Upload folder using huggingface_hub
Browse files- config.json +42 -0
- generation_config.json +6 -0
- latest +1 -0
- pytorch_model.bin +3 -0
- rng_state_0.pth +3 -0
- rng_state_1.pth +3 -0
- rng_state_2.pth +3 -0
- rng_state_3.pth +3 -0
- trainer_state.json +868 -0
- training_args.bin +3 -0
- zero_to_fp32.py +578 -0
config.json
ADDED
@@ -0,0 +1,42 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_name_or_path": "Salesforce/codegen-6B-multi",
|
3 |
+
"activation_function": "gelu_new",
|
4 |
+
"architectures": [
|
5 |
+
"CodeGenForCausalLM"
|
6 |
+
],
|
7 |
+
"attn_pdrop": 0.0,
|
8 |
+
"bos_token_id": 1,
|
9 |
+
"embd_pdrop": 0.0,
|
10 |
+
"eos_token_id": 50256,
|
11 |
+
"gradient_checkpointing": false,
|
12 |
+
"initializer_range": 0.02,
|
13 |
+
"layer_norm_epsilon": 1e-05,
|
14 |
+
"model_type": "codegen",
|
15 |
+
"n_ctx": 2048,
|
16 |
+
"n_embd": 4096,
|
17 |
+
"n_head": 16,
|
18 |
+
"n_inner": null,
|
19 |
+
"n_layer": 33,
|
20 |
+
"n_positions": 2048,
|
21 |
+
"resid_pdrop": 0.0,
|
22 |
+
"rotary_dim": 64,
|
23 |
+
"scale_attn_weights": true,
|
24 |
+
"summary_activation": null,
|
25 |
+
"summary_first_dropout": 0.1,
|
26 |
+
"summary_proj_to_labels": true,
|
27 |
+
"summary_type": "cls_index",
|
28 |
+
"summary_use_proj": true,
|
29 |
+
"task_specific_params": {
|
30 |
+
"text-generation": {
|
31 |
+
"do_sample": true,
|
32 |
+
"max_length": 50,
|
33 |
+
"temperature": 1.0
|
34 |
+
}
|
35 |
+
},
|
36 |
+
"tie_word_embeddings": false,
|
37 |
+
"tokenizer_class": "GPT2Tokenizer",
|
38 |
+
"torch_dtype": "float16",
|
39 |
+
"transformers_version": "4.27.1",
|
40 |
+
"use_cache": true,
|
41 |
+
"vocab_size": 51200
|
42 |
+
}
|
generation_config.json
ADDED
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_from_model_config": true,
|
3 |
+
"bos_token_id": 1,
|
4 |
+
"eos_token_id": 50256,
|
5 |
+
"transformers_version": "4.27.1"
|
6 |
+
}
|
latest
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
global_step142
|
pytorch_model.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:48b00581814ac8051be2131ac28cbee490d93a540b44b231fb63775c07053101
|
3 |
+
size 28810600769
|
rng_state_0.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e75c96f06b249e57a701db73ce821398e69672027a86d3a44063830602a29ab4
|
3 |
+
size 14583
|
rng_state_1.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:6dae7f45b6bac644ac207a61f43cba6d4b919a4cac22022bbb02907914422f5d
|
3 |
+
size 14583
|
rng_state_2.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e53c770fe48635faad7fa341007d771781f1397cd47daab5b58f879ffb65f178
|
3 |
+
size 14583
|
rng_state_3.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:68692af1001e65d02e07ac9974ccf4c332cfb23bc8f89566e1a908b1f2c4a1ed
|
3 |
+
size 14583
|
trainer_state.json
ADDED
@@ -0,0 +1,868 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": null,
|
3 |
+
"best_model_checkpoint": null,
|
4 |
+
"epoch": 1.0,
|
5 |
+
"global_step": 142,
|
6 |
+
"is_hyper_param_search": false,
|
7 |
+
"is_local_process_zero": true,
|
8 |
+
"is_world_process_zero": true,
|
9 |
+
"log_history": [
|
10 |
+
{
|
11 |
+
"epoch": 0.01,
|
12 |
+
"learning_rate": 0,
|
13 |
+
"loss": 3.8301,
|
14 |
+
"step": 1
|
15 |
+
},
|
16 |
+
{
|
17 |
+
"epoch": 0.01,
|
18 |
+
"learning_rate": 0,
|
19 |
+
"loss": 3.648,
|
20 |
+
"step": 2
|
21 |
+
},
|
22 |
+
{
|
23 |
+
"epoch": 0.02,
|
24 |
+
"learning_rate": 0,
|
25 |
+
"loss": 3.6693,
|
26 |
+
"step": 3
|
27 |
+
},
|
28 |
+
{
|
29 |
+
"epoch": 0.03,
|
30 |
+
"learning_rate": 0,
|
31 |
+
"loss": 3.6809,
|
32 |
+
"step": 4
|
33 |
+
},
|
34 |
+
{
|
35 |
+
"epoch": 0.04,
|
36 |
+
"learning_rate": 0,
|
37 |
+
"loss": 3.6776,
|
38 |
+
"step": 5
|
39 |
+
},
|
40 |
+
{
|
41 |
+
"epoch": 0.04,
|
42 |
+
"learning_rate": 0.0,
|
43 |
+
"loss": 3.6498,
|
44 |
+
"step": 6
|
45 |
+
},
|
46 |
+
{
|
47 |
+
"epoch": 0.05,
|
48 |
+
"learning_rate": 4.306765580733931e-06,
|
49 |
+
"loss": 2.0524,
|
50 |
+
"step": 7
|
51 |
+
},
|
52 |
+
{
|
53 |
+
"epoch": 0.06,
|
54 |
+
"learning_rate": 6.826061944859854e-06,
|
55 |
+
"loss": 1.9607,
|
56 |
+
"step": 8
|
57 |
+
},
|
58 |
+
{
|
59 |
+
"epoch": 0.06,
|
60 |
+
"learning_rate": 8.613531161467863e-06,
|
61 |
+
"loss": 0.8966,
|
62 |
+
"step": 9
|
63 |
+
},
|
64 |
+
{
|
65 |
+
"epoch": 0.07,
|
66 |
+
"learning_rate": 1e-05,
|
67 |
+
"loss": 0.292,
|
68 |
+
"step": 10
|
69 |
+
},
|
70 |
+
{
|
71 |
+
"epoch": 0.08,
|
72 |
+
"learning_rate": 1e-05,
|
73 |
+
"loss": 0.2447,
|
74 |
+
"step": 11
|
75 |
+
},
|
76 |
+
{
|
77 |
+
"epoch": 0.08,
|
78 |
+
"learning_rate": 9.97624703087886e-06,
|
79 |
+
"loss": 0.1913,
|
80 |
+
"step": 12
|
81 |
+
},
|
82 |
+
{
|
83 |
+
"epoch": 0.09,
|
84 |
+
"learning_rate": 9.95249406175772e-06,
|
85 |
+
"loss": 0.1367,
|
86 |
+
"step": 13
|
87 |
+
},
|
88 |
+
{
|
89 |
+
"epoch": 0.1,
|
90 |
+
"learning_rate": 9.92874109263658e-06,
