mwitiderrick
commited on
Commit
•
2175f24
1
Parent(s):
cd6039f
Upload folder using huggingface_hub
Browse filesThis view is limited to 50 files because it contains too many changes.
See raw diff
- all_results.json +8 -0
- checkpoint-1000/config.json +29 -0
- checkpoint-1000/generation_config.json +7 -0
- checkpoint-1000/optimizer.pt +3 -0
- checkpoint-1000/pytorch_model.bin +3 -0
- checkpoint-1000/recipe.yaml +104 -0
- checkpoint-1000/rng_state.pth +3 -0
- checkpoint-1000/scheduler.pt +3 -0
- checkpoint-1000/special_tokens_map.json +30 -0
- checkpoint-1000/tokenizer.json +0 -0
- checkpoint-1000/tokenizer.model +3 -0
- checkpoint-1000/tokenizer_config.json +40 -0
- checkpoint-1000/trainer_state.json +31 -0
- checkpoint-1000/training_args.bin +3 -0
- checkpoint-1500/config.json +29 -0
- checkpoint-1500/generation_config.json +7 -0
- checkpoint-1500/optimizer.pt +3 -0
- checkpoint-1500/pytorch_model.bin +3 -0
- checkpoint-1500/recipe.yaml +104 -0
- checkpoint-1500/rng_state.pth +3 -0
- checkpoint-1500/scheduler.pt +3 -0
- checkpoint-1500/special_tokens_map.json +30 -0
- checkpoint-1500/tokenizer.json +0 -0
- checkpoint-1500/tokenizer.model +3 -0
- checkpoint-1500/tokenizer_config.json +40 -0
- checkpoint-1500/trainer_state.json +37 -0
- checkpoint-1500/training_args.bin +3 -0
- checkpoint-2000/config.json +29 -0
- checkpoint-2000/generation_config.json +7 -0
- checkpoint-2000/optimizer.pt +3 -0
- checkpoint-2000/pytorch_model.bin +3 -0
- checkpoint-2000/recipe.yaml +104 -0
- checkpoint-2000/rng_state.pth +3 -0
- checkpoint-2000/scheduler.pt +3 -0
- checkpoint-2000/special_tokens_map.json +30 -0
- checkpoint-2000/tokenizer.json +0 -0
- checkpoint-2000/tokenizer.model +3 -0
- checkpoint-2000/tokenizer_config.json +40 -0
- checkpoint-2000/trainer_state.json +43 -0
- checkpoint-2000/training_args.bin +3 -0
- checkpoint-2500/config.json +29 -0
- checkpoint-2500/generation_config.json +7 -0
- checkpoint-2500/optimizer.pt +3 -0
- checkpoint-2500/pytorch_model.bin +3 -0
- checkpoint-2500/recipe.yaml +104 -0
- checkpoint-2500/rng_state.pth +3 -0
- checkpoint-2500/scheduler.pt +3 -0
- checkpoint-2500/special_tokens_map.json +30 -0
- checkpoint-2500/tokenizer.json +0 -0
- checkpoint-2500/tokenizer.model +3 -0
all_results.json
ADDED
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 2.0,
|
3 |
+
"train_loss": 31291.547015404365,
|
4 |
+
"train_runtime": 1788.4277,
|
5 |
+
"train_samples": 12463,
|
6 |
+
"train_samples_per_second": 13.937,
|
7 |
+
"train_steps_per_second": 1.742
|
8 |
+
}
|
checkpoint-1000/config.json
ADDED
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_name_or_path": "./obcq_deployment",
|
3 |
+
"architectures": [
|
4 |
+
"LlamaForCausalLM"
|
5 |
+
],
|
6 |
+
"attention_bias": false,
|
7 |
+
"bos_token_id": 1,
|
8 |
+
"eos_token_id": 2,
|
9 |
+
"hidden_act": "silu",
|
10 |
+
"hidden_size": 2048,
|
11 |
+
"initializer_range": 0.02,
|
12 |
+
"intermediate_size": 5632,
|
13 |
+
"is_decoder": true,
|
14 |
+
"max_position_embeddings": 2048,
|
15 |
+
"model_type": "llama",
|
16 |
+
"num_attention_heads": 32,
|
17 |
+
"num_hidden_layers": 22,
|
18 |
+
"num_key_value_heads": 4,
|
19 |
+
"pretraining_tp": 1,
|
20 |
+
"rms_norm_eps": 1e-05,
|
21 |
+
"rope_scaling": null,
|
22 |
+
"rope_theta": 10000.0,
|
23 |
+
