import os import re import json import torch import torch.nn as nn from collections import defaultdict from transformers import ( AutoConfig, AutoModelForCausalLM, AutoTokenizer, ) from transformers.models.mixtral.modeling_mixtral import MixtralSparseMoeBlock from transformers.models.mixtral.configuration_mixtral import MixtralConfig from transformers.modeling_utils import WEIGHTS_INDEX_NAME, WEIGHTS_NAME, shard_checkpoint, load_sharded_checkpoint from megatron.initialize import initialize_megatron from megatron import get_args from megatron.model import ModelType from megatron.checkpointing import get_checkpoint_names, get_checkpoint_tracker_filename, read_metadata import sys path_dir = os.path.dirname(os.path.dirname(os.path.dirname(os.path.dirname(os.path.realpath(__file__))))) sys.path.append(os.path.join(path_dir, "examples")) from qwen1_5.pretrain_megablocks_qwen import model_provider from megatron_patch.arguments import get_patch_args torch.backends.cudnn.deterministic = True torch.backends.cudnn.benchmark = False torch.backends.cuda.matmul.allow_tf32 = False torch.backends.cudnn.allow_tf32 = False import numpy as np from collections.abc import Mapping, Sequence @torch.inference_mode() def clone_state_dict(elem): """clone all tensors in the elem to cpu device. """ elem_type = type(elem) if isinstance(elem, torch.Tensor): elem = elem.clone() elif isinstance(elem, (np.ndarray, str)): pass elif isinstance(elem, Mapping): elem = dict(elem) for k, v in elem.items(): elem[k] = clone_state_dict(v) elem = elem_type(elem) elif isinstance(elem, Sequence): elem = list(elem) for i in range(len(elem)): elem[i] = clone_state_dict(elem[i]) elem = elem_type(elem) return elem def add_checkpointing_args(parser): parser.add_argument('--megatron-path', type=str, default=None, help='Base directory of Megatron repository') parser.add_argument( '--convert_checkpoint_from_megatron_to_transformers', action='store_true', help= ('If True, convert a Megatron checkpoint to a Transformers checkpoint. ' 'If False, convert a Transformers checkpoint to a Megatron checkpoint.' ), ) parser.add_argument( '--load_path', type=str, required=True, help='Path to the checkpoint to convert.', ) parser.add_argument( '--save_path', type=str, required=True, help='Path to the converted checkpoint.', ) parser.add_argument( '--huggingface_model_path', type=str, required= True, ) return parser def add_megatron_checkpoint_args(parser): parser.add_argument( "--target_tensor_model_parallel_size", type=int, default=1, help=( "The tensor model parallel size of the converted checkpoint. " "Only used when converting a Transformers checkpoint to a Megatron checkpoint." ), ) parser.add_argument( "--target_pipeline_model_parallel_size", type=int, default=1, help=( "The pipeline model parallel size of the converted checkpoint. " "Only used when converting a Transformers checkpoint to a Megatron checkpoint." ), ) parser.add_argument( "--target_expert_model_parallel_size", type=int, default=1, help=( "The data parallel size of the converted checkpoint. " "Only used when converting a Transformers checkpoint to a Megatron checkpoint." ), ) return parser def add_transformers_checkpoint_args(parser): parser.add_argument( "--max_shard_size", type=str, default="10GB", help=( "The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size " "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`). " "Only used when converting a Megatron checkpoint to a Transformers checkpoint." ), ) return parser def build_huggingface_model(model_to_load, compute_dtype, random_init=False): config = AutoConfig.from_pretrained( model_to_load, trust_remote_code=True, ) if random_init: model = AutoModelForCausalLM.from_config( config=config, torch_dtype=compute_dtype, trust_remote_code=True ) else: model = AutoModelForCausalLM.from_pretrained( model_to_load, torch_dtype=compute_dtype, trust_remote_code=True ) return config, model.eval() def replace_mlp_with_moe(args, model): config = MixtralConfig( intermediate_size=args.intermediate_size, hidden_size=args.hidden_size, num_attention_heads=args.num_attention_heads, num_local_experts=args.num_local_experts, num_key_value_heads=args.num_key_value_heads, rope_theta=args.rope_theta, rms_norm_eps=args.rms_norm_eps, num_experts_per_tok=1, ) def get_hidden_output(module, args, output): return output[0] for layer in model.model.layers: mlp = MixtralSparseMoeBlock(config).to(args.torch_dtype) mlp.register_forward_hook(get_hidden_output) layer.mlp = mlp return model def create_huggingface_model(args): if not args.convert_checkpoint_from_megatron_to_transformers or args.num_experts is None: copy_huggingface_tokenizer(args.huggingface_model_path, args.save_path) config, model = build_huggingface_model(args.huggingface_model_path, args.params_dtype) else: copy_huggingface_tokenizer(args.huggingface_model_path, args.save_path, with_code=True) config, model = build_huggingface_model(args.save_path, args.params_dtype, random_init=True) model = replace_mlp_with_moe(config, model) return config, model.eval() def create_megatron_model(args, hf_config): args = get_args() args.model_type = ModelType.encoder_or_decoder_with_lbl args.hidden_size = hf_config.hidden_size args.num_layers = hf_config.num_hidden_layers args.num_attention_heads = hf_config.num_attention_heads args.kv_channels = args.hidden_size // args.num_attention_heads args.ffn_hidden_size = hf_config.intermediate_size args.num_query_groups = hf_config.num_key_value_heads model = model_provider() return model.eval() def copy_huggingface_tokenizer(src_path, dst_path, with_code=False): assert os.path.exists(src_path) os.makedirs(dst_path, exist_ok=True) os.system("cp -rf " + src_path + "/config*.json " + dst_path) os.system("cp -rf " + src_path + "/tokenizer* " + dst_path) os.system("cp -rf " + src_path + "/vocab.json " + dst_path) os.system("cp -rf " + src_path + "/merges.txt " + dst_path) if with_code: cur_dir = os.path.dirname(os.path.abspath(__file__)) code_path = os.path.join(cur_dir, 'hf_qwen1.5_moe') os.system("cp -rf " + code_path + "/*.py " + dst_path) os.system("cp -rf " + code_path + "/*.json " + dst_path) def name_to_expert_rank(key): pattern = r'local_experts\.(\d+)\.' expert_rank = int(re.findall(pattern, key)[0]) return expert_rank def load_megatron_model(args, model): model_path = args.load_path tracker_filename = get_checkpoint_tracker_filename(model_path) iteration, release = read_metadata(tracker_filename) head_dim = args.hidden_size // args.num_attention_heads group_per_split = args.num_query_groups // args.target_tensor_model_parallel_size num_local_experts = args.moe_num_experts // args.target_expert_model_parallel_size if args.moe_num_experts else 0 state_dict = {} mid_state = defaultdict(list) if ( args.target_tensor_model_parallel_size == 1 and args.target_pipeline_model_parallel_size == 1 and args.target_expert_model_parallel_size == 1 ): checkpoint_name = get_checkpoint_names(model_path, iteration, release, None, None, None, None, None) state_dict = torch.load(checkpoint_name)['model'] elif ( args.target_tensor_model_parallel_size == 1 and args.target_pipeline_model_parallel_size == 1 and args.moe_num_experts and args.moe_num_experts % args.target_expert_model_parallel_size == 0 ): for ep_rank in range(args.target_expert_model_parallel_size): checkpoint_name = get_checkpoint_names(model_path, iteration, release, None, None, None, True, ep_rank) print(f'load {checkpoint_name}') split_state = torch.load(checkpoint_name, map_location="cpu")['model'] for k, v in split_state.items(): if 'local_experts' in k: expert_local_rank = name_to_expert_rank(k) expert_rank = expert_local_rank + num_local_experts * ep_rank k = k.replace(f'local_experts.{expert_local_rank}', f'local_experts.{expert_rank}') state_dict[k] = v elif ( args.target_tensor_model_parallel_size > 1 and args.target_pipeline_model_parallel_size == 1 and args.moe_num_experts is None ): for tp_rank in range(args.target_tensor_model_parallel_size): checkpoint_name = get_checkpoint_names(model_path, iteration, release, None, tp_rank, None, None, None) print(f'load {checkpoint_name}') split_state = torch.load(checkpoint_name, map_location="cpu")['model'] for k, v in split_state.items(): mid_state[k].append(v) for k, v in mid_state.items(): if not isinstance(v[0], torch.Tensor) or 'norm' in k: target_v = v[0] elif 'embedding' in k or 'output_layer' in k: target_v = torch.cat(v, dim=0) elif 'linear_proj' in k or 'linear_fc2' in k: target_v = torch.cat(v, dim=1) elif 'linear_qkv.weight' in k: viewed = [x.view(group_per_split, -1, head_dim, args.hidden_size) for x in v] target_v = torch.cat(viewed, dim=0).view(-1, args.hidden_size) elif 'linear_qkv.bias' in k: viewed = [x.view(group_per_split, -1) for x in v] target_v = torch.cat(viewed, dim=0).view(-1) elif 'linear_fc1' in k: viewed = [x.view(2, -1, args.hidden_size) for x in v] target_v = torch.cat(viewed, dim=1).view(-1, args.hidden_size) else: raise ValueError state_dict[k] = target_v elif ( args.target_tensor_model_parallel_size > 1 and args.target_pipeline_model_parallel_size == 1 and args.moe_num_experts and args.moe_num_experts % args.target_expert_model_parallel_size == 0 ): for tp_rank in range(args.target_tensor_model_parallel_size): for ep_rank in range(args.target_expert_model_parallel_size): checkpoint_name = get_checkpoint_names(model_path, iteration, release, None, tp_rank, None, True, ep_rank) print(f'load {checkpoint_name}') split_state = torch.load(checkpoint_name, map_location="cpu")['model'] for k, v in split_state.items(): if 'local_experts' in k and 'norm' not in k: local_expert_rank = name_to_expert_rank(k) expert_rank = local_expert_rank + num_local_experts * ep_rank k = k.replace(f'local_experts.{local_expert_rank}', f'local_experts.{expert_rank}') mid_state[k].append(v) elif ep_rank == 0: mid_state[k].append(v) for k, v in mid_state.items(): if not isinstance(v[0], torch.Tensor) or 'norm' in k or 'router' in k: target_v = v[0] elif 'embedding' in k or 'output_layer' in k: target_v = torch.cat(v, dim=0) elif 'linear_proj' in k or 'linear_fc2' in k: target_v = torch.cat(v, dim=1) elif 'linear_qkv.weight' in k: viewed = [x.view(group_per_split, -1, head_dim, args.hidden_size) for x in v] target_v = torch.cat(viewed, dim=0).view(-1, args.hidden_size) elif 'linear_qkv.bias' in k: viewed = [x.view(group_per_split, -1) for x in v] target_v = torch.cat(viewed, dim=0).view(-1) elif 'linear_fc1' in k: viewed = [x.view(2, -1, args.hidden_size) for x in v] target_v = torch.cat(viewed, dim=1).view(-1, args.hidden_size) else: print('passed', k) state_dict[k] = target_v else: raise ValueError('not