|
91 |
+
"loss": 0.1032,
|
92 |
+
"step": 14
|
93 |
+
},
|
94 |
+
{
|
95 |
+
"epoch": 0.11,
|
96 |
+
"learning_rate": 9.90498812351544e-06,
|
97 |
+
"loss": 0.0797,
|
98 |
+
"step": 15
|
99 |
+
},
|
100 |
+
{
|
101 |
+
"epoch": 0.11,
|
102 |
+
"learning_rate": 9.8812351543943e-06,
|
103 |
+
"loss": 0.0641,
|
104 |
+
"step": 16
|
105 |
+
},
|
106 |
+
{
|
107 |
+
"epoch": 0.12,
|
108 |
+
"learning_rate": 9.857482185273159e-06,
|
109 |
+
"loss": 0.0501,
|
110 |
+
"step": 17
|
111 |
+
},
|
112 |
+
{
|
113 |
+
"epoch": 0.13,
|
114 |
+
"learning_rate": 9.83372921615202e-06,
|
115 |
+
"loss": 0.0494,
|
116 |
+
"step": 18
|
117 |
+
},
|
118 |
+
{
|
119 |
+
"epoch": 0.13,
|
120 |
+
"learning_rate": 9.80997624703088e-06,
|
121 |
+
"loss": 0.0497,
|
122 |
+
"step": 19
|
123 |
+
},
|
124 |
+
{
|
125 |
+
"epoch": 0.14,
|
126 |
+
"learning_rate": 9.78622327790974e-06,
|
127 |
+
"loss": 0.0453,
|
128 |
+
"step": 20
|
129 |
+
},
|
130 |
+
{
|
131 |
+
"epoch": 0.15,
|
132 |
+
"learning_rate": 9.7624703087886e-06,
|
133 |
+
"loss": 0.0384,
|
134 |
+
"step": 21
|
135 |
+
},
|
136 |
+
{
|
137 |
+
"epoch": 0.15,
|
138 |
+
"learning_rate": 9.73871733966746e-06,
|
139 |
+
"loss": 0.0332,
|
140 |
+
"step": 22
|
141 |
+
},
|
142 |
+
{
|
143 |
+
"epoch": 0.16,
|
144 |
+
"learning_rate": 9.714964370546319e-06,
|
145 |
+
"loss": 0.0328,
|
146 |
+
"step": 23
|
147 |
+
},
|
148 |
+
{
|
149 |
+
"epoch": 0.17,
|
150 |
+
"learning_rate": 9.69121140142518e-06,
|
151 |
+
"loss": 0.0335,
|
152 |
+
"step": 24
|
153 |
+
},
|
154 |
+
{
|
155 |
+
"epoch": 0.18,
|
156 |
+
"learning_rate": 9.66745843230404e-06,
|
157 |
+
"loss": 0.0337,
|
158 |
+
"step": 25
|
159 |
+
},
|
160 |
+
{
|
161 |
+
"epoch": 0.18,
|
162 |
+
"learning_rate": 9.643705463182899e-06,
|
163 |
+
"loss": 0.0335,
|
164 |
+
"step": 26
|
165 |
+
},
|
166 |
+
{
|
167 |
+
"epoch": 0.19,
|
168 |
+
"learning_rate": 9.619952494061758e-06,
|
169 |
+
"loss": 0.0348,
|
170 |
+
"step": 27
|
171 |
+
},
|
172 |
+
{
|
173 |
+
"epoch": 0.2,
|
174 |
+
"learning_rate": 9.596199524940617e-06,
|
175 |
+
"loss": 0.0318,
|
176 |
+
"step": 28
|
177 |
+
},
|
178 |
+
{
|
179 |
+
"epoch": 0.2,
|
180 |
+
"learning_rate": 9.572446555819479e-06,
|
181 |
+
"loss": 0.0309,
|
182 |
+
"step": 29
|
183 |
+
},
|
184 |
+
{
|
185 |
+
"epoch": 0.21,
|
186 |
+
"learning_rate": 9.548693586698338e-06,
|
187 |
+
"loss": 0.0303,
|
188 |
+
"step": 30
|
189 |
+
},
|
190 |
+
{
|
191 |
+
"epoch": 0.22,
|
192 |
+
"learning_rate": 9.524940617577197e-06,
|
193 |
+
"loss": 0.032,
|
194 |
+
"step": 31
|
195 |
+
},
|
196 |
+
{
|
197 |
+
"epoch": 0.23,
|
198 |
+
"learning_rate": 9.501187648456057e-06,
|
199 |
+
"loss": 0.0303,
|
200 |
+
"step": 32
|
201 |
+
},
|
202 |
+
{
|
203 |
+
"epoch": 0.23,
|
204 |
+
"learning_rate": 9.477434679334918e-06,
|
205 |
+
"loss": 0.0294,
|
206 |
+
"step": 33
|
207 |
+
},
|
208 |
+
{
|
209 |
+
"epoch": 0.24,
|
210 |
+
"learning_rate": 9.453681710213777e-06,
|
211 |
+
"loss": 0.0288,
|
212 |
+
"step": 34
|
213 |
+
},
|
214 |
+
{
|
215 |
+
"epoch": 0.25,
|
216 |
+
"learning_rate": 9.429928741092638e-06,
|
217 |
+
"loss": 0.0288,
|
218 |
+
"step": 35
|
219 |
+
},
|
220 |
+
{
|
221 |
+
"epoch": 0.25,
|
222 |
+
"learning_rate": 9.406175771971498e-06,
|
223 |
+
"loss": 0.0301,
|
224 |
+
"step": 36
|
225 |
+
},
|
226 |
+
{
|
227 |
+
"epoch": 0.26,
|
228 |
+
"learning_rate": 9.382422802850357e-06,
|
229 |
+
"loss": 0.0314,
|
230 |
+
"step": 37
|
231 |
+
},
|
232 |
+
{
|
233 |
+
"epoch": 0.27,
|
234 |
+
"learning_rate": 9.358669833729217e-06,
|
235 |
+
"loss": 0.0288,
|
236 |
+
"step": 38
|
237 |
+
},
|
238 |
+
{
|
239 |
+
"epoch": 0.27,
|
240 |
+
"learning_rate": 9.334916864608076e-06,
|
241 |
+
"loss": 0.0287,
|
242 |
+
"step": 39
|
243 |
+
},
|
244 |
+
{
|
245 |
+
"epoch": 0.28,
|
246 |
+
"learning_rate": 9.311163895486937e-06,
|
247 |
+
"loss": 0.0299,
|
248 |
+
"step": 40
|
249 |
+
},
|
250 |
+
{
|
251 |
+
"epoch": 0.29,
|
252 |
+
"learning_rate": 9.287410926365797e-06,
|
253 |
+
"loss": 0.0288,
|
254 |
+
"step": 41
|
255 |
+
},
|
256 |
+
{
|
257 |
+
"epoch": 0.3,
|
258 |
+
"learning_rate": 9.263657957244656e-06,
|
259 |
+
"loss": 0.0302,
|
260 |
+
"step": 42
|
261 |
+
},
|
262 |
+
{
|
263 |
+
"epoch": 0.3,
|
264 |
+
"learning_rate": 9.239904988123515e-06,
|
265 |
+
"loss": 0.0277,
|
266 |
+
"step": 43
|
267 |
+
},
|
268 |
+
{
|
269 |
+
"epoch": 0.31,
|
270 |
+
"learning_rate": 9.216152019002376e-06,
|
271 |
+
"loss": 0.0279,
|
272 |
+
"step": 44
|
273 |
+
},
|
274 |
+
{
|
275 |
+
"epoch": 0.32,
|
276 |
+
"learning_rate": 9.192399049881236e-06,
|
277 |
+
"loss": 0.0285,
|
278 |
+
"step": 45
|
279 |
+
},
|
280 |
+
{
|
281 |
+
"epoch": 0.32,
|
282 |
+
"learning_rate": 9.168646080760095e-06,
|
283 |
+
"loss": 0.0299,
|
284 |
+
"step": 46
|
285 |
+
},
|
286 |
+
{
|
287 |
+
"epoch": 0.33,
|
288 |