"tie_word_embeddings": false,
|
24 |
+
"torch_dtype": "float16",
|
25 |
+
"transformers_version": "1.7.0.20240122",
|
26 |
+
"use_cache": false,
|
27 |
+
"use_past": false,
|
28 |
+
"vocab_size": 32000
|
29 |
+
}
|
checkpoint-1000/generation_config.json
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"bos_token_id": 1,
|
3 |
+
"eos_token_id": 2,
|
4 |
+
"max_length": 2048,
|
5 |
+
"pad_token_id": 0,
|
6 |
+
"transformers_version": "1.7.0.20240122"
|
7 |
+
}
|
checkpoint-1000/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:46dc44b489506d6d651b0c56aaa95da647b83d92187c5df3f9a1c77332b881fa
|
3 |
+
size 4400363238
|
checkpoint-1000/pytorch_model.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:ed6df076e5b420888f4eb6c7bfd39c01e7ca41f0404f945ff69d86071774fffd
|
3 |
+
size 2207873726
|
checkpoint-1000/recipe.yaml
ADDED
@@ -0,0 +1,104 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
test_stage_0:
|
2 |
+
obcq_modifiers:
|
3 |
+
LogarithmicEqualizationModifier:
|
4 |
+
mappings:
|
5 |
+
- - - re:.*q_proj
|
6 |
+
- re:.*k_proj
|
7 |
+
- re:.*v_proj
|
8 |
+
- re:.*input_layernorm
|
9 |
+
- - - re:.*gate_proj
|
10 |
+
- re:.*up_proj
|
11 |
+
- re:.*post_attention_layernorm
|
12 |
+
QuantizationModifier:
|
13 |
+
ignore:
|
14 |
+
- LlamaRotaryEmbedding
|
15 |
+
- LlamaRMSNorm
|
16 |
+
- SiLUActivation
|
17 |
+
- MatMulOutput_QK
|
18 |
+
- MatMulOutput_PV
|
19 |
+
- model.layers.21.mlp.down_proj
|
20 |
+
- model.layers.7.mlp.down_proj
|
21 |
+
- model.layers.2.mlp.down_proj
|
22 |
+
- model.layers.8.self_attn.q_proj
|
23 |
+
- model.layers.8.self_attn.k_proj
|
24 |
+
post_oneshot_calibration: true
|
25 |
+
scheme_overrides:
|
26 |
+
Linear:
|
27 |
+
weights:
|
28 |
+
num_bits: 8
|
29 |
+
symmetric: true
|
30 |
+
strategy: channel
|
31 |
+
MatMulLeftInput_QK:
|
32 |
+
input_activations:
|
33 |
+
num_bits: 8
|
34 |
+
symmetric: true
|
35 |
+
MatMulLeftInput_PV:
|
36 |
+
input_activations:
|
37 |
+
num_bits: 8
|
38 |
+
symmetric: true
|
39 |
+
Embedding:
|
40 |
+
input_activations: null
|
41 |
+
weights:
|
42 |
+
num_bits: 8
|
43 |
+
symmetric: false
|
44 |
+
SparseGPTModifier:
|
45 |
+
sparsity: 0.5
|
46 |
+
block_size: 128
|
47 |
+
sequential_update: true
|
48 |
+
quantize: true
|
49 |
+
percdamp: 0.01
|
50 |
+
mask_structure: 0:0
|
51 |
+
targets:
|
52 |
+
- re:model.layers.\d*$
|
53 |
+
test_stage_1:
|
54 |
+
obcq_modifiers:
|
55 |
+
LogarithmicEqualizationModifier:
|
56 |
+
mappings:
|
57 |
+
- - - re:.*q_proj
|
58 |
+
- re:.*k_proj
|
59 |
+
- re:.*v_proj
|
60 |
+
- re:.*input_layernorm
|
61 |
+
- - - re:.*gate_proj
|
62 |
+
- re:.*up_proj
|
63 |
+
- re:.*post_attention_layernorm
|
64 |
+
QuantizationModifier:
|
65 |
+
ignore:
|
66 |
+
- LlamaRotaryEmbedding
|
67 |
+
- LlamaRMSNorm
|
68 |
+
- SiLUActivation
|
69 |
+
- MatMulOutput_QK
|
70 |
+
- MatMulOutput_PV
|
71 |
+
- model.layers.21.mlp.down_proj
|
72 |
+
- model.layers.7.mlp.down_proj
|
73 |
+
- model.layers.2.mlp.down_proj
|
74 |
+
- model.layers.8.self_attn.q_proj
|
75 |
+
- model.layers.8.self_attn.k_proj
|
76 |
+
post_oneshot_calibration: true
|
77 |
+
scheme_overrides:
|
78 |
+
Linear:
|
79 |
+
weights:
|
80 |
+
num_bits: 8
|
81 |
+
symmetric: true
|
82 |
+
strategy: channel
|
83 |
+
MatMulLeftInput_QK:
|
84 |
+
input_activations:
|
85 |
+
num_bits: 8
|
