ctrltokyo commited on
Commit
3600b7d
1 Parent(s): 8549658

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +20 -1
README.md CHANGED
@@ -6,8 +6,27 @@ language:
6
  - en
7
  pipeline_tag: text-generation
8
  ---
9
- ## Training procedure
10
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
11
 
12
  The following `bitsandbytes` quantization config was used during training:
13
  - load_in_8bit: False
 
6
  - en
7
  pipeline_tag: text-generation
8
  ---
 
9
 
10
+ # ctrltokyo/llama-2-7b-hf-dolly-flash-attention
11
+
12
+ This model is a fine-tuned version of [NousResearch/Llama-2-7b-hf](https://huggingface.co/NousResearch/Llama-2-7b-hf) on the databricks/databricks-dolly-15k dataset with all training performed using Flash Attention 2.
13
+
14
+ No further testing or optimisation has been performed.
15
+
16
+ ## Model description
17
+
18
+ Just like [ctrltokyo/llm_prompt_mask_fill_model](https://huggingface.co/ctrltokyo/llm_prompt_mask_fill_model), this model could be used for live autocompletion of PROMPTS, but is more designed for a generalized chatbot (hence the usage of the Dolly 15k dataset). Don't try this on code, because it won't work.
19
+ I plan to release a further fine-tuned version using the [code_instructions_120k](https://huggingface.co/datasets/sahil2801/code_instructions_120k) dataset.
20
+
21
+ ## Intended uses & limitations
22
+
23
+ Use as intended.
24
+
25
+ ## Training and evaluation data
26
+
27
+ No evaluation was performed. Trained on NVIDIA A100, but appears to use around 20GB of VRAM when performing inference on the raw model.
28
+
29
+ ## Training procedure
30
 
31
  The following `bitsandbytes` quantization config was used during training:
32
  - load_in_8bit: False