+
"learning_rate": 9.144893111638956e-06,
|
289 |
+
"loss": 0.029,
|
290 |
+
"step": 47
|
291 |
+
},
|
292 |
+
{
|
293 |
+
"epoch": 0.34,
|
294 |
+
"learning_rate": 9.121140142517816e-06,
|
295 |
+
"loss": 0.0293,
|
296 |
+
"step": 48
|
297 |
+
},
|
298 |
+
{
|
299 |
+
"epoch": 0.35,
|
300 |
+
"learning_rate": 9.097387173396675e-06,
|
301 |
+
"loss": 0.0293,
|
302 |
+
"step": 49
|
303 |
+
},
|
304 |
+
{
|
305 |
+
"epoch": 0.35,
|
306 |
+
"learning_rate": 9.073634204275536e-06,
|
307 |
+
"loss": 0.029,
|
308 |
+
"step": 50
|
309 |
+
},
|
310 |
+
{
|
311 |
+
"epoch": 0.36,
|
312 |
+
"learning_rate": 9.049881235154396e-06,
|
313 |
+
"loss": 0.0295,
|
314 |
+
"step": 51
|
315 |
+
},
|
316 |
+
{
|
317 |
+
"epoch": 0.37,
|
318 |
+
"learning_rate": 9.026128266033255e-06,
|
319 |
+
"loss": 0.0284,
|
320 |
+
"step": 52
|
321 |
+
},
|
322 |
+
{
|
323 |
+
"epoch": 0.37,
|
324 |
+
"learning_rate": 9.002375296912114e-06,
|
325 |
+
"loss": 0.0298,
|
326 |
+
"step": 53
|
327 |
+
},
|
328 |
+
{
|
329 |
+
"epoch": 0.38,
|
330 |
+
"learning_rate": 8.978622327790974e-06,
|
331 |
+
"loss": 0.0286,
|
332 |
+
"step": 54
|
333 |
+
},
|
334 |
+
{
|
335 |
+
"epoch": 0.39,
|
336 |
+
"learning_rate": 8.954869358669835e-06,
|
337 |
+
"loss": 0.0292,
|
338 |
+
"step": 55
|
339 |
+
},
|
340 |
+
{
|
341 |
+
"epoch": 0.39,
|
342 |
+
"learning_rate": 8.931116389548694e-06,
|
343 |
+
"loss": 0.0283,
|
344 |
+
"step": 56
|
345 |
+
},
|
346 |
+
{
|
347 |
+
"epoch": 0.4,
|
348 |
+
"learning_rate": 8.907363420427554e-06,
|
349 |
+
"loss": 0.0286,
|
350 |
+
"step": 57
|
351 |
+
},
|
352 |
+
{
|
353 |
+
"epoch": 0.41,
|
354 |
+
"learning_rate": 8.883610451306413e-06,
|
355 |
+
"loss": 0.029,
|
356 |
+
"step": 58
|
357 |
+
},
|
358 |
+
{
|
359 |
+
"epoch": 0.42,
|
360 |
+
"learning_rate": 8.859857482185273e-06,
|
361 |
+
"loss": 0.0282,
|
362 |
+
"step": 59
|
363 |
+
},
|
364 |
+
{
|
365 |
+
"epoch": 0.42,
|
366 |
+
"learning_rate": 8.836104513064134e-06,
|
367 |
+
"loss": 0.0281,
|
368 |
+
"step": 60
|
369 |
+
},
|
370 |
+
{
|
371 |
+
"epoch": 0.43,
|
372 |
+
"learning_rate": 8.812351543942995e-06,
|
373 |
+
"loss": 0.0281,
|
374 |
+
"step": 61
|
375 |
+
},
|
376 |
+
{
|
377 |
+
"epoch": 0.44,
|
378 |
+
"learning_rate": 8.788598574821854e-06,
|
379 |
+
"loss": 0.028,
|
380 |
+
"step": 62
|
381 |
+
},
|
382 |
+
{
|
383 |
+
"epoch": 0.44,
|
384 |
+
"learning_rate": 8.764845605700714e-06,
|
385 |
+
"loss": 0.0286,
|
386 |
+
"step": 63
|
387 |
+
},
|
388 |
+
{
|
389 |
+
"epoch": 0.45,
|
390 |
+
"learning_rate": 8.741092636579573e-06,
|
391 |
+
"loss": 0.0293,
|
392 |
+
"step": 64
|
393 |
+
},
|
394 |
+
{
|
395 |
+
"epoch": 0.46,
|
396 |
+
"learning_rate": 8.717339667458432e-06,
|
397 |
+
"loss": 0.0285,
|
398 |
+
"step": 65
|
399 |
+
},
|
400 |
+
{
|
401 |
+
"epoch": 0.46,
|
402 |
+
"learning_rate": 8.693586698337293e-06,
|
403 |
+
"loss": 0.0281,
|
404 |
+
"step": 66
|
405 |
+
},
|
406 |
+
{
|
407 |
+
"epoch": 0.47,
|
408 |
+
"learning_rate": 8.669833729216153e-06,
|
409 |
+
"loss": 0.0284,
|
410 |
+
"step": 67
|
411 |
+
},
|
412 |
+
{
|
413 |
+
"epoch": 0.48,
|
414 |
+
"learning_rate": 8.646080760095012e-06,
|
415 |
+
"loss": 0.0288,
|
416 |
+
"step": 68
|
417 |
+
},
|
418 |
+
{
|
419 |
+
"epoch": 0.49,
|
420 |
+
"learning_rate": 8.622327790973872e-06,
|
421 |
+
"loss": 0.0288,
|
422 |
+
"step": 69
|
423 |
+
},
|
424 |
+
{
|
425 |
+
"epoch": 0.49,
|
426 |
+
"learning_rate": 8.598574821852733e-06,
|
427 |
+
"loss": 0.0283,
|
428 |
+
"step": 70
|
429 |
+
},
|
430 |
+
{
|
431 |
+
"epoch": 0.5,
|
432 |
+
"learning_rate": 8.574821852731592e-06,
|
433 |
+
"loss": 0.0278,
|
434 |
+
"step": 71
|
435 |
+
},
|
436 |
+
{
|
437 |
+
"epoch": 0.51,
|
438 |
+
"learning_rate": 8.551068883610452e-06,
|
439 |
+
"loss": 0.0281,
|
440 |
+
"step": 72
|
441 |
+
},
|
442 |
+
{
|
443 |
+
"epoch": 0.51,
|
444 |
+
"learning_rate": 8.527315914489311e-06,
|
445 |
+
"loss": 0.0276,
|
446 |
+
"step": 73
|
447 |
+
},
|
448 |
+
{
|
449 |
+
"epoch": 0.52,
|
450 |
+
"learning_rate": 8.50356294536817e-06,
|
451 |
+
"loss": 0.0282,
|
452 |
+
"step": 74
|
453 |
+
},
|
454 |
+
{
|
455 |
+
"epoch": 0.53,
|
456 |
+
"learning_rate": 8.479809976247032e-06,
|
457 |
+
"loss": 0.0281,
|
458 |
+
"step": 75
|
459 |
+
},
|
460 |
+
{
|
461 |
+
"epoch": 0.54,
|
462 |
+
"learning_rate": 8.456057007125893e-06,
|
463 |
+
"loss": 0.0282,
|
464 |
+
"step": 76
|
465 |
+
},
|
466 |
+
{
|
467 |
+
"epoch": 0.54,
|
468 |
+
"learning_rate": 8.432304038004752e-06,
|
469 |
+
"loss": 0.0282,
|
470 |
+
"step": 77
|
471 |
+
},
|
472 |
+
{
|
473 |
+
"epoch": 0.55,
|
474 |
+
"learning_rate": 8.408551068883611e-06,
|
475 |
+
"loss": 0.0281,
|
476 |
+
"step": 78
|
477 |
+
},
|
478 |
+
{
|
479 |
+
"epoch": 0.56,
|
480 |
+
"learning_rate": 8.38479809976247e-06,
|
481 |
+
"loss": 0.0282,
|
482 |
+
"step": 79
|
483 |
+
},