86 |
+
symmetric: true
|
87 |
+
MatMulLeftInput_PV:
|
88 |
+
input_activations:
|
89 |
+
num_bits: 8
|
90 |
+
symmetric: true
|
91 |
+
Embedding:
|
92 |
+
input_activations: null
|
93 |
+
weights:
|
94 |
+
num_bits: 8
|
95 |
+
symmetric: false
|
96 |
+
SparseGPTModifier:
|
97 |
+
sparsity: 0.5
|
98 |
+
block_size: 128
|
99 |
+
sequential_update: true
|
100 |
+
quantize: true
|
101 |
+
percdamp: 0.01
|
102 |
+
mask_structure: 0:0
|
103 |
+
targets:
|
104 |
+
- re:model.layers.\d*$
|
checkpoint-1000/rng_state.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:386fcc8cc1089aade9450d86fb239ea3483f455fd2d78d8378645feecfec9d69
|
3 |
+
size 14244
|
checkpoint-1000/scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c34ba9e6f50c2313df84e91c7dfed2be6b77dec7394adba34d3bd12d40d5ba2d
|
3 |
+
size 1064
|
checkpoint-1000/special_tokens_map.json
ADDED
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"bos_token": {
|
3 |
+
"content": "<s>",
|
4 |
+
"lstrip": false,
|
5 |
+
"normalized": false,
|
6 |
+
"rstrip": false,
|
7 |
+
"single_word": false
|
8 |
+
},
|
9 |
+
"eos_token": {
|
10 |
+
"content": "</s>",
|
11 |
+
"lstrip": false,
|
12 |
+
"normalized": false,
|
13 |
+
"rstrip": false,
|
14 |
+
"single_word": false
|
15 |
+
},
|
16 |
+
"pad_token": {
|
17 |
+
"content": "</s>",
|
18 |
+
"lstrip": false,
|
19 |
+
"normalized": false,
|
20 |
+
"rstrip": false,
|
21 |
+
"single_word": false
|
22 |
+
},
|
23 |
+
"unk_token": {
|
24 |
+
"content": "<unk>",
|
25 |
+
"lstrip": false,
|
26 |
+
"normalized": false,
|
27 |
+
"rstrip": false,
|
28 |
+
"single_word": false
|
29 |
+
}
|
30 |
+
}
|
checkpoint-1000/tokenizer.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-1000/tokenizer.model
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
|
3 |
+
size 499723
|
checkpoint-1000/tokenizer_config.json
ADDED
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"added_tokens_decoder": {
|
3 |
+
"0": {
|
4 |
+
"content": "<unk>",
|
5 |
+
"lstrip": false,
|
6 |
+
"normalized": false,
|
7 |
+
"rstrip": false,
|
8 |
+
"single_word": false,
|
9 |
+
"special": true
|
10 |
+
},
|
11 |
+
"1": {
|
12 |
+
"content": "<s>",
|
13 |
+
"lstrip": false,
|
14 |
+
"normalized": false,
|
15 |
+
"rstrip": false,
|
16 |
+
"single_word": false,
|
17 |
+
"special": true
|
18 |
+
},
|
19 |
+
"2": {
|
20 |
+
"content": "</s>",
|
21 |
+
"lstrip": false,
|
22 |
+
"normalized": false,
|
23 |
+
"rstrip": false,
|
24 |
+
"single_word": false,
|
25 |
+
"special": true
|
26 |
+
}
|
27 |
+
},
|
28 |
+
"bos_token": "<s>",
|
29 |
+
"chat_template": "{% for message in messages %}\n{% if message['role'] == 'user' %}\n{{ '<|user|>\n' + message['content'] + eos_token }}\n{% elif message['role'] == 'system' %}\n{{ '<|system|>\n' + message['content'] + eos_token }}\n{% elif message['role'] == 'assistant' %}\n{{ '<|assistant|>\n' + message['content'] + eos_token }}\n{% endif %}\n{% if loop.last and add_generation_prompt %}\n{{ '<|assistant|>' }}\n{% endif %}\n{% endfor %}",
|
30 |
+
"clean_up_tokenization_spaces": false,
|
31 |
+
"eos_token": "</s>",
|
32 |
+
"legacy": false,
|
33 |
+
"model_max_length": 2048,
|
34 |
+
"pad_token": "</s>",
|
35 |
+
"padding_side": "right",
|
36 |
+
"sp_model_kwargs": {},
|
37 |
+
"tokenizer_class": "LlamaTokenizer",
|
38 |
+
