Muennighoff
commited on
Commit
•
9f8fbfd
1
Parent(s):
ef09465
Update README.md
Browse files
README.md
CHANGED
@@ -122,13 +122,15 @@ Please see [the BLOOM training README](https://github.com/bigscience-workshop/bi
|
|
122 |
|
123 |
* ALiBI positional encodings (see [paper](https://arxiv.org/pdf/2108.12409.pdf)), with GeLU activation functions
|
124 |
|
125 |
-
*
|
|
|
|
|
126 |
|
127 |
* 24 layers, 16 attention heads
|
128 |
|
129 |
* Hidden layers are 1024-dimensional
|
130 |
|
131 |
-
* Sequence length of 2048 tokens
|
132 |
|
133 |
**Objective Function:** Cross Entropy with mean reduction (see [API documentation](https://pytorch.org/docs/stable/generated/torch.nn.CrossEntropyLoss.html#torch.nn.CrossEntropyLoss)).
|
134 |
|
@@ -165,18 +167,9 @@ Please see [the BLOOM training README](https://github.com/bigscience-workshop/bi
|
|
165 |
|
166 |
#### **Training**
|
167 |
|
168 |
-
|
169 |
-
_In progress._
|
170 |
-
|
171 |
-
Current training logs: [Tensorboard link](https://huggingface.co/tensorboard/bigscience/tr11-176B-ml-logs/)
|
172 |
-
|
173 |
-
- Checkpoint size:
|
174 |
-
|
175 |
-
- Bf16 weights: 329GB
|
176 |
-
|
177 |
-
- Full checkpoint with optimizer states: 2.3TB
|
178 |
|
179 |
-
- Training throughput: About 150
|
180 |
|
181 |
- Number of epochs: 1 (*current target*)
|
182 |
|
@@ -184,9 +177,9 @@ Current training logs: [Tensorboard link](https://huggingface.co/tensorboard/big
|
|
184 |
|
185 |
- Started 11th March, 2022 11:42am PST
|
186 |
|
187 |
-
-
|
188 |
|
189 |
-
- Estimated cost of training: Equivalent of $2-5M in cloud computing (including preliminary experiments)
|
190 |
|
191 |
- Server training location: Île-de-France, France
|
192 |
|
|
|
122 |
|
123 |
* ALiBI positional encodings (see [paper](https://arxiv.org/pdf/2108.12409.pdf)), with GeLU activation functions
|
124 |
|
125 |
+
* 559,214,592 parameters:
|
126 |
+
|
127 |
+
* 256,901,120 embedding parameters
|
128 |
|
129 |
* 24 layers, 16 attention heads
|
130 |
|
131 |
* Hidden layers are 1024-dimensional
|
132 |
|
133 |
+
* Sequence length of 2048 tokens (see [BLOOM tokenizer](https://huggingface.co/bigscience/tokenizer), [tokenizer description](#tokenization))
|
134 |
|
135 |
**Objective Function:** Cross Entropy with mean reduction (see [API documentation](https://pytorch.org/docs/stable/generated/torch.nn.CrossEntropyLoss.html#torch.nn.CrossEntropyLoss)).
|
136 |
|
|
|
167 |
|
168 |
#### **Training**
|
169 |
|
170 |
+
Training logs: [Tensorboard link](https://huggingface.co/bigscience/tr11e-350M-logs)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
171 |
|
172 |
+
- Training throughput: About 150 TFLOPs per GPU per second
|
173 |
|
174 |
- Number of epochs: 1 (*current target*)
|
175 |
|
|
|
177 |
|
178 |
- Started 11th March, 2022 11:42am PST
|
179 |
|
180 |
+
- Ended 5th July, 2022
|
181 |
|
182 |
+
- Estimated cost of training: Equivalent of $2-5M in cloud computing (including preliminary experiments and other model sizes)
|
183 |
|
184 |
- Server training location: Île-de-France, France
|
185 |
|