|
484 |
+
{
|
485 |
+
"epoch": 0.56,
|
486 |
+
"learning_rate": 8.36104513064133e-06,
|
487 |
+
"loss": 0.0293,
|
488 |
+
"step": 80
|
489 |
+
},
|
490 |
+
{
|
491 |
+
"epoch": 0.57,
|
492 |
+
"learning_rate": 8.337292161520191e-06,
|
493 |
+
"loss": 0.0284,
|
494 |
+
"step": 81
|
495 |
+
},
|
496 |
+
{
|
497 |
+
"epoch": 0.58,
|
498 |
+
"learning_rate": 8.31353919239905e-06,
|
499 |
+
"loss": 0.028,
|
500 |
+
"step": 82
|
501 |
+
},
|
502 |
+
{
|
503 |
+
"epoch": 0.58,
|
504 |
+
"learning_rate": 8.28978622327791e-06,
|
505 |
+
"loss": 0.0281,
|
506 |
+
"step": 83
|
507 |
+
},
|
508 |
+
{
|
509 |
+
"epoch": 0.59,
|
510 |
+
"learning_rate": 8.26603325415677e-06,
|
511 |
+
"loss": 0.0285,
|
512 |
+
"step": 84
|
513 |
+
},
|
514 |
+
{
|
515 |
+
"epoch": 0.6,
|
516 |
+
"learning_rate": 8.24228028503563e-06,
|
517 |
+
"loss": 0.0271,
|
518 |
+
"step": 85
|
519 |
+
},
|
520 |
+
{
|
521 |
+
"epoch": 0.61,
|
522 |
+
"learning_rate": 8.21852731591449e-06,
|
523 |
+
"loss": 0.0284,
|
524 |
+
"step": 86
|
525 |
+
},
|
526 |
+
{
|
527 |
+
"epoch": 0.61,
|
528 |
+
"learning_rate": 8.19477434679335e-06,
|
529 |
+
"loss": 0.0294,
|
530 |
+
"step": 87
|
531 |
+
},
|
532 |
+
{
|
533 |
+
"epoch": 0.62,
|
534 |
+
"learning_rate": 8.171021377672209e-06,
|
535 |
+
"loss": 0.0286,
|
536 |
+
"step": 88
|
537 |
+
},
|
538 |
+
{
|
539 |
+
"epoch": 0.63,
|
540 |
+
"learning_rate": 8.14726840855107e-06,
|
541 |
+
"loss": 0.0276,
|
542 |
+
"step": 89
|
543 |
+
},
|
544 |
+
{
|
545 |
+
"epoch": 0.63,
|
546 |
+
"learning_rate": 8.12351543942993e-06,
|
547 |
+
"loss": 0.028,
|
548 |
+
"step": 90
|
549 |
+
},
|
550 |
+
{
|
551 |
+
"epoch": 0.64,
|
552 |
+
"learning_rate": 8.099762470308789e-06,
|
553 |
+
"loss": 0.0274,
|
554 |
+
"step": 91
|
555 |
+
},
|
556 |
+
{
|
557 |
+
"epoch": 0.65,
|
558 |
+
"learning_rate": 8.07600950118765e-06,
|
559 |
+
"loss": 0.028,
|
560 |
+
"step": 92
|
561 |
+
},
|
562 |
+
{
|
563 |
+
"epoch": 0.65,
|
564 |
+
"learning_rate": 8.05225653206651e-06,
|
565 |
+
"loss": 0.0287,
|
566 |
+
"step": 93
|
567 |
+
},
|
568 |
+
{
|
569 |
+
"epoch": 0.66,
|
570 |
+
"learning_rate": 8.028503562945369e-06,
|
571 |
+
"loss": 0.0297,
|
572 |
+
"step": 94
|
573 |
+
},
|
574 |
+
{
|
575 |
+
"epoch": 0.67,
|
576 |
+
"learning_rate": 8.004750593824228e-06,
|
577 |
+
"loss": 0.028,
|
578 |
+
"step": 95
|
579 |
+
},
|
580 |
+
{
|
581 |
+
"epoch": 0.68,
|
582 |
+
"learning_rate": 7.98099762470309e-06,
|
583 |
+
"loss": 0.0288,
|
584 |
+
"step": 96
|
585 |
+
},
|
586 |
+
{
|
587 |
+
"epoch": 0.68,
|
588 |
+
"learning_rate": 7.957244655581949e-06,
|
589 |
+
"loss": 0.0278,
|
590 |
+
"step": 97
|
591 |
+
},
|
592 |
+
{
|
593 |
+
"epoch": 0.69,
|
594 |
+
"learning_rate": 7.933491686460808e-06,
|
595 |
+
"loss": 0.0281,
|
596 |
+
"step": 98
|
597 |
+
},
|
598 |
+
{
|
599 |
+
"epoch": 0.7,
|
600 |
+
"learning_rate": 7.909738717339667e-06,
|
601 |
+
"loss": 0.0286,
|
602 |
+
"step": 99
|
603 |
+
},
|
604 |
+
{
|
605 |
+
"epoch": 0.7,
|
606 |
+
"learning_rate": 7.885985748218527e-06,
|
607 |
+
"loss": 0.0282,
|
608 |
+
"step": 100
|
609 |
+
},
|
610 |
+
{
|
611 |
+
"epoch": 0.71,
|
612 |
+
"learning_rate": 7.862232779097388e-06,
|
613 |
+
"loss": 0.0279,
|
614 |
+
"step": 101
|
615 |
+
},
|
616 |
+
{
|
617 |
+
"epoch": 0.72,
|
618 |
+
"learning_rate": 7.838479809976247e-06,
|
619 |
+
"loss": 0.0283,
|
620 |
+
"step": 102
|
621 |
+
},
|
622 |
+
{
|
623 |
+
"epoch": 0.73,
|
624 |
+
"learning_rate": 7.814726840855108e-06,
|
625 |
+
"loss": 0.0293,
|
626 |
+
"step": 103
|
627 |
+
},
|
628 |
+
{
|
629 |
+
"epoch": 0.73,
|
630 |
+
"learning_rate": 7.790973871733968e-06,
|
631 |
+
"loss": 0.0287,
|
632 |
+
"step": 104
|
633 |
+
},
|
634 |
+
{
|
635 |
+
"epoch": 0.74,
|
636 |
+
"learning_rate": 7.767220902612827e-06,
|
637 |
+
"loss": 0.0272,
|
638 |
+
"step": 105
|
639 |
+
},
|
640 |
+
{
|
641 |
+
"epoch": 0.75,
|
642 |
+
"learning_rate": 7.743467933491687e-06,
|
643 |
+
"loss": 0.0288,
|
644 |
+
"step": 106
|
645 |
+
},
|
646 |
+
{
|
647 |
+
"epoch": 0.75,
|
648 |
+
"learning_rate": 7.719714964370548e-06,
|
649 |
+
"loss": 0.0292,
|
650 |
+
"step": 107
|
651 |
+
},
|
652 |
+
{
|
653 |
+
"epoch": 0.76,
|
654 |
+
"learning_rate": 7.695961995249407e-06,
|
655 |
+
"loss": 0.0287,
|
656 |
+
"step": 108
|
657 |
+
},
|
658 |
+
{
|
659 |
+
"epoch": 0.77,
|
660 |
+
"learning_rate": 7.672209026128267e-06,
|
661 |
+
"loss": 0.0286,
|
662 |
+
"step": 109
|
663 |
+
},
|
664 |
+
{
|
665 |
+
"epoch": 0.77,
|
666 |
+
"learning_rate": 7.648456057007126e-06,
|
667 |
+
"loss": 0.0293,
|
668 |
+
"step": 110
|
669 |
+
},
|
670 |
+
{
|
671 |
+
"epoch": 0.78,
|
672 |
+
"learning_rate": 7.624703087885986e-06,
|
673 |
+
"loss": 0.0278,
|
674 |
+
"step": 111
|
675 |
+
},
|
676 |
+
{
|
677 |
+
"epoch": 0.79,
|
678 |
+
"learning_rate": 7.600950118764846e-06,
|
679 |
+