"unk_token": "<unk>",
|
39 |
+
"use_default_system_prompt": false
|
40 |
+
}
|
checkpoint-1000/trainer_state.json
ADDED
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": null,
|
3 |
+
"best_model_checkpoint": null,
|
4 |
+
"epoch": 0.6418485237483954,
|
5 |
+
"eval_steps": 500,
|
6 |
+
"global_step": 1000,
|
7 |
+
"is_hyper_param_search": false,
|
8 |
+
"is_local_process_zero": true,
|
9 |
+
"is_world_process_zero": true,
|
10 |
+
"log_history": [
|
11 |
+
{
|
12 |
+
"epoch": 0.32,
|
13 |
+
"learning_rate": 5e-05,
|
14 |
+
"loss": 31251.434,
|
15 |
+
"step": 500
|
16 |
+
},
|
17 |
+
{
|
18 |
+
"epoch": 0.64,
|
19 |
+
"learning_rate": 5e-05,
|
20 |
+
"loss": 31304.524,
|
21 |
+
"step": 1000
|
22 |
+
}
|
23 |
+
],
|
24 |
+
"logging_steps": 500,
|
25 |
+
"max_steps": 3116,
|
26 |
+
"num_train_epochs": 2,
|
27 |
+
"save_steps": 500,
|
28 |
+
"total_flos": 1.9068132261888e+16,
|
29 |
+
"trial_name": null,
|
30 |
+
"trial_params": null
|
31 |
+
}
|
checkpoint-1000/training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d259bb3b980ce004ad43f80c2d7f127a6d085a805ca671c63ea8e38660b643bb
|
3 |
+
size 4728
|
checkpoint-1500/config.json
ADDED
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_name_or_path": "./obcq_deployment",
|
3 |
+
"architectures": [
|
4 |
+
"LlamaForCausalLM"
|
5 |
+
],
|
6 |
+
"attention_bias": false,
|
7 |
+
"bos_token_id": 1,
|
8 |
+
"eos_token_id": 2,
|
9 |
+
"hidden_act": "silu",
|
10 |
+
"hidden_size": 2048,
|
11 |
+
"initializer_range": 0.02,
|
12 |
+
"intermediate_size": 5632,
|
13 |
+
"is_decoder": true,
|
14 |
+
"max_position_embeddings": 2048,
|
15 |
+
"model_type": "llama",
|
16 |
+
"num_attention_heads": 32,
|
17 |
+
"num_hidden_layers": 22,
|
18 |
+
"num_key_value_heads": 4,
|
19 |
+
"pretraining_tp": 1,
|
20 |
+
"rms_norm_eps": 1e-05,
|
21 |
+
"rope_scaling": null,
|
22 |
+
"rope_theta": 10000.0,
|
23 |
+
"tie_word_embeddings": false,
|
24 |
+
"torch_dtype": "float16",
|
25 |
+
"transformers_version": "1.7.0.20240122",
|
26 |
+
"use_cache": false,
|
27 |
+
"use_past": false,
|
28 |
+
"vocab_size": 32000
|
29 |
+
}
|
checkpoint-1500/generation_config.json
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"bos_token_id": 1,
|
3 |
+
"eos_token_id": 2,
|
4 |
+
"max_length": 2048,
|
5 |
+
"pad_token_id": 0,
|
6 |
+
"transformers_version": "1.7.0.20240122"
|
7 |
+
}
|
checkpoint-1500/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:1a11d5c1f1f864e472301a75cf017aa749805955847d0b1e5a5a20a2d7e2a773
|
3 |
+
size 4400363238
|
checkpoint-1500/pytorch_model.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:ed6df076e5b420888f4eb6c7bfd39c01e7ca41f0404f945ff69d86071774fffd
|
3 |
+
size 2207873726
|
checkpoint-1500/recipe.yaml
ADDED
@@ -0,0 +1,104 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
test_stage_0:
|
2 |
+
obcq_modifiers:
|
3 |
+
LogarithmicEqualizationModifier:
|
4 |
+
mappings:
|
5 |
+
- - - re:.*q_proj
|
6 |
+
- re:.*k_proj
|
7 |
+
- re:.*v_proj
|
8 |
+
- re:.*input_layernorm
|
9 |
+
- - - re:.*gate_proj
|
10 |
+
- re:.*up_proj
|
11 |
+
- re:.*post_attention_layernorm
|
12 |
+
QuantizationModifier:
|
13 |
+
ignore:
|
14 |
+
- LlamaRotaryEmbedding
|
15 |
+
- LlamaRMSNorm
|
16 |
+
- SiLUActivation
|
17 |
+
- MatMulOutput_QK
|
18 |
+
- MatMulOutput_PV
|
19 |
+
- model.layers.21.mlp.down_proj
