Update README.md
Browse files
README.md
CHANGED
|
@@ -4,6 +4,8 @@ license: apache-2.0
|
|
| 4 |
### huihui-ai/grok-2
|
| 5 |
This Python [script](https://huggingface.co/huihui-ai/grok-2/blob/main/convert_safetensors.py) is designed to process and merge sharded weight files (in safetensors format) for a machine learning model, specifically targeting the `xai-org/grok-2` model. The main functionalities include:
|
| 6 |
|
|
|
|
|
|
|
| 7 |
1. **Collecting safetensors files**: Locates all `pytorch_model-*.safetensors` files in the specified model directory.
|
| 8 |
2. **Loading files into cache**: Loads all safetensors files into memory and builds a key-to-file mapping.
|
| 9 |
3. **Merging Tensor Parallel (TP) shards**: Merges shards for tensor parallelism (TP=8) along specific dimensions and verifies the merged tensor shapes.
|
|
|
|
| 4 |
### huihui-ai/grok-2
|
| 5 |
This Python [script](https://huggingface.co/huihui-ai/grok-2/blob/main/convert_safetensors.py) is designed to process and merge sharded weight files (in safetensors format) for a machine learning model, specifically targeting the `xai-org/grok-2` model. The main functionalities include:
|
| 6 |
|
| 7 |
+
Just a simple merge, without any inference code, and does not indicate whether the final model is reasonable or correct.
|
| 8 |
+
|
| 9 |
1. **Collecting safetensors files**: Locates all `pytorch_model-*.safetensors` files in the specified model directory.
|
| 10 |
2. **Loading files into cache**: Loads all safetensors files into memory and builds a key-to-file mapping.
|
| 11 |
3. **Merging Tensor Parallel (TP) shards**: Merges shards for tensor parallelism (TP=8) along specific dimensions and verifies the merged tensor shapes.
|