"loss": 0.028,
|
680 |
+
"step": 112
|
681 |
+
},
|
682 |
+
{
|
683 |
+
"epoch": 0.8,
|
684 |
+
"learning_rate": 7.577197149643706e-06,
|
685 |
+
"loss": 0.0284,
|
686 |
+
"step": 113
|
687 |
+
},
|
688 |
+
{
|
689 |
+
"epoch": 0.8,
|
690 |
+
"learning_rate": 7.553444180522565e-06,
|
691 |
+
"loss": 0.0272,
|
692 |
+
"step": 114
|
693 |
+
},
|
694 |
+
{
|
695 |
+
"epoch": 0.81,
|
696 |
+
"learning_rate": 7.5296912114014255e-06,
|
697 |
+
"loss": 0.028,
|
698 |
+
"step": 115
|
699 |
+
},
|
700 |
+
{
|
701 |
+
"epoch": 0.82,
|
702 |
+
"learning_rate": 7.505938242280285e-06,
|
703 |
+
"loss": 0.0292,
|
704 |
+
"step": 116
|
705 |
+
},
|
706 |
+
{
|
707 |
+
"epoch": 0.82,
|
708 |
+
"learning_rate": 7.482185273159146e-06,
|
709 |
+
"loss": 0.0273,
|
710 |
+
"step": 117
|
711 |
+
},
|
712 |
+
{
|
713 |
+
"epoch": 0.83,
|
714 |
+
"learning_rate": 7.458432304038005e-06,
|
715 |
+
"loss": 0.0287,
|
716 |
+
"step": 118
|
717 |
+
},
|
718 |
+
{
|
719 |
+
"epoch": 0.84,
|
720 |
+
"learning_rate": 7.434679334916866e-06,
|
721 |
+
"loss": 0.0297,
|
722 |
+
"step": 119
|
723 |
+
},
|
724 |
+
{
|
725 |
+
"epoch": 0.85,
|
726 |
+
"learning_rate": 7.410926365795725e-06,
|
727 |
+
"loss": 0.0299,
|
728 |
+
"step": 120
|
729 |
+
},
|
730 |
+
{
|
731 |
+
"epoch": 0.85,
|
732 |
+
"learning_rate": 7.387173396674585e-06,
|
733 |
+
"loss": 0.0276,
|
734 |
+
"step": 121
|
735 |
+
},
|
736 |
+
{
|
737 |
+
"epoch": 0.86,
|
738 |
+
"learning_rate": 7.363420427553445e-06,
|
739 |
+
"loss": 0.0288,
|
740 |
+
"step": 122
|
741 |
+
},
|
742 |
+
{
|
743 |
+
"epoch": 0.87,
|
744 |
+
"learning_rate": 7.339667458432305e-06,
|
745 |
+
"loss": 0.0276,
|
746 |
+
"step": 123
|
747 |
+
},
|
748 |
+
{
|
749 |
+
"epoch": 0.87,
|
750 |
+
"learning_rate": 7.315914489311164e-06,
|
751 |
+
"loss": 0.0278,
|
752 |
+
"step": 124
|
753 |
+
},
|
754 |
+
{
|
755 |
+
"epoch": 0.88,
|
756 |
+
"learning_rate": 7.292161520190024e-06,
|
757 |
+
"loss": 0.0283,
|
758 |
+
"step": 125
|
759 |
+
},
|
760 |
+
{
|
761 |
+
"epoch": 0.89,
|
762 |
+
"learning_rate": 7.268408551068884e-06,
|
763 |
+
"loss": 0.0282,
|
764 |
+
"step": 126
|
765 |
+
},
|
766 |
+
{
|
767 |
+
"epoch": 0.89,
|
768 |
+
"learning_rate": 7.2446555819477435e-06,
|
769 |
+
"loss": 0.0288,
|
770 |
+
"step": 127
|
771 |
+
},
|
772 |
+
{
|
773 |
+
"epoch": 0.9,
|
774 |
+
"learning_rate": 7.220902612826604e-06,
|
775 |
+
"loss": 0.0289,
|
776 |
+
"step": 128
|
777 |
+
},
|
778 |
+
{
|
779 |
+
"epoch": 0.91,
|
780 |
+
"learning_rate": 7.197149643705463e-06,
|
781 |
+
"loss": 0.0283,
|
782 |
+
"step": 129
|
783 |
+
},
|
784 |
+
{
|
785 |
+
"epoch": 0.92,
|
786 |
+
"learning_rate": 7.173396674584323e-06,
|
787 |
+
"loss": 0.0284,
|
788 |
+
"step": 130
|
789 |
+
},
|
790 |
+
{
|
791 |
+
"epoch": 0.92,
|
792 |
+
"learning_rate": 7.149643705463184e-06,
|
793 |
+
"loss": 0.0291,
|
794 |
+
"step": 131
|
795 |
+
},
|
796 |
+
{
|
797 |
+
"epoch": 0.93,
|
798 |
+
"learning_rate": 7.125890736342044e-06,
|
799 |
+
"loss": 0.0286,
|
800 |
+
"step": 132
|
801 |
+
},
|
802 |
+
{
|
803 |
+
"epoch": 0.94,
|
804 |
+
"learning_rate": 7.102137767220903e-06,
|
805 |
+
"loss": 0.0276,
|
806 |
+
"step": 133
|
807 |
+
},
|
808 |
+
{
|
809 |
+
"epoch": 0.94,
|
810 |
+
"learning_rate": 7.0783847980997635e-06,
|
811 |
+
"loss": 0.0283,
|
812 |
+
"step": 134
|
813 |
+
},
|
814 |
+
{
|
815 |
+
"epoch": 0.95,
|
816 |
+
"learning_rate": 7.054631828978623e-06,
|
817 |
+
"loss": 0.0296,
|
818 |
+
"step": 135
|
819 |
+
},
|
820 |
+
{
|
821 |
+
"epoch": 0.96,
|
822 |
+
"learning_rate": 7.030878859857483e-06,
|
823 |
+
"loss": 0.028,
|
824 |
+
"step": 136
|
825 |
+
},
|
826 |
+
{
|
827 |
+
"epoch": 0.96,
|
828 |
+
"learning_rate": 7.007125890736343e-06,
|
829 |
+
"loss": 0.0289,
|
830 |
+
"step": 137
|
831 |
+
},
|
832 |
+
{
|
833 |
+
"epoch": 0.97,
|
834 |
+
"learning_rate": 6.983372921615203e-06,
|
835 |
+
"loss": 0.0289,
|
836 |
+
"step": 138
|
837 |
+
},
|
838 |
+
{
|
839 |
+
"epoch": 0.98,
|
840 |
+
"learning_rate": 6.959619952494062e-06,
|
841 |
+
"loss": 0.0286,
|
842 |
+
"step": 139
|
843 |
+
},
|
844 |
+
{
|
845 |
+
"epoch": 0.99,
|
846 |
+
"learning_rate": 6.935866983372922e-06,
|
847 |
+
"loss": 0.0276,
|
848 |
+
"step": 140
|
849 |
+
},
|
850 |
+
{
|
851 |
+
"epoch": 0.99,
|
852 |
+
"learning_rate": 6.912114014251782e-06,
|
853 |
+
"loss": 0.0277,
|
854 |
+
"step": 141
|
855 |
+
},
|
856 |
+
{
|
857 |
+
"epoch": 1.0,
|
858 |
+
"learning_rate": 6.888361045130641e-06,
|
859 |
+
"loss": 0.0279,
|
860 |
+
"step": 142
|
861 |
+
}
|
862 |
+
],
|
863 |
+
"max_steps": 426,
|
864 |
+
"num_train_epochs": 3,
|
865 |
+
"total_flos": 86302131093504.0,
|
866 |
+
"trial_name": null,
|
867 |
+
"trial_params": null
|
868 |
+
}
|
training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f470866f2ac29d8436396050d4c60e557eea6d7c62c788720e9543fe379beae3