|
20 |
+
- model.layers.7.mlp.down_proj
|
21 |
+
- model.layers.2.mlp.down_proj
|
22 |
+
- model.layers.8.self_attn.q_proj
|
23 |
+
- model.layers.8.self_attn.k_proj
|
24 |
+
post_oneshot_calibration: true
|
25 |
+
scheme_overrides:
|
26 |
+
Linear:
|
27 |
+
weights:
|
28 |
+
num_bits: 8
|
29 |
+
symmetric: true
|
30 |
+
strategy: channel
|
31 |
+
MatMulLeftInput_QK:
|
32 |
+
input_activations:
|
33 |
+
num_bits: 8
|
34 |
+
symmetric: true
|
35 |
+
MatMulLeftInput_PV:
|
36 |
+
input_activations:
|
37 |
+
num_bits: 8
|
38 |
+
symmetric: true
|
39 |
+
Embedding:
|
40 |
+
input_activations: null
|
41 |
+
weights:
|
42 |
+
num_bits: 8
|
43 |
+
symmetric: false
|
44 |
+
SparseGPTModifier:
|
45 |
+
sparsity: 0.5
|
46 |
+
block_size: 128
|
47 |
+
sequential_update: true
|
48 |
+
quantize: true
|
49 |
+
percdamp: 0.01
|
50 |
+
mask_structure: 0:0
|
51 |
+
targets:
|
52 |
+
- re:model.layers.\d*$
|
53 |
+
test_stage_1:
|
54 |
+
obcq_modifiers:
|
55 |
+
LogarithmicEqualizationModifier:
|
56 |
+
mappings:
|
57 |
+
- - - re:.*q_proj
|
58 |
+
- re:.*k_proj
|
59 |
+
- re:.*v_proj
|
60 |
+
- re:.*input_layernorm
|
61 |
+
- - - re:.*gate_proj
|
62 |
+
- re:.*up_proj
|
63 |
+
- re:.*post_attention_layernorm
|
64 |
+
QuantizationModifier:
|
65 |
+
ignore:
|
66 |
+
- LlamaRotaryEmbedding
|
67 |
+
- LlamaRMSNorm
|
68 |
+
- SiLUActivation
|
69 |
+
- MatMulOutput_QK
|
70 |
+
- MatMulOutput_PV
|
71 |
+
- model.layers.21.mlp.down_proj
|
72 |
+
- model.layers.7.mlp.down_proj
|
73 |
+
- model.layers.2.mlp.down_proj
|
74 |
+
- model.layers.8.self_attn.q_proj
|
75 |
+
- model.layers.8.self_attn.k_proj
|
76 |
+
post_oneshot_calibration: true
|
77 |
+
scheme_overrides:
|
78 |
+
Linear:
|
79 |
+
weights:
|
80 |
+
num_bits: 8
|
81 |
+
symmetric: true
|
82 |
+
strategy: channel
|
83 |
+
MatMulLeftInput_QK:
|
84 |
+
input_activations:
|
85 |
+
num_bits: 8
|
86 |
+
symmetric: true
|
87 |
+
MatMulLeftInput_PV:
|
88 |
+
input_activations:
|
89 |
+
num_bits: 8
|
90 |
+
symmetric: true
|
91 |
+
Embedding:
|
92 |
+
input_activations: null
|
93 |
+
weights:
|
94 |
+
num_bits: 8
|
95 |
+
symmetric: false
|
96 |
+
SparseGPTModifier:
|
97 |
+
sparsity: 0.5
|
98 |
+
block_size: 128
|
99 |
+
sequential_update: true
|
100 |
+
quantize: true
|
101 |
+
percdamp: 0.01
|
102 |
+
mask_structure: 0:0
|
103 |
+
targets:
|
104 |
+
- re:model.layers.\d*$
|
checkpoint-1500/rng_state.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:386fcc8cc1089aade9450d86fb239ea3483f455fd2d78d8378645feecfec9d69
|
3 |
+
size 14244
|
checkpoint-1500/scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:084c8aecb9d3ac85c94f614df2060324a90d5a59efa400ab27a1733b0fc8abfb
|
3 |
+
size 1064
|
checkpoint-1500/special_tokens_map.json
ADDED
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"bos_token": {
|
3 |
+
"content": "<s>",
|
4 |
+
"lstrip": false,
|
5 |
+
"normalized": false,
|
6 |
+
"rstrip": false,
|
7 |
+
"single_word": false
|
8 |
+
},
|
9 |
+
"eos_token": {
|
10 |
+
"content": "</s>",
|
11 |
+
"lstrip": false,
|
12 |
+
"normalized": false,
|
13 |
+
"rstrip": false,
|
14 |
+
"single_word": false
|
15 |
+
},
|
16 |
+
"pad_token": {
|
17 |
+
"content": "</s>",
|
18 |
+
"lstrip": false,
|
19 |
+
"normalized": false,
|
20 |
+
"rstrip": false,
|
21 |
+
"single_word": false
