modified docs

Signed-off-by: ftgreat <[email protected]>
silchan · Jun 8, 2023 · b9961b0 · b9961b0
1 parent 5fe0b59
commit b9961b0
Show file tree

Hide file tree

Showing 23 changed files with 2,650 additions and 834 deletions.
diff --git a/examples/Aquila/Aquila-code/generate_code.py b/examples/Aquila/Aquila-code/generate_code.py
@@ -18,30 +18,23 @@
 
 print(f"building model...")
 loader = AutoLoader("lm", model_name="aquilacode-7b-nv",
-                    only_download_config=True, 
                     use_cache=True, 
-                    fp16=True,
                     model_dir=model_dir)
 
 model = loader.get_model()
 tokenizer = loader.get_tokenizer()
-
+model.half()
 model.eval()
-
+model.cuda()
 model.to(device)
 
 vocab = tokenizer.get_vocab()
 
 id2word = {v:k for k, v in vocab.items()}
-
 predictor = Predictor(model, tokenizer)
 
 max_new_tokens = 256
 
-# test_file = "./datasets/code_test.txt"
-# with open(test_file) as fin:
-#     prompt = '\n'+fin.read()+'\n'
-
 texts = ["#补全代码\ndef quick_sort(x):", 
            '"""\n向用户询问他们的名字并说“你好”\m"""',
            '"""\nAsk the user for their name and say "Hello\n""""' ]
@@ -52,8 +45,11 @@
 
     max_length = input_length+max_new_tokens
     with torch.no_grad():
-        res = predictor.predict_generate_randomsample(prompt, 
+        res = predictor.predict_generate_randomsample(text, 
                                                         out_max_length=max_length, 
                                                         top_p=0.95, 
                                                         temperature=0.7)
-        print(res)
+        print(res)
+
+
+
diff --git a/examples/Aquila/Aquila-pretrain/README_Aquila.md b/examples/Aquila/Aquila-pretrain/README_Aquila.md
@@ -33,11 +33,11 @@ We also support [Huggingface](hflink)
 
 |   Model          |  License    | Commercial use?  |  GPU   | Model link                                     
 | :---------------- | :------- | :-- |:-- | :-- |    
-| <font color=red>Aquila-7B </font>         | Apache 2.0  |  ✅   | Nvidia-A100  | mhlink
+| <font color=red>Aquila-7B </font>         | Apache 2.0  |  ✅   | Nvidia-A100  | https://model.baai.ac.cn/model-detail/100098
 | <font color=red>Aquila-33B          | Apache 2.0  |  ✅  |  Nvidia-A100  | mhlink
-| AquilaCode-7B-nv          | Apache 2.0  |  ✅   |   Nvidia-A100   | mhlink
-| AquilaCode-7B-ts           | Apache 2.0  |  ✅    |  Tianshu-BI-V100   |    mhlink                
-| AquilaChat-7B           | Apache 2.0  |  ✅    | Nvidia-A100  | mhlink
+| AquilaCode-7B-nv          | Apache 2.0  |  ✅   |   Nvidia-A100   | https://model.baai.ac.cn/model-detail/100102
+| AquilaCode-7B-ts           | Apache 2.0  |  ✅    |  Tianshu-BI-V100   | https://model.baai.ac.cn/model-detail/100099                
+| AquilaChat-7B           | Apache 2.0  |  ✅    | Nvidia-A100  | https://model.baai.ac.cn/model-detail/100101
 
 我们使用了一系列更高效的底层算子来辅助模型训练，其中包括参考[flash-attention](https://github.com/HazyResearch/flash-attention)的方法并替换了一些中间计算，同时还使用了RMSNorm。在此基础上，我们升级了[BMtrain](https://github.com/OpenBMB/BMTrain)技术进行轻量化的并行训练，该技术采用了数据并行、ZeRO（零冗余优化器）、优化器卸载、检查点和操作融合、通信-计算重叠等方法来优化模型训练过程。
 

diff --git a/examples/Aquila/Aquila-sft/Aquila-sft.yaml b/examples/Aquila/Aquila-sft/Aquila-sft.yaml
@@ -1,13 +1,12 @@
-epochs: 3
 batch_size: 4
-gradient_accumulation_steps: 1
-lr: 9.65e-6
-warm_up: 0.1
-save_interval: 1000
+gradient_accumulation_steps: 4
+lr: 3.0e-4
+warm_up: 0.01
 
-bmt_lr_decay_style: "linear"
 bmt_cpu_offload: False
-
 bmt_pre_load: True
-enable_sft_dataset_dir: './data/'
-enable_sft_dataset_file: 'sft_samples.jsonl'
+
+save_optim: True
+save_rng: True
+enable_sft_dataset_dir: '../datasets/'
+enable_sft_dataset_file: 'convo_v2.jsonl'
diff --git a/examples/Aquila/Aquila-sft/aquila_sft.py b/examples/Aquila/Aquila-sft/aquila_sft.py
@@ -15,6 +15,7 @@
 import jsonlines
 import numpy as np
 from examples.Aquila import cyg_conversation as conversation_lib
+from flagai.model.tools.lora.prepare_lora import lora_transfer
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 
 # You can input all parameters by the command line.
@@ -69,20 +70,6 @@
 
 print('*'*20, "model_name", model_name, flush=True)
 
-'''
-auto_loader = AutoLoader(
-    "lm",
-    model_name=model_name,
-    model_dir=checkpoints,
-    only_download_config=True,
-)
-model = auto_loader.get_model()
-tokenizer = auto_loader.get_tokenizer()
-print('*'*20, "model", model)
-trainer.pre_train(model)
-print('*'*20, "model", model)
-
-'''
 
 cache_dir = os.path.join(checkpoints, model_name)
 print('*'*20, "cache_dir", cache_dir)
@@ -100,6 +87,11 @@
 model = AQUILAModel.init_from_json(config_file=config_file)
 print('*'*20, "model", model)
 
+#lora
+if env_args.lora:
+    model = lora_transfer(model,env_args)
+    model.print_trainable_parameters()
+
 ## bmt_pre_load
 checkpoint_path = os.path.join(cache_dir, "pytorch_model.bin")
 if env_args.bmt_pre_load:

diff --git a/examples/Aquila/Aquila-sft/generate_sft.py b/examples/Aquila/Aquila-sft/generate_sft.py
@@ -5,7 +5,6 @@
 from flagai.model.predictor.predictor import Predictor
 from flagai.model.predictor.aquila import aquila_generate
 from flagai.data.tokenizer import Tokenizer
-import bminf
 
 state_dict = "./checkpoints_in"
 model_name = 'aquilachat-7b'
@@ -21,8 +20,6 @@
 
 model.eval()
 model.half()
-# with torch.cuda.device(0):
-#     model = bminf.wrapper(model, quantization=False, memory_limit=2 << 30)
 model.cuda()
 
 predictor = Predictor(model, tokenizer)
@@ -107,5 +104,4 @@ def convo_tokenize(convo_obj, tokenizer):
 
     with torch.no_grad():
         out = aquila_generate(tokenizer, model, [text], max_gen_len:=200, top_p=0.95, prompts_tokens=[tokens])
-        print(f"pred is {out}")
-
+        print(f"pred is {out}")