text-generation-webui/modules/llamacpp_hf.py

134 lines
4.7 KiB
Python
Raw Normal View History

2023-07-16 07:21:13 +02:00
import os
from pathlib import Path
from typing import Any, Dict, Optional, Union
import torch
from torch.nn import CrossEntropyLoss
from transformers import GenerationConfig, PretrainedConfig, PreTrainedModel
from transformers.modeling_outputs import CausalLMOutputWithPast
from modules import shared
from modules.logging_colors import logger
import llama_cpp
if torch.cuda.is_available() and not torch.version.hip:
try:
import llama_cpp_cuda
except:
llama_cpp_cuda = None
else:
llama_cpp_cuda = None
def llama_cpp_lib():
if shared.args.cpu or llama_cpp_cuda is None:
return llama_cpp
else:
return llama_cpp_cuda
2023-07-16 07:21:13 +02:00
2023-07-16 07:21:13 +02:00
class LlamacppHF(PreTrainedModel):
def __init__(self, model):
super().__init__(PretrainedConfig())
self.model = model
self.generation_config = GenerationConfig()
self.cache = None
def _validate_model_class(self):
pass
def _validate_model_kwargs(self, model_kwargs: Dict[str, Any]):
pass
def prepare_inputs_for_generation(self, input_ids, **kwargs):
return {'input_ids': input_ids, **kwargs}
@property
def device(self) -> torch.device:
return torch.device(0)
def __call__(self, *args, **kwargs):
input_ids = args[0] if len(args) > 0 else kwargs['input_ids']
2023-07-16 07:21:13 +02:00
use_cache = kwargs.get('use_cache', True)
labels = kwargs.get('labels', None)
cache = kwargs.get('past_key_values', None)
seq = input_ids[0].tolist()
2023-07-16 07:21:13 +02:00
# Make the forward call
seq_tensor = torch.tensor(seq)
if labels is None:
if self.cache is None or not torch.equal(self.cache, seq_tensor[:-1]):
self.model.reset()
self.model.eval(seq)
else:
self.model.eval([seq[-1]])
2023-08-07 02:49:27 +02:00
logits = torch.tensor(self.model.scores[self.model.n_tokens - 1, :]).view(1, 1, -1).to(kwargs['input_ids'].device)
2023-07-16 07:21:13 +02:00
else:
self.model.reset()
self.model.eval(seq)
logits = torch.tensor(self.model.eval_logits)
logits = logits.view(1, logits.shape[0], logits.shape[1]).to(input_ids.device)
2023-07-16 07:21:13 +02:00
2023-07-17 05:49:48 +02:00
self.cache = seq_tensor
2023-07-16 07:21:13 +02:00
# Based on transformers/models/llama/modeling_llama.py
loss = None
if labels is not None:
# Shift so that tokens < n predict n
shift_logits = logits[..., :-1, :].contiguous()
shift_labels = labels[..., 1:].contiguous()
# Flatten the tokens
loss_fct = CrossEntropyLoss()
shift_logits = shift_logits.view(-1, logits.shape[-1])
shift_labels = shift_labels.view(-1)
# Enable model parallelism
shift_labels = shift_labels.to(shift_logits.device)
loss = loss_fct(shift_logits, shift_labels)
return CausalLMOutputWithPast(logits=logits, past_key_values=cache if use_cache else None, loss=loss)
@classmethod
def from_pretrained(cls, pretrained_model_name_or_path: Optional[Union[str, os.PathLike]], *model_args, **kwargs):
assert len(model_args) == 0 and len(kwargs) == 0, "extra args is currently not supported"
if isinstance(pretrained_model_name_or_path, str):
pretrained_model_name_or_path = Path(pretrained_model_name_or_path)
path = Path(f'{shared.args.model_dir}') / Path(pretrained_model_name_or_path)
if path.is_file():
model_file = path
else:
model_file = list(path.glob('*ggml*.bin'))[0]
logger.info(f"llama.cpp weights detected: {model_file}\n")
2023-08-18 06:39:26 +02:00
if shared.args.tensor_split is None or shared.args.tensor_split.strip() == '':
tensor_split_list = None
else:
tensor_split_list = [float(x) for x in shared.args.tensor_split.strip().split(",")]
2023-07-16 07:21:13 +02:00
params = {
'model_path': str(model_file),
'n_ctx': shared.args.n_ctx,
'seed': int(shared.args.llama_cpp_seed),
'n_threads': shared.args.threads or None,
'n_batch': shared.args.n_batch,
'use_mmap': not shared.args.no_mmap,
'use_mlock': shared.args.mlock,
2023-08-18 16:58:20 +02:00
'mul_mat_q': shared.args.mul_mat_q,
2023-07-16 07:21:13 +02:00
'low_vram': shared.args.low_vram,
'n_gpu_layers': shared.args.n_gpu_layers,
2023-08-07 02:49:27 +02:00
'rope_freq_base': 10000 * shared.args.alpha_value ** (64 / 63.),
2023-08-18 06:39:26 +02:00
'tensor_split': tensor_split_list,
'rope_freq_scale': 1.0 / shared.args.compress_pos_emb,
2023-07-24 21:37:03 +02:00
'n_gqa': shared.args.n_gqa or None,
'rms_norm_eps': shared.args.rms_norm_eps or None,
2023-07-16 07:21:13 +02:00
'logits_all': True,
}
Llama = llama_cpp_lib().Llama
2023-07-16 07:21:13 +02:00
model = Llama(**params)
2023-07-16 07:21:13 +02:00
return LlamacppHF(model)