|
22 |
+
},
|
23 |
+
"unk_token": {
|
24 |
+
"content": "<unk>",
|
25 |
+
"lstrip": false,
|
26 |
+
"normalized": false,
|
27 |
+
"rstrip": false,
|
28 |
+
"single_word": false
|
29 |
+
}
|
30 |
+
}
|
checkpoint-1500/tokenizer.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-1500/tokenizer.model
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
|
3 |
+
size 499723
|
checkpoint-1500/tokenizer_config.json
ADDED
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"added_tokens_decoder": {
|
3 |
+
"0": {
|
4 |
+
"content": "<unk>",
|
5 |
+
"lstrip": false,
|
6 |
+
"normalized": false,
|
7 |
+
"rstrip": false,
|
8 |
+
"single_word": false,
|
9 |
+
"special": true
|
10 |
+
},
|
11 |
+
"1": {
|
12 |
+
"content": "<s>",
|
13 |
+
"lstrip": false,
|
14 |
+
"normalized": false,
|
15 |
+
"rstrip": false,
|
16 |
+
"single_word": false,
|
17 |
+
"special": true
|
18 |
+
},
|
19 |
+
"2": {
|
20 |
+
"content": "</s>",
|
21 |
+
"lstrip": false,
|
22 |
+
"normalized": false,
|
23 |
+
"rstrip": false,
|
24 |
+
"single_word": false,
|
25 |
+
"special": true
|
26 |
+
}
|
27 |
+
},
|
28 |
+
"bos_token": "<s>",
|
29 |
+
"chat_template": "{% for message in messages %}\n{% if message['role'] == 'user' %}\n{{ '<|user|>\n' + message['content'] + eos_token }}\n{% elif message['role'] == 'system' %}\n{{ '<|system|>\n' + message['content'] + eos_token }}\n{% elif message['role'] == 'assistant' %}\n{{ '<|assistant|>\n' + message['content'] + eos_token }}\n{% endif %}\n{% if loop.last and add_generation_prompt %}\n{{ '<|assistant|>' }}\n{% endif %}\n{% endfor %}",
|
30 |
+
"clean_up_tokenization_spaces": false,
|
31 |
+
"eos_token": "</s>",
|
32 |
+
"legacy": false,
|
33 |
+
"model_max_length": 2048,
|
34 |
+
"pad_token": "</s>",
|
35 |
+
"padding_side": "right",
|
36 |
+
"sp_model_kwargs": {},
|
37 |
+
"tokenizer_class": "LlamaTokenizer",
|
38 |
+
"unk_token": "<unk>",
|
39 |
+
"use_default_system_prompt": false
|
40 |
+
}
|
checkpoint-1500/trainer_state.json
ADDED
@@ -0,0 +1,37 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": null,
|
3 |
+
"best_model_checkpoint": null,
|
4 |
+
"epoch": 0.962772785622593,
|
5 |
+
"eval_steps": 500,
|
6 |
+
"global_step": 1500,
|
7 |
+
"is_hyper_param_search": false,
|
8 |
+
"is_local_process_zero": true,
|
9 |
+
"is_world_process_zero": true,
|
10 |
+
"log_history": [
|
11 |
+
{
|
12 |
+
"epoch": 0.32,
|
13 |
+
"learning_rate": 5e-05,
|
14 |
+
"loss": 31251.434,
|
15 |
+
"step": 500
|
16 |
+
},
|
17 |
+
{
|
18 |
+
"epoch": 0.64,
|
19 |
+
"learning_rate": 5e-05,
|
20 |
+
"loss": 31304.524,
|
21 |
+
"step": 1000
|
22 |
+
},
|
23 |
+
{
|
24 |
+
"epoch": 0.96,
|
25 |
+
"learning_rate": 5e-05,
|
26 |
+
"loss": 31288.94,
|
27 |
+
"step": 1500
|
28 |
+
}
|
29 |
+
],
|
30 |
+
"logging_steps": 500,
|
31 |
+
"max_steps": 3116,
|
32 |
+
"num_train_epochs": 2,
|
33 |
+
"save_steps": 500,
|
34 |
+
"total_flos": 2.8602198392832e+16,
|
35 |
+
"trial_name": null,
|
36 |
+
"trial_params": null
|
37 |
+
}
|
checkpoint-1500/training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d259bb3b980ce004ad43f80c2d7f127a6d085a805ca671c63ea8e38660b643bb
|
3 |
+
size 4728
|
checkpoint-2000/config.json
ADDED
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_name_or_path": "./obcq_deployment",
|
3 |
+
"architectures": [
