Searched refs:sampling_kwargs (Results 1 – 2 of 2) sorted by relevance
| /external/executorch/examples/models/llama/experimental/ |
| D | generate.py | 50 model: torch.nn.Module, x: torch.Tensor, **sampling_kwargs argument
53 return sample(logits, **sampling_kwargs)
56 def prefill(model: torch.nn.Module, x: torch.Tensor, **sampling_kwargs) -> torch.Tensor: argument
57 return decode_one_token(model, x, **sampling_kwargs)[0]
65 **sampling_kwargs, argument
74 model, cur_token.view(1, -1), **sampling_kwargs
94 **sampling_kwargs, argument
119 next_token = prefill(model, prompt.view(1, -1), **sampling_kwargs)
132 **sampling_kwargs,
|
| /external/pytorch/benchmarks/gpt_fast/ |
| D | generate.py | 79 model: torch.nn.Module, x: torch.Tensor, input_pos: torch.Tensor, **sampling_kwargs argument
83 return sample(logits, **sampling_kwargs)[0]
88 model: torch.nn.Module, x: torch.Tensor, input_pos: torch.Tensor, **sampling_kwargs argument
93 return sample(logits, **sampling_kwargs)
101 **sampling_kwargs, argument
109 model, cur_token, input_pos, **sampling_kwargs
121 model: torch.nn.Module, prompt: torch.Tensor, max_new_tokens: int, **sampling_kwargs argument
137 next_token = prefill(model, prompt.view(1, -1), input_pos, **sampling_kwargs)
143 model, next_token.view(1, -1), input_pos, max_new_tokens - 1, **sampling_kwargs
|