|
3 |
+
size 5115
|
zero_to_fp32.py
ADDED
@@ -0,0 +1,578 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
#!/usr/bin/env python
|
2 |
+
|
3 |
+
# Copyright (c) Microsoft Corporation.
|
4 |
+
# SPDX-License-Identifier: Apache-2.0
|
5 |
+
|
6 |
+
# DeepSpeed Team
|
7 |
+
|
8 |
+
# This script extracts fp32 consolidated weights from a zero 2 and 3 DeepSpeed checkpoints. It gets
|
9 |
+
# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
|
10 |
+
# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
|
11 |
+
# application.
|
12 |
+
#
|
13 |
+
# example: python zero_to_fp32.py . pytorch_model.bin
|
14 |
+
|
15 |
+
import argparse
|
16 |
+
import torch
|
17 |
+
import glob
|
18 |
+
import math
|
19 |
+
import os
|
20 |
+
import re
|
21 |
+
from collections import OrderedDict
|
22 |
+
from dataclasses import dataclass
|
23 |
+
|
24 |
+
# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
|
25 |
+
# DeepSpeed data structures it has to be available in the current python environment.
|
26 |
+
from deepspeed.utils import logger
|
27 |
+
from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
|
28 |
+
FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
|
29 |
+
FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
|
30 |
+
|
31 |
+
|
32 |
+
@dataclass
|
33 |
+
class zero_model_state:
|
34 |
+
buffers: dict()
|
35 |
+
param_shapes: dict()
|
36 |
+
shared_params: list
|
37 |
+
ds_version: int
|
38 |
+
frozen_param_shapes: dict()
|
39 |
+
frozen_param_fragments: dict()
|
40 |
+
|
41 |
+
|
42 |
+
debug = 0
|
43 |
+
|
44 |
+
# load to cpu
|
45 |
+
device = torch.device('cpu')
|
46 |
+
|
47 |
+
|
48 |
+
def atoi(text):
|
49 |
+
return int(text) if text.isdigit() else text
|
50 |
+
|
51 |
+
|
52 |
+
def natural_keys(text):
|
53 |
+
'''
|
54 |
+
alist.sort(key=natural_keys) sorts in human order
|
55 |
+
http://nedbatchelder.com/blog/200712/human_sorting.html
|
56 |
+
(See Toothy's implementation in the comments)
|
57 |
+
'''
|
58 |
+
return [atoi(c) for c in re.split(r'(\d+)', text)]
|
59 |
+
|
60 |
+
|
61 |
+
def get_model_state_file(checkpoint_dir, zero_stage):
|
62 |
+
if not os.path.isdir(checkpoint_dir):
|
63 |
+
raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
|
64 |
+
|
65 |
+
# there should be only one file
|
66 |
+
if zero_stage == 2:
|
67 |
+
file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
|
68 |
+
elif zero_stage == 3:
|
69 |
+
file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
|
70 |
+
|
71 |
+
if not os.path.exists(file):
|
72 |
+
raise FileNotFoundError(f"can't find model states file at '{file}'")
|
73 |
+
|
74 |
+
return file
|
75 |
+
|
76 |
+
|
77 |
+
def get_checkpoint_files(checkpoint_dir, glob_pattern):
|
78 |
+
# XXX: need to test that this simple glob rule works for multi-node setup too
|
79 |
+
ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
|
80 |
+
|
81 |
+
if len(ckpt_files) == 0:
|
82 |
+
raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
|
83 |
+
|
84 |
+
return ckpt_files
|
85 |
+
|
86 |
+
|
87 |
+
def get_optim_files(checkpoint_dir):
|
88 |
+
return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
|
89 |
+
|
90 |
+
|
91 |
+
def get_model_state_files(checkpoint_dir):
|
92 |
+
return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
|
93 |
+
|
94 |
+
|
95 |
+
def parse_model_states(files):
|
96 |
+
zero_model_states = []
|
97 |
+
for file in files:
|
98 |
+
state_dict = torch.load(file, map_location=device)
|
99 |
+
|
100 |
+
if BUFFER_NAMES not in state_dict:
|
101 |
+
raise ValueError(f"{file} is not a model state checkpoint")
|
102 |
+
buffer_names = state_dict[BUFFER_NAMES]
|
103 |
+
if debug:
|
104 |
+
print("Found buffers:", buffer_names)
|
105 |
+
|
106 |
+
# recover just the buffers while restoring them to fp32 if they were saved in fp16
|
107 |
+
buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
|
108 |
+
param_shapes = state_dict[PARAM_SHAPES]
|
109 |
+
|
110 |
+
# collect parameters that are included in param_shapes
|
111 |
+
param_names = []
|
112 |
+
for s in param_shapes:
|
113 |
+
for name in s.keys():
|
114 |
+
param_names.append(name)
|
115 |
+
|
116 |
+
# update with frozen parameters
|
117 |
+
frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
|
118 |
+
if frozen_param_shapes is not None:
|
119 |
+
if debug:
|
120 |
+
print(f"Found frozen_param_shapes: {frozen_param_shapes}")
|
121 |
+
param_names += list(frozen_param_shapes.keys())
|
122 |
+
|
123 |
+
# handle shared params
|
124 |
+
shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
|
125 |
+
|
126 |
+
ds_version = state_dict.get(DS_VERSION, None)
|
127 |
+
|
128 |
+
frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
|
129 |
+
|
130 |
+
z_model_state = zero_model_state(buffers=buffers,
|
131 |
+
param_shapes=param_shapes,
|
132 |
+
shared_params=shared_params,
|
133 |
+
ds_version=ds_version,
|
134 |
+
frozen_param_shapes=frozen_param_shapes,
|
135 |
+
frozen_param_fragments=frozen_param_fragments)
|
136 |
+
zero_model_states.append(z_model_state)
|
137 |
+
|
138 |
+
return zero_model_states
|
139 |
+
|
140 |
+
|
141 |
+
def parse_optim_states(files, ds_checkpoint_dir):
|
142 |
+
|
143 |
+
total_files = len(files)
|
144 |
+
state_dicts = []
|
145 |
+
for f in files:
|
146 |
+
state_dicts.append(torch.load(f, map_location=device))
|
147 |
+
|
148 |
+
if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
|
149 |
+
raise ValueError(f"{files[0]} is not a zero checkpoint")
