@@ -161,10 +161,10 @@ You can download the model weights from the [Hugging Face Hub](https://huggingfa
161
161
162
162
``` shell
163
163
# gpt-oss-120b
164
- huggingface-cli download openai/gpt-oss-120b --include " original/*" --local-dir gpt-oss-120b/
164
+ hf download openai/gpt-oss-120b --include " original/*" --local-dir gpt-oss-120b/
165
165
166
166
# gpt-oss-20b
167
- huggingface-cli download openai/gpt-oss-20b --include " original/*" --local-dir gpt-oss-20b/
167
+ hf download openai/gpt-oss-20b --include " original/*" --local-dir gpt-oss-20b/
168
168
```
169
169
170
170
## Reference PyTorch implementation
@@ -174,7 +174,7 @@ We include an inefficient reference PyTorch implementation in [gpt_oss/torch/mod
174
174
To run the reference implementation. Install dependencies:
175
175
176
176
``` shell
177
- pip install -e .[torch]
177
+ pip install -e " .[torch]"
178
178
```
179
179
180
180
And then run:
@@ -198,7 +198,7 @@ pip install -r python/requirements.txt
198
198
pip install -e . --verbose --no-build-isolation
199
199
200
200
# Install the gpt-oss triton implementation
201
- pip install -e .[triton]
201
+ pip install -e " .[triton]"
202
202
```
203
203
204
204
And then run:
@@ -218,7 +218,7 @@ Additionally we are providing a reference implementation for Metal to run on App
218
218
The implementation will get automatically compiled when running the ` .[metal] ` installation on an Apple Silicon device:
219
219
220
220
``` shell
221
- pip install -e .[metal]
221
+ pip install -e " .[metal]"
222
222
```
223
223
224
224
To perform inference you'll need to first convert the SafeTensor weights from Hugging Face into the right format using:
@@ -230,8 +230,8 @@ python gpt_oss/metal/scripts/create-local-model.py -s <model_dir> -d <output_fil
230
230
Or downloaded the pre-converted weight:
231
231
232
232
``` shell
233
- huggingface-cli download openai/gpt-oss-120b --include " metal/*" --local-dir gpt-oss-120b/metal/
234
- huggingface-cli download openai/gpt-oss-20b --include " metal/*" --local-dir gpt-oss-20b/metal/
233
+ hf download openai/gpt-oss-120b --include " metal/*" --local-dir gpt-oss-120b/metal/
234
+ hf download openai/gpt-oss-20b --include " metal/*" --local-dir gpt-oss-20b/metal/
235
235
```
236
236
237
237
To test it you can run:
0 commit comments