Hugging Face's logo

newmindai
/
Llama-3.1-8B-Instruct-w16a8-rw

Text Generation
Transformers
Safetensors
PyTorch
Turkish
llama
legal
turkish
llama-3.1
fp8
bfloat16
mixed-precision
question-answering
fsdp-v2
distributed-training
conversational
text-generation-inference
Model card Files
xet
Community
nmmursit commited on
Commit
7d032f6
·
verified ·
1 Parent(s): 61e051c

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +1 -1
README.md CHANGED
@@ -99,7 +99,7 @@ model = convert_to_float8_training(model, config=config)
99
 
100
  ## **Performance Evaluation**
101
  ### *2-models trained on 1Node with fp8 recipes*
102
- | Loss metric results for w16a16 & rowwise recipe| Memory allocation for w16a16 & rowwise recipe | Utilization for w16a16 & rowwise recipe |
103
  |---------|---------|---------|
104
  | ![lossRW](https://cdn-uploads.huggingface.co/production/uploads/683d4880e639f8d647355997/vGNOYTIGlHPpyBEf4ZBrY.png) | ![MemAlRW](https://cdn-uploads.huggingface.co/production/uploads/683d4880e639f8d647355997/diC7BXnBaL4U9AjtSz8OG.png)|![gpuUtilsRW](https://cdn-uploads.huggingface.co/production/uploads/683d4880e639f8d647355997/Bxiu9g7YbxDI9X1A-8kfA.png) |
105
 
 
99
 
100
  ## **Performance Evaluation**
101
  ### *2-models trained on 1Node with fp8 recipes*
102
+ | Loss metric results for w16a16 tensorwise & w16a8 rowwise recipe| Memory allocation for w16a16 tensorwise & w16a8 rowwise recipe | Utilization for w16a16 tensorwise & w16a8 rowwise recipe |
103
  |---------|---------|---------|
104
  | ![lossRW](https://cdn-uploads.huggingface.co/production/uploads/683d4880e639f8d647355997/vGNOYTIGlHPpyBEf4ZBrY.png) | ![MemAlRW](https://cdn-uploads.huggingface.co/production/uploads/683d4880e639f8d647355997/diC7BXnBaL4U9AjtSz8OG.png)|![gpuUtilsRW](https://cdn-uploads.huggingface.co/production/uploads/683d4880e639f8d647355997/Bxiu9g7YbxDI9X1A-8kfA.png) |
105