Update README.md
Browse files
README.md
CHANGED
@@ -15,8 +15,8 @@ tags:
|
|
15 |
|
16 |
# Info
|
17 |
|
18 |
-
|
19 |
-
|
20 |
* 📜 Model license: [Llama 2 Community License Agreement](https://ai.meta.com/resources/models-and-libraries/llama-downloads/)
|
21 |
* 🏛️ Base Model: [Llama-2-70b-hf](https://huggingface.co/meta-llama/Llama-2-70b-hf)
|
22 |
* 🖥️ Machine: Nvidia A100 (40 GB vRAM)
|
@@ -24,6 +24,8 @@ Adapter model trained with the [**QloRA**](https://arxiv.org/abs/2305.14314) te
|
|
24 |
* ⌛ Training Time: 3 hour 22 minutes
|
25 |
* 📊 Dataset Used: [vicgalle/alpaca-gpt4](https://huggingface.co/datasets/vicgalle/alpaca-gpt4)
|
26 |
|
|
|
|
|
27 |
# Evulation Results ([Open LLM Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard))
|
28 |
|
29 |
| | Average | ARC (25-shot) | HellaSwag (10-shot) | MMLU (5-shot) | TruthfulQA (0-shot) |
|
@@ -36,17 +38,3 @@ Adapter model trained with the [**QloRA**](https://arxiv.org/abs/2305.14314) te
|
|
36 |
|
37 |
![](https://i.imgur.com/xPRcRyM.png)
|
38 |
|
39 |
-
|
40 |
-
## Training procedure
|
41 |
-
|
42 |
-
The following `bitsandbytes` quantization config was used during training:
|
43 |
-
|
44 |
-
- load_in_8bit: False
|
45 |
-
- load_in_4bit: True
|
46 |
-
- llm_int8_threshold: 6.0
|
47 |
-
- llm_int8_skip_modules: None
|
48 |
-
- llm_int8_enable_fp32_cpu_offload: False
|
49 |
-
- llm_int8_has_fp16_weight: False
|
50 |
-
- bnb_4bit_quant_type: nf4
|
51 |
-
- bnb_4bit_use_double_quant: True
|
52 |
-
- bnb_4bit_compute_dtype: bfloat16
|
|
|
15 |
|
16 |
# Info
|
17 |
|
18 |
+
This model is an adapter model trained with [**QloRA**](https://arxiv.org/abs/2305.14314) technique.
|
19 |
+
|
20 |
* 📜 Model license: [Llama 2 Community License Agreement](https://ai.meta.com/resources/models-and-libraries/llama-downloads/)
|
21 |
* 🏛️ Base Model: [Llama-2-70b-hf](https://huggingface.co/meta-llama/Llama-2-70b-hf)
|
22 |
* 🖥️ Machine: Nvidia A100 (40 GB vRAM)
|
|
|
24 |
* ⌛ Training Time: 3 hour 22 minutes
|
25 |
* 📊 Dataset Used: [vicgalle/alpaca-gpt4](https://huggingface.co/datasets/vicgalle/alpaca-gpt4)
|
26 |
|
27 |
+
You can acces Llama-2 paper by clicking [here](https://arxiv.org/abs/2307.09288)
|
28 |
+
|
29 |
# Evulation Results ([Open LLM Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard))
|
30 |
|
31 |
| | Average | ARC (25-shot) | HellaSwag (10-shot) | MMLU (5-shot) | TruthfulQA (0-shot) |
|
|
|
38 |
|
39 |
![](https://i.imgur.com/xPRcRyM.png)
|
40 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|