|
4 |
+
"LlamaForCausalLM"
|
5 |
+
],
|
6 |
+
"attention_bias": false,
|
7 |
+
"bos_token_id": 1,
|
8 |
+
"eos_token_id": 2,
|
9 |
+
"hidden_act": "silu",
|
10 |
+
"hidden_size": 2048,
|
11 |
+
"initializer_range": 0.02,
|
12 |
+
"intermediate_size": 5632,
|
13 |
+
"is_decoder": true,
|
14 |
+
"max_position_embeddings": 2048,
|
15 |
+
"model_type": "llama",
|
16 |
+
"num_attention_heads": 32,
|
17 |
+
"num_hidden_layers": 22,
|
18 |
+
"num_key_value_heads": 4,
|
19 |
+
"pretraining_tp": 1,
|
20 |
+
"rms_norm_eps": 1e-05,
|
21 |
+
"rope_scaling": null,
|
22 |
+
"rope_theta": 10000.0,
|
23 |
+
"tie_word_embeddings": false,
|
24 |
+
"torch_dtype": "float16",
|
25 |
+
"transformers_version": "1.7.0.20240122",
|
26 |
+
"use_cache": false,
|
27 |
+
"use_past": false,
|
28 |
+
"vocab_size": 32000
|
29 |
+
}
|
checkpoint-2000/generation_config.json
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"bos_token_id": 1,
|
3 |
+
"eos_token_id": 2,
|
4 |
+
"max_length": 2048,
|
5 |
+
"pad_token_id": 0,
|
6 |
+
"transformers_version": "1.7.0.20240122"
|
7 |
+
}
|
checkpoint-2000/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f57ffa4ef5e2ac9e25aa77eebf2bfdf5471d5a27603568403fbda96e4b2bdcfd
|
3 |
+
size 4400363238
|
checkpoint-2000/pytorch_model.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:ed6df076e5b420888f4eb6c7bfd39c01e7ca41f0404f945ff69d86071774fffd
|
3 |
+
size 2207873726
|
checkpoint-2000/recipe.yaml
ADDED
@@ -0,0 +1,104 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
test_stage_0:
|
2 |
+
obcq_modifiers:
|
3 |
+
LogarithmicEqualizationModifier:
|
4 |
+
mappings:
|
5 |
+
- - - re:.*q_proj
|
6 |
+
- re:.*k_proj
|
7 |
+
- re:.*v_proj
|
8 |
+
- re:.*input_layernorm
|
9 |
+
- - - re:.*gate_proj
|
10 |
+
- re:.*up_proj
|
11 |
+
- re:.*post_attention_layernorm
|
12 |
+
QuantizationModifier:
|
13 |
+
ignore:
|
14 |
+
- LlamaRotaryEmbedding
|
15 |
+
- LlamaRMSNorm
|
16 |
+
- SiLUActivation
|
17 |
+
- MatMulOutput_QK
|
18 |
+
- MatMulOutput_PV
|
19 |
+
- model.layers.21.mlp.down_proj
|
20 |
+
- model.layers.7.mlp.down_proj
|
21 |
+
- model.layers.2.mlp.down_proj
|
22 |
+
- model.layers.8.self_attn.q_proj
|
23 |
+
- model.layers.8.self_attn.k_proj
|
24 |
+
post_oneshot_calibration: true
|
25 |
+
scheme_overrides:
|
26 |
+
Linear:
|
27 |
+
weights:
|
28 |
+
num_bits: 8
|
29 |
+
symmetric: true
|
30 |
+
strategy: channel
|
31 |
+
MatMulLeftInput_QK:
|
32 |
+
input_activations:
|
33 |
+
num_bits: 8
|
34 |
+
symmetric: true
|
35 |
+
MatMulLeftInput_PV:
|
36 |
+
input_activations:
|
37 |
+
num_bits: 8
|
38 |
+
symmetric: true
|
39 |
+
Embedding:
|
40 |
+
input_activations: null
|
41 |
+
weights:
|
42 |
+
num_bits: 8
|
43 |
+
symmetric: false
|
44 |
+
SparseGPTModifier:
|
45 |
+
sparsity: 0.5
|
46 |
+
block_size: 128
|
47 |
+
sequential_update: true
|
48 |
+
quantize: true
|
49 |
+
percdamp: 0.01
|
50 |
+
mask_structure: 0:0
|
51 |
+
targets:
|
52 |
+
- re:model.layers.\d*$
|
53 |
+
test_stage_1:
|
54 |
+
obcq_modifiers:
|
55 |
+
LogarithmicEqualizationModifier:
|
56 |
+
mappings:
|
57 |
+
- - - re:.*q_proj
|
58 |
+
- re:.*k_proj
|
59 |
+
- re:.*v_proj
|
60 |
+
- re:.*input_layernorm
|
61 |
+
- - - re:.*gate_proj
|
62 |
+
- re:.*up_proj
|
63 |
+
- re:.*post_attention_layernorm
|
64 |
+
QuantizationModifier:
|
65 |
+
