb.ai/links/ruslandev/rilj60ra)
|
64 |
+
|
65 |
+
2400 steps took 7 hours on a single H100
|
66 |
+
|
67 |
+
[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)
|