support yet') model.load_state_dict(state_dict) return model def convert_checkpoint_from_megatron_to_transformers(mgmodel, hgmodel, args): query_group = args.num_query_groups hidden_size = args.hidden_size head_dim = hidden_size // args.num_attention_heads num_experts = args.moe_num_experts value_num_per_group = args.num_attention_heads // query_group with torch.no_grad(): hgmodel.model.embed_tokens.weight.copy_(mgmodel.embedding.word_embeddings.weight) for mglayer, hglayer in zip(mgmodel.decoder.layers, hgmodel.model.layers): hglayer.input_layernorm.weight.copy_(mglayer.self_attention.linear_qkv.layer_norm_weight) qkv_weight = mglayer.self_attention.linear_qkv.weight.view(query_group, -1, head_dim, hidden_size) q_weight, k_weight, v_weight = torch.split(qkv_weight, split_size_or_sections=[value_num_per_group, 1, 1], dim=1) hglayer.self_attn.q_proj.weight.copy_(q_weight.reshape(-1, hidden_size)) hglayer.self_attn.k_proj.weight.copy_(k_weight.reshape(-1, hidden_size)) hglayer.self_attn.v_proj.weight.copy_(v_weight.reshape(-1, hidden_size)) qkv_bias = mglayer.self_attention.linear_qkv.bias.view(query_group, -1) q_bias, k_bias, v_bias = torch.split(qkv_bias, split_size_or_sections=[q_weight.shape[2], k_weight.shape[2], v_weight.shape[2]], dim=1) hglayer.self_attn.q_proj.bias.copy_(q_bias.reshape(-1)) hglayer.self_attn.k_proj.bias.copy_(k_bias.reshape(-1)) hglayer.self_attn.v_proj.bias.copy_(v_bias.reshape(-1)) hglayer.self_attn.o_proj.weight.copy_(mglayer.self_attention.linear_proj.weight) if num_experts is None: gate_weight, fc1_weight = torch.split(mglayer.mlp.linear_fc1.weight, split_size_or_sections=args.ffn_hidden_size) hglayer.mlp.gate_proj.weight.copy_(gate_weight) hglayer.mlp.up_proj.weight.copy_(fc1_weight) hglayer.mlp.down_proj.weight.copy_(mglayer.mlp.linear_fc2.weight) hglayer.post_attention_layernorm.weight.copy_(mglayer.mlp.linear_fc1.layer_norm_weight) else: hglayer.post_attention_layernorm.weight.copy_(mglayer.pre_mlp_layernorm.weight) hglayer.mlp.gate.weight.copy_(mglayer.mlp.router.weight) for mgexpert, hgexpert in zip(mglayer.mlp.experts.local_experts, hglayer.mlp.experts): gate_weight, fc1_weight = torch.split(mgexpert.linear_fc1.weight, split_size_or_sections=args.ffn_hidden_size) hgexpert.w1.weight.copy_(gate_weight) hgexpert.w3.weight.copy_(fc1_weight) hgexpert.w2.weight.copy_(mgexpert.linear_fc2.weight) hgmodel.model.norm.weight.copy_(mgmodel.decoder.final_layernorm.weight) hgmodel.lm_head.weight.copy_(mgmodel.output_layer.weight) def transformers_to_megatron_fix_query_key_value_ordering( param, checkpoint_version, num_splits, num_heads, hidden_size ): """ Permutes layout of param tensor to the one compatible with respective NVIDIA Megatron-LM chekpoint versions. Input is [num_splits * num_heads * hidden_size, :] and output is [num_heads * hidden_size * num_splits, :] for version 1.0 and [num_heads * num_splits * hidden_size, :] for version 2.0 and later. If param is the weight tensor of the self-attention block, the param needs to be already transposed before calling this function. Args: param (torch.Tensor): the tensor to permute checkpoint_version (int): the version of the checkpoint. num_splits (int): the number of projections, usually 3 for (Query, Key, Value) num_heads (int): the number of attention heads hidden_size (int): the hidden size per head """ # Input is [num_splits * num_heads * hidden_size, :] input_shape = param.size() if checkpoint_version == 1.0: # version 1.0 stores [num_heads * hidden_size * num_splits, :] current_shape = (num_splits, num_heads, hidden_size) + input_shape[1:] param = param.view(*current_shape) param = param.transpose(0, 2) param = param.transpose(1, 2).contiguous() elif checkpoint_version >= 2.0: # other versions store [num_heads * num_splits * hidden_size, :] current_shape = (num_splits, num_heads, hidden_size) + input_shape[1:] param = param.view(*current_shape) param = param.transpose(0, 1).contiguous() param = param.view(*input_shape) return param def convert_checkpoint_from_transformers_to_megatron(mgmodel, hgmodel, args, hf_config): num_query_groups = hf_config.num_key_value_heads hidden_dim = hf_config.hidden_size head_dim = hidden_dim // hf_config.num_attention_heads num_experts = args.moe_num_experts num_local_experts = args.moe_num_experts // args.target_expert_model_parallel_size if args.moe_num_experts else 0 with torch.no_grad(): mgmodel.language_model.embedding.word_embeddings.weight.copy_(hgmodel.model.embed_tokens.weight) for mglayer, hglayer in zip(mgmodel.language_model.encoder.layers, hgmodel.model.layers): mglayer.input_norm.weight.copy_(hglayer.input_layernorm.weight) q = hglayer.self_attn.q_proj.weight.view([num_query_groups, -1, head_dim, hidden_dim]) k = hglayer.self_attn.k_proj.weight.view([num_query_groups, -1, head_dim, hidden_dim]) v = hglayer.self_attn.v_proj.weight.view([num_query_groups, -1, head_dim, hidden_dim]) qkv = torch.cat([q, k, v], dim=1).view(-1, hidden_dim).contiguous() q_bias = hglayer.self_attn.q_proj.bias.view([num_query_groups, -1]) k_bias = hglayer.self_attn.k_proj.bias.view([num_query_groups, -1]) v_bias = hglayer.self_attn.v_proj.bias.view([num_query_groups, -1]) qkv_bias = torch.cat([q_bias, k_bias, v_bias], dim=1).view(-1).contiguous() mglayer.self_attention.query_key_value.weight.copy_(qkv) mglayer.self_attention.query_key_value.bias.copy_(qkv_bias) mglayer.self_attention.dense.weight.copy_(hglayer.self_attn.o_proj.weight) fc1_weight = torch.cat([hglayer.mlp.gate_proj.weight, hglayer.mlp.up_proj.weight]) """ mlp layers.23.mlp.moe.router.layer.weight', 'layers.23.mlp.moe.experts.bias', 'layers.23.mlp.moe.experts.mlp.w1', 'layers.23.mlp.moe.experts.mlp.w2 glu 'layers.0.mlp.moe.router.layer.weight', : torch.Size([8, 1024]) 'layers.0.mlp.moe.experts.bias', : torch.Size([1024]) 'layers.0.mlp.moe.experts.mlp.w1',: torch.Size([2816, 1024]) 'layers.0.mlp.moe.experts.mlp.w2', : torch.Size([2816, 1024]) 'layers.0.mlp.moe.experts.mlp.v1', : torch.Size([2816, 1024]) """ #self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x)) #down->w2 , gate->v1, up->w1 if num_experts is None: mglayer.mlp.linear_fc1.weight.copy_(fc1_weight) mglayer.mlp.linear_fc2.weight.copy_(hglayer.mlp.down_proj.weight) mglayer.mlp.linear_fc1.layer_norm_weight.copy_(hglayer.post_attention_layernorm.weight) else: mglayer.post_attention_norm.weight.copy_(hglayer.post_attention_layernorm.weight) nn.init.normal_(mglayer.mlp.moe.router.layer.weight, mean=0, std=0.02) mglayer.mlp.moe.experts.mlp.w1.copy_(torch.cat([hglayer.mlp.up_proj.weight] * num_local_experts)) mglayer.mlp.moe.experts.mlp.w2.copy_(torch.cat([hglayer.mlp.down_proj.weight.transpose(0,1)] * num_local_experts)) mglayer.mlp.moe.experts.mlp.v1.copy_(torch.cat([hglayer.mlp.gate_proj.weight] * num_local_experts)) mgmodel.language_model.encoder.final_norm.weight.copy_(hgmodel.model.norm.weight) mgmodel.language_model.output_layer.weight.copy_(hgmodel.lm_head.weight) def save_state_dict(args, model, checkpoint_name): args.tensor_model_parallel_size = args.target_tensor_model_parallel_size args.pipeline_model_parallel_size = args.target_pipeline_model_parallel_size state_dict = {} state_dict['args'] = args state_dict['checkpoint_version'] = 3.0 state_dict['iteration'] = 0 state_dict['model'] = model os.makedirs(os.path.dirname(checkpoint_name), exist_ok=True) print(f'save model part {checkpoint_name}') torch.save(clone_state_dict(state_dict), checkpoint_name) def save_mgmodel(args, mgmodel, load_path, save_path): # Saving config and tokenzier files copy_huggingface_tokenizer(load_path, save_path) tracker_filepath = os.path.join(save_path, 'latest_checkpointed_iteration.txt') with open(tracker_filepath, "w") as f: f.write("release") head_dim = args.hidden_size // args.num_attention_heads group_per_split = args.num_query_groups // args.target_tensor_model_parallel_size full_model = mgmodel.state_dict_for_save_checkpoint() for k in list(full_model.keys()): if full_model[k] is None or "_extra_state" in k: full_model.pop(k) pattern = r'local_experts\.(\d+)\.' num_local_experts = args.moe_num_experts // args.target_expert_model_parallel_size if args.moe_num_experts else 0 if ( args.target_tensor_model_parallel_size == 1 and args.target_pipeline_model_parallel_size == 1 and args.target_expert_model_parallel_size == 1 ): checkpoint_name = get_checkpoint_names(save_path, 0, True) save_state_dict(args, full_model, checkpoint_name) elif ( args.target_tensor_model_parallel_size == 1 and args.target_pipeline_model_parallel_size == 1 and args.moe_num_experts and args.moe_num_experts % args.target_expert_model_parallel_size == 0 ): checkpoint_name = get_checkpoint_names(save_path, 0, False, True)[0] save_state_dict(args, full_model, checkpoint_name) elif ( args.target_tensor_model_parallel_size > 1 and args.target_pipeline_model_parallel_size == 1 and args.moe_num_experts is None ): for tp_rank in range(args.target_tensor_model_parallel_size): model_split = {} checkpoint_name = get_checkpoint_names(save_path, 0, True, None, tp_rank) print(f'tensor_parallel, save model to {checkpoint_name}') for k, v in full_model.items(): if not isinstance(v, torch.Tensor): target_v = v elif 'linear_qkv.weight' in k and 'norm' not in k: viewed = v.view(args.num_query_groups, -1, head_dim, args.hidden_size) viewed = viewed[group_per_split*tp_rank : group_per_split*(tp_rank + 1)] target_v = viewed.view(-1, args.hidden_size) elif 'linear_qkv.bias' in k and 'norm' not in k: viewed = v.view(args.num_query_groups, -1, head_dim) viewed = viewed[group_per_split * tp_rank: group_per_split * (tp_rank + 1)] target_v = viewed.view(-1) elif 'linear_proj' in k or 'linear_fc2' in k: seg = v.shape[1] // args.target_tensor_model_parallel_size target_v = v[:, seg*tp_rank : seg*(tp_rank + 1)] elif 'embedding' in k or 'output_layer' in k: seg = v.shape[0] // args.target_tensor_model_parallel_size target_v = v[seg*tp_rank : seg*(tp_rank + 1)] elif 'linear_fc1' in k and 'norm' not in k: viewed = v.view(-1, args.ffn_hidden_size, args.hidden_size) seg = args.ffn_hidden_size // args.target_tensor_model_parallel_size target_v = viewed[:, seg*tp_rank: seg*(tp_rank+1), :].reshape(-1, args.hidden_size) else: target_v = v model_split[k] = target_v save_state_dict(args, model_split, checkpoint_name) elif ( args.target_tensor_model_parallel_size > 1 and args.target_pipeline_model_parallel_size == 1 and args.moe_num_experts and args.moe_num_experts % args.target_expert_model_parallel_size == 0 ): for tp_rank in range(args.target_tensor_model_parallel_size): for ep_rank in range(args.target_expert_model_parallel_size): model_split = {} checkpoint_name = get_checkpoint_names(save_path, 