|
150 |
+
zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
|
151 |
+
world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
|
152 |
+
|
153 |
+
# For ZeRO-2 each param group can have different partition_count as data parallelism for expert
|
154 |
+
# parameters can be different from data parallelism for non-expert parameters. So we can just
|
155 |
+
# use the max of the partition_count to get the dp world_size.
|
156 |
+
|
157 |
+
if type(world_size) is list:
|
158 |
+
world_size = max(world_size)
|
159 |
+
|
160 |
+
if world_size != total_files:
|
161 |
+
raise ValueError(
|
162 |
+
f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
|
163 |
+
"Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
|
164 |
+
)
|
165 |
+
|
166 |
+
# the groups are named differently in each stage
|
167 |
+
if zero_stage == 2:
|
168 |
+
fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
|
169 |
+
elif zero_stage == 3:
|
170 |
+
fp32_groups_key = FP32_FLAT_GROUPS
|
171 |
+
else:
|
172 |
+
raise ValueError(f"unknown zero stage {zero_stage}")
|
173 |
+
|
174 |
+
if zero_stage == 2:
|
175 |
+
fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
|
176 |
+
elif zero_stage == 3:
|
177 |
+
# if there is more than one param group, there will be multiple flattened tensors - one
|
178 |
+
# flattened tensor per group - for simplicity merge them into a single tensor
|
179 |
+
#
|
180 |
+
# XXX: could make the script more memory efficient for when there are multiple groups - it
|
181 |
+
# will require matching the sub-lists of param_shapes for each param group flattened tensor
|
182 |
+
|
183 |
+
fp32_flat_groups = [
|
184 |
+
torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
|
185 |
+
]
|
186 |
+
|
187 |
+
return zero_stage, world_size, fp32_flat_groups
|
188 |
+
|
189 |
+
|
190 |
+
def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir):
|
191 |
+
"""
|
192 |
+
Returns fp32 state_dict reconstructed from ds checkpoint
|
193 |
+
|
194 |
+
Args:
|
195 |
+
- ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
|
196 |
+
|
197 |
+
"""
|
198 |
+
print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
|
199 |
+
|
200 |
+
optim_files = get_optim_files(ds_checkpoint_dir)
|
201 |
+
zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
|
202 |
+
print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
|
203 |
+
|
204 |
+
model_files = get_model_state_files(ds_checkpoint_dir)
|
205 |
+
|
206 |
+
zero_model_states = parse_model_states(model_files)
|
207 |
+
print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
|
208 |
+
|
209 |
+
if zero_stage == 2:
|
210 |
+
return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states)
|
211 |
+
elif zero_stage == 3:
|
212 |
+
return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states)
|
213 |
+
|
214 |
+
|
215 |
+
def _zero2_merge_frozen_params(state_dict, zero_model_states):
|
216 |
+
if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
|
217 |
+
return
|
218 |
+
|
219 |
+
frozen_param_shapes = zero_model_states[0].frozen_param_shapes
|
220 |
+
frozen_param_fragments = zero_model_states[0].frozen_param_fragments
|
221 |
+
|
222 |
+
if debug:
|
223 |
+
num_elem = sum(s.numel() for s in frozen_param_shapes.values())
|
224 |
+
print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
|
225 |
+
|
226 |
+
wanted_params = len(frozen_param_shapes)
|
227 |
+
wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
|
228 |
+
avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
|
229 |
+
print(f'Frozen params: Have {avail_numel} numels to process.')
|
230 |
+
print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
|
231 |
+
|
232 |
+
total_params = 0
|
233 |
+
total_numel = 0
|
234 |
+
for name, shape in frozen_param_shapes.items():
|
235 |
+
total_params += 1
|
236 |
+
unpartitioned_numel = shape.numel()
|
237 |
+
total_numel += unpartitioned_numel
|
238 |
+
|
239 |
+
state_dict[name] = frozen_param_fragments[name]
|
240 |
+
|
241 |
+
if debug:
|
242 |
+
print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
|
243 |
+
|
244 |
+
print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
|
245 |
+
|
246 |
+
|
247 |
+
def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
|
248 |
+
param_shapes = zero_model_states[0].param_shapes
|
249 |
+
|
250 |
+
# Reconstruction protocol:
|
251 |
+
#
|
252 |
+
# XXX: document this
|
253 |
+
|
254 |
+
if debug:
|
255 |
+
for i in range(world_size):
|
256 |
+
for j in range(len(fp32_flat_groups[0])):
|
257 |
+
print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
|
258 |
+
|
259 |
+
# XXX: memory usage doubles here (zero2)
|
260 |
+
num_param_groups = len(fp32_flat_groups[0])
|
261 |
+
merged_single_partition_of_fp32_groups = []
|
262 |
+
for i in range(num_param_groups):
|
263 |
+
merged_partitions = [sd[i] for sd in fp32_flat_groups]
|
264 |
+
full_single_fp32_vector = torch.cat(merged_partitions, 0)
|
265 |
+
merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
|
266 |
+
avail_numel = sum(
|
267 |
+
[full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
|
268 |
+
|
269 |
+
if debug:
|
270 |
+
wanted_params = sum([len(shapes) for shapes in param_shapes])
|
271 |
+
wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
|
272 |
+
# not asserting if there is a mismatch due to possible padding
|
273 |
+
print(f"Have {avail_numel} numels to process.")
|
274 |
+
print(f"Need {wanted_numel} numels in {wanted_params} params.")
|
275 |
+
|
276 |
+
# params
|
277 |
+
# XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
|
278 |
+
# out-of-core computing solution
|
279 |
+
total_numel = 0
|
280 |
+
total_params = 0
|
281 |
+
for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
|
282 |
+
offset = 0
|
283 |
+
avail_numel = full_single_fp32_vector.numel()
|
284 |
+
for name, shape in shapes.items():
|
285 |
+
|
286 |
+
unpartitioned_numel = shape.numel()
|
287 |
+
total_numel += unpartitioned_numel
|
288 |
+
total_params += 1
|
289 |
+
|
290 |
+
if debug:
|
291 |
+
print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
|
292 |
+
state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
|
293 |
+
offset += unpartitioned_numel
|
294 |
+
|
295 |
+
# Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
|
296 |
+
# avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
|
297 |
+
# paddings performed in the code it's almost impossible to predict the exact numbers w/o the
|
298 |
+
# live optimizer object, so we are checking that the numbers are within the right range
|
299 |
+
align_to = 2 * world_size
|
300 |
+
|
301 |
+
def zero2_align(x):
|
302 |
+
return align_to * math.ceil(x / align_to)
|
303 |
+
|
304 |
+
if debug:
|
305 |
+
print(f"original offset={offset}, avail_numel={avail_numel}")
|
306 |
+
|
307 |
+
offset = zero2_align(offset)
|
308 |
+
avail_numel = zero2_align(avail_numel)
|
309 |
+
|
310 |
+
if debug:
|
311 |
+
print(f"aligned offset={offset}, avail_numel={avail_numel}")
|
312 |
+
|
313 |
+
# Sanity check
|
314 |
+
if offset != avail_numel:
|
315 |
+
raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
|
316 |
+
|
317 |
+
print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
|
318 |
+
|
319 |
+
|
320 |
+
def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states):
|
321 |
+
state_dict = OrderedDict()
|
322 |
+
|
323 |
+
# buffers
|
324 |
+
buffers = zero_model_states[0].buffers
|
325 |
+
state_dict.update(buffers)
|
326 |
+
if debug:
|
327 |
+
print(f"added {len(buffers)} buffers")
|
328 |
+
|
329 |
+
_zero2_merge_frozen_params(state_dict, zero_model_states)
|
330 |
+
|
331 |
+
_zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
|
332 |
+
|
333 |
+
# recover shared parameters
|
334 |
+
for pair in zero_model_states[0].shared_params:
|
335 |
+
if pair[1] in state_dict:
|
336 |
+
state_dict[pair[0]] = state_dict[pair[1]]
|
337 |
+
|
338 |
+
return state_dict
|
339 |
+
|
340 |
+
|
341 |
+
def zero3_partitioned_param_info(unpartitioned_numel, world_size):
|
342 |
+
remainder = unpartitioned_numel % world_size
|
343 |
+
padding_numel = (world_size - remainder) if remainder else 0
|
344 |
+
partitioned_numel = math.ceil(unpartitioned_numel / world_size)
|
345 |
+
return partitioned_numel, padding_numel
|
346 |
+
|
347 |
+
|
348 |
+
def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
|
349 |
+
if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
|
350 |
+
return
|
351 |
+
|
352 |
+
if debug:
|
353 |
+
for i in range(world_size):
|
354 |
+
num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
|
355 |
+
print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
|
356 |
+
|
357 |
+
frozen_param_shapes = zero_model_states[0].frozen_param_shapes
|
358 |
+
wanted_params = len(frozen_param_shapes)
|
359 |
+
wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
|
360 |
+
avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
|
361 |
+
print(f'Frozen params: Have {avail_numel} numels to process.')
