TheBloke

/

Llama-2-70B-GPTQ

like 83

could you provide some code?

I'm just using the example code in the readme for the autogptq section

When I have tested inference before, I have had code to change the seed, like so:

    @property
    def seed(self):
        return self._current_seed

    

@seed

	.setter
    def seed(self, seed):
        self._seed = int(seed)

    def update_seed(self):
        self._current_seed = (self._seed == -1 ) and random.randint(1, 2**31) or self._seed
        random.seed(self._current_seed)
        torch.manual_seed(self._current_seed)
        torch.cuda.manual_seed_all(self._current_seed)

    def generate(self, prompt):
        self.update_seed()
        input_ids, len_input_ids = self.encode(prompt)

        with self.do_timing(True) as timing:
            with torch.no_grad():
                tokens = self.model.generate(inputs=input_ids, generation_config=self.generation_config)[0].cuda()
            len_reply = len(tokens) - len_input_ids
            response = self.tokenizer.decode(tokens)
            reply_tokens = tokens[-len_reply:]
            reply = self.tokenizer.decode(reply_tokens)

        result = {
            'response': response,   # The response in full, including prompt
            'reply': reply,         # Just the reply, no prompt
            'len_reply': len_reply, # The length of the reply tokens
            'seed': self.seed,      # The seed used to generate this response
            'time': timing['time']  # The time in seconds to generate the response
        }
        return result

You could try the same to get a different seed for each generation.