0, True, None, tp_rank, None, True, ep_rank) for k, v in full_model.items(): if not isinstance(v, torch.Tensor): target_v = v elif 'linear_qkv.weight' in k and 'norm' not in k: viewed = v.view(args.num_query_groups, -1, head_dim, args.hidden_size) viewed = viewed[group_per_split*tp_rank : group_per_split*(tp_rank + 1)] target_v = viewed.view(-1, args.hidden_size) elif 'linear_qkv.bias' in k and 'norm' not in k: viewed = v.view(args.num_query_groups, -1, head_dim) viewed = viewed[group_per_split*tp_rank : group_per_split*(tp_rank + 1)] target_v = viewed.view(-1) elif 'linear_proj' in k: seg = v.shape[1] // args.target_tensor_model_parallel_size target_v = v[:, seg*tp_rank : seg*(tp_rank + 1)] elif 'embedding' in k or 'output_layer' in k: seg = v.shape[0] // args.target_tensor_model_parallel_size target_v = v[seg*tp_rank : seg*(tp_rank + 1)] elif 'local_experts' in k: expert_rank = int(re.findall(pattern, k)[0]) if expert_rank // num_local_experts != ep_rank: continue expert_local_rank = expert_rank % num_local_experts if 'linear_fc1' in k and 'norm' not in k: viewed = v.view(-1, args.ffn_hidden_size, args.hidden_size) seg = args.ffn_hidden_size // args.target_tensor_model_parallel_size target_v = viewed[:, seg*tp_rank: seg*(tp_rank+1), :].reshape(-1, args.hidden_size) elif 'linear_fc2' in k: seg = v.shape[1] // args.target_tensor_model_parallel_size target_v = v[:, seg*tp_rank : seg*(tp_rank + 1)] k = k.replace(f'local_experts.{expert_rank}', f'local_experts.{expert_local_rank}') else: target_v = v model_split[k] = target_v save_state_dict(args, model_split, checkpoint_name) else: raise ValueError('not support pp convert') print(f'megatron model is save to {save_path}') def save_hgmodel(args, model): output_state_dict = model.state_dict() max_shard_size=args.max_shard_size shards, index = shard_checkpoint(output_state_dict, max_shard_size=max_shard_size) os.makedirs(args.save_path, exist_ok=True) for shard_file, shard in shards.items(): target_file = os.path.join(args.save_path, shard_file) print(f'huggingface model is save to {target_file}') torch.save(clone_state_dict(shard), target_file) if index is None: print(f"Model weights saved in {os.path.join(args.save_path, WEIGHTS_NAME)}") else: save_index_file = os.path.join(args.save_path, WEIGHTS_INDEX_NAME) # Save the index as well with open(save_index_file, "w", encoding="utf-8") as f: content = json.dumps(index, indent=2, sort_keys=True) + "\n" f.write(content) print( f"The model is bigger than the maximum size per checkpoint ({args.max_shard_size}) and is going to be " f"split in {len(shards)} checkpoint shards. You can find where each parameters has been saved in the " f"index located at {save_index_file}." ) def add_ckpt_args(parser): parser = get_patch_args(parser) parser = add_checkpointing_args(parser) parser = add_megatron_checkpoint_args(parser) parser = add_transformers_checkpoint_args(parser) return parser def main(): initialize_megatron(extra_args_provider=add_ckpt_args) args = get_args() hf_config, hf_model = create_huggingface_model(args) mg_model = create_megatron_model(args, hf_config) if args.convert_checkpoint_from_megatron_to_transformers: load_megatron_model(args, mg_model) convert_checkpoint_from_megatron_to_transformers(mg_model, hf_model, args) save_hgmodel(args, hf_model) else: hf_model.from_pretrained(args.load_path) convert_checkpoint_from_transformers_to_megatron(mg_model, hf_model, args, hf_config) save_mgmodel(args, mg_model, args.load_path, args.save_path) if __name__ == "__main__": main()