|
362 |
+
print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
|
363 |
+
|
364 |
+
total_params = 0
|
365 |
+
total_numel = 0
|
366 |
+
for name, shape in zero_model_states[0].frozen_param_shapes.items():
|
367 |
+
total_params += 1
|
368 |
+
unpartitioned_numel = shape.numel()
|
369 |
+
total_numel += unpartitioned_numel
|
370 |
+
|
371 |
+
param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
|
372 |
+
state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
|
373 |
+
|
374 |
+
partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
|
375 |
+
|
376 |
+
if debug:
|
377 |
+
print(
|
378 |
+
f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
|
379 |
+
)
|
380 |
+
|
381 |
+
print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
|
382 |
+
|
383 |
+
|
384 |
+
def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
|
385 |
+
param_shapes = zero_model_states[0].param_shapes
|
386 |
+
avail_numel = fp32_flat_groups[0].numel() * world_size
|
387 |
+
# Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
|
388 |
+
# param, re-consolidating each param, while dealing with padding if any
|
389 |
+
|
390 |
+
# merge list of dicts, preserving order
|
391 |
+
param_shapes = {k: v for d in param_shapes for k, v in d.items()}
|
392 |
+
|
393 |
+
if debug:
|
394 |
+
for i in range(world_size):
|
395 |
+
print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
|
396 |
+
|
397 |
+
wanted_params = len(param_shapes)
|
398 |
+
wanted_numel = sum(shape.numel() for shape in param_shapes.values())
|
399 |
+
# not asserting if there is a mismatch due to possible padding
|
400 |
+
avail_numel = fp32_flat_groups[0].numel() * world_size
|
401 |
+
print(f"Trainable params: Have {avail_numel} numels to process.")
|
402 |
+
print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
|
403 |
+
|
404 |
+
# params
|
405 |
+
# XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
|
406 |
+
# out-of-core computing solution
|
407 |
+
offset = 0
|
408 |
+
total_numel = 0
|
409 |
+
total_params = 0
|
410 |
+
for name, shape in param_shapes.items():
|
411 |
+
|
412 |
+
unpartitioned_numel = shape.numel()
|
413 |
+
total_numel += unpartitioned_numel
|
414 |
+
total_params += 1
|
415 |
+
|
416 |
+
partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
|
417 |
+
|
418 |
+
if debug:
|
419 |
+
print(
|
420 |
+
f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
|
421 |
+
)
|
422 |
+
|
423 |
+
# XXX: memory usage doubles here
|
424 |
+
state_dict[name] = torch.cat(
|
425 |
+
tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
|
426 |
+
0).narrow(0, 0, unpartitioned_numel).view(shape)
|
427 |
+
offset += partitioned_numel
|
428 |
+
|
429 |
+
offset *= world_size
|
430 |
+
|
431 |
+
# Sanity check
|
432 |
+
if offset != avail_numel:
|
433 |
+
raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
|
434 |
+
|
435 |
+
print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
|
436 |
+
|
437 |
+
|
438 |
+
def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states):
|
439 |
+
state_dict = OrderedDict()
|
440 |
+
|
441 |
+
# buffers
|
442 |
+
buffers = zero_model_states[0].buffers
|
443 |
+
state_dict.update(buffers)
|
444 |
+
if debug:
|
445 |
+
print(f"added {len(buffers)} buffers")
|
446 |
+
|
447 |
+
_zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
|
448 |
+
|
449 |
+
_zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
|
450 |
+
|
451 |
+
# recover shared parameters
|
452 |
+
for pair in zero_model_states[0].shared_params:
|
453 |
+
if pair[1] in state_dict:
|
454 |
+
state_dict[pair[0]] = state_dict[pair[1]]
|
455 |
+
|
456 |
+
return state_dict
|
457 |
+
|
458 |
+
|
459 |
+
def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None):
|
460 |
+
"""
|
461 |
+
Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
|
462 |
+
``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
|
463 |
+
via a model hub.
|
464 |
+
|
465 |
+
Args:
|
466 |
+
- ``checkpoint_dir``: path to the desired checkpoint folder
|
467 |
+
- ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
|
468 |
+
|
469 |
+
Returns:
|
470 |
+
- pytorch ``state_dict``
|
471 |
+
|
472 |
+
Note: this approach may not work if your application doesn't have sufficient free CPU memory and
|
473 |
+
you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
|
474 |
+
the checkpoint.
|
475 |
+
|
476 |
+
A typical usage might be ::
|
477 |
+
|
478 |
+
from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
|
479 |
+
# do the training and checkpoint saving
|
480 |
+
state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
|
481 |
+
model = model.cpu() # move to cpu
|
482 |
+
model.load_state_dict(state_dict)
|
483 |
+
# submit to model hub or save the model to share with others
|
484 |
+
|
485 |
+
In this example the ``model`` will no longer be usable in the deepspeed context of the same
|
486 |
+
application. i.e. you will need to re-initialize the deepspeed engine, since
|
487 |
+
``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
|
488 |
+
|
489 |
+
If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
|
490 |
+
|
491 |
+
"""
|
492 |
+
if tag is None:
|
493 |
+
latest_path = os.path.join(checkpoint_dir, 'latest')
|
494 |
+
if os.path.isfile(latest_path):
|
495 |
+
with open(latest_path, 'r') as fd:
|
496 |
+
tag = fd.read().strip()
|
497 |
+
else:
|
498 |
+
raise ValueError(f"Unable to find 'latest' file at {latest_path}")
|
499 |
+
|
500 |
+
ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
|
501 |
+
|
502 |
+
if not os.path.isdir(ds_checkpoint_dir):
|
503 |
+
raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
|
504 |
+
|
505 |
+
return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir)
|
506 |
+
|
507 |
+
|
508 |
+
def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None):
|
509 |
+
"""
|
510 |
+
Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
|
511 |
+
loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
|
512 |
+
|
513 |
+
Args:
|
514 |
+
- ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
|
515 |
+
- ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
|
516 |
+
- ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
|
517 |
+
"""
|
518 |
+
|
519 |
+
state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
|
520 |
+
print(f"Saving fp32 state dict to {output_file}")
|
521 |
+
torch.save(state_dict, output_file)
|
522 |
+
|
523 |
+
|
524 |
+
def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
|
525 |
+
"""
|
526 |
+
1. Put the provided model to cpu
|
527 |
+
2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
|
528 |
+
3. Load it into the provided model
|
529 |
+
|
530 |
+
Args:
|
531 |
+
- ``model``: the model object to update
|
532 |
+
- ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
|
533 |
+
- ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
|
534 |
+
|
535 |
+
Returns:
|
536 |
+
- ``model`: modified model
|
537 |
+
|
538 |
+
Make sure you have plenty of CPU memory available before you call this function. If you don't
|
539 |
+
have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
|
540 |
+
conveniently placed for you in the checkpoint folder.
|
541 |
+
|
542 |
+
A typical usage might be ::
|
543 |
+
|
544 |
+
from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
|
545 |
+
model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
|
546 |
+
# submit to model hub or save the model to share with others
|
547 |
+
|
548 |
+
Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
|
549 |
+
of the same application. i.e. you will need to re-initialize the deepspeed engine, since
|
550 |
+
``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
|
551 |
+
|
552 |
+
"""
|
553 |
+
logger.info(f"Extracting fp32 weights")
|
554 |
+
state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
|
555 |
+
|
556 |
+
logger.info(f"Overwriting model with fp32 weights")
|
557 |
+
model = model.cpu()
|
558 |
+
model.load_state_dict(state_dict, strict=False)
|
559 |
+
|
560 |
+
return model
|
561 |
+
|
562 |
+
|
563 |
+
if __name__ == "__main__":
|
564 |
+
|
565 |
+
parser = argparse.ArgumentParser()
|
566 |
+
parser.add_argument("checkpoint_dir",
|
567 |
+
type=str,
|
568 |
+
help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
|
569 |
+
parser.add_argument(
|
570 |
+
"output_file",
|
571 |
+
type=str,
|
572 |
+
help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
|
573 |
+
parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
|
574 |
+
args = parser.parse_args()
|
575 |
+
|
576 |
+
debug = args.debug
|
577 |
+
|
578 |
+
convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir, args.output_file)
|