Update README.md

README.md

@@ -27,7 +27,73 @@ language:
 library_name: transformers
 ---
 
-```Python
+```
+reference-data-model:
+
+  datasets:
+    - OpenAssistant/oasst_top1_2023-08-25:
+      lang: "bg,ca,cs,da,de,en,es,fr,hr,hu,it,nl,pl,pt,ro,ru,sl,sr,sv,uk"
+      link: https://huggingface.co/datasets/OpenAssistant/oasst_top1_2023-08-25
+
+  model:
+    - Open-Orca/Mistral-7B-OpenOrca
+      Link:
+        https://huggingface.co/Open-Orca/Mistral-7B-OpenOrca
+
+  100 examples of generating:
+    - Link:
+      https://huggingface.co/NickyNicky/Mistral-7B-OpenOrca-oasst_top1_2023-08-25-v3/blob/main/output.xlsx
+
+  Activated training with:
+    - Link:
+        https://huggingface.co/blog/tomaarsen/attention-sinks
+        https://github.com/tomaarsen/attention_sinks
+        https://arxiv.org/abs/2309.17453
+
+  TRL:
+    - Link:
+        https://huggingface.co/docs/trl/index
+        https://huggingface.co/docs/trl/sft_trainer
+
+  flash-attention:
+    - Link:
+        https://github.com/Dao-AILab/flash-attention
+        https://arxiv.org/abs/2205.14135
+
+  Version:
+    - Link:
+        https://huggingface.co/NickyNicky/Mistral-7B-OpenOrca-oasst_top1_2023-08-25-v1
+        https://huggingface.co/NickyNicky/Mistral-7B-OpenOrca-oasst_top1_2023-08-25-v2
+        https://huggingface.co/NickyNicky/Mistral-7B-OpenOrca-oasst_top1_2023-08-25-v3
+
+
+```
+
+
+## 
+
+
+```py
+# attention-sinks
+pip install attention_sinks
+
+# flash-attn
+!export CUDA_HOME=/usr/local/cuda-11.8
+!MAX_JOBS=4 pip install flash-attn --no-build-isolation -qqq
+!pip install git+"https://github.com/HazyResearch/flash-attention.git#subdirectory=csrc/rotary" -qqq
+```
+
+
+## Version
+```py
+import torch, transformers,torchvision
+torch.__version__,transformers.__version__, torchvision.__version__
+#OUTPUTS: ('2.0.1+cu118', '4.34.0', '0.15.2+cu118')
+```
+
+## How to use
+```py
+
 from transformers import (
     AutoModelForCausalLM,
     AutoTokenizer,
@@ -99,4 +165,4 @@ generation_config = GenerationConfig(
 outputs = model.generate(generation_config=generation_config,
                                 input_ids=inputs,)
 tokenizer.decode(outputs[0], skip_special_tokens=False) #True
-```
+```