ignore:
|
66 |
+
- LlamaRotaryEmbedding
|
67 |
+
- LlamaRMSNorm
|
68 |
+
- SiLUActivation
|
69 |
+
- MatMulOutput_QK
|
70 |
+
- MatMulOutput_PV
|
71 |
+
- model.layers.21.mlp.down_proj
|
72 |
+
- model.layers.7.mlp.down_proj
|
73 |
+
- model.layers.2.mlp.down_proj
|
74 |
+
- model.layers.8.self_attn.q_proj
|
75 |
+
- model.layers.8.self_attn.k_proj
|
76 |
+
post_oneshot_calibration: true
|
77 |
+
scheme_overrides:
|
78 |
+
Linear:
|
79 |
+
weights:
|
80 |
+
num_bits: 8
|
81 |
+
symmetric: true
|
82 |
+
strategy: channel
|
83 |
+
MatMulLeftInput_QK:
|
84 |
+
input_activations:
|
85 |
+
num_bits: 8
|
86 |
+
symmetric: true
|
87 |
+
MatMulLeftInput_PV:
|
88 |
+
input_activations:
|
89 |
+
num_bits: 8
|
90 |
+
symmetric: true
|
91 |
+
Embedding:
|
92 |
+
input_activations: null
|
93 |
+
weights:
|
94 |
+
num_bits: 8
|
95 |
+
symmetric: false
|
96 |
+
SparseGPTModifier:
|
97 |
+
sparsity: 0.5
|
98 |
+
block_size: 128
|
99 |
+
sequential_update: true
|
100 |
+
quantize: true
|
101 |
+
percdamp: 0.01
|
102 |
+
mask_structure: 0:0
|
103 |
+
targets:
|
104 |
+
- re:model.layers.\d*$
|
checkpoint-2000/rng_state.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:1ff264f99d31b522cc7e2a4eac9d38606d0c58a34c0adc74d71e0ca8b371dc36
|
3 |
+
size 14244
|
checkpoint-2000/scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:115fc5d3c71fc8b112d5c1701359950aa4e48ea559451a73de1d1d8ca1116d89
|
3 |
+
size 1064
|
checkpoint-2000/special_tokens_map.json
ADDED
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"bos_token": {
|
3 |
+
"content": "<s>",
|
4 |
+
"lstrip": false,
|
5 |
+
"normalized": false,
|
6 |
+
"rstrip": false,
|
7 |
+
"single_word": false
|
8 |
+
},
|
9 |
+
"eos_token": {
|
10 |
+
"content": "</s>",
|
11 |
+
"lstrip": false,
|
12 |
+
"normalized": false,
|
13 |
+
"rstrip": false,
|
14 |
+
"single_word": false
|
15 |
+
},
|
16 |
+
"pad_token": {
|
17 |
+
"content": "</s>",
|
18 |
+
"lstrip": false,
|
19 |
+
"normalized": false,
|
20 |
+
"rstrip": false,
|
21 |
+
"single_word": false
|
22 |
+
},
|
23 |
+
"unk_token": {
|
24 |
+
"content": "<unk>",
|
25 |
+
"lstrip": false,
|
26 |
+
"normalized": false,
|
27 |
+
"rstrip": false,
|
28 |
+
"single_word": false
|
29 |
+
}
|
30 |
+
}
|
checkpoint-2000/tokenizer.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-2000/tokenizer.model
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
|
3 |
+
size 499723
|
checkpoint-2000/tokenizer_config.json
ADDED
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"added_tokens_decoder": {
|
3 |
+
"0": {
|
4 |
+
"content": "<unk>",
|
5 |
+
"lstrip": false,
|
6 |
+
"normalized": false,
|
7 |
+
"rstrip": false,
|
8 |
+
"single_word": false,
|
9 |
+
"special": true
|
10 |
+
},
|
11 |
+
"1": {
|
12 |
+
"content": "<s>",
|
13 |
+
"lstrip": false,
|
14 |
+
"normalized": false,
|
15 |
+
"rstrip": false,
|
16 |
+
"single_word": false,
|
17 |
+
"special": true
|
18 |
+
},
|
19 |
+
"2": {
|
20 |
+
"content": "</s>",
|
21 |
+
"lstrip": false,
|
22 |
+
"normalized": false,
|
23 |
+
"rstrip": false,
|
24 |
+
"single_word": false,
|
25 |
+
"special": true
|
26 |
+
}
|
27 |
+
},
|
28 |
+
"bos_token": "<s>",
|
29 |
+
"chat_template": "{% for message in messages %}\n{% if message['role'] == 'user' %}\n{{ '<|user|>\n' + message['content'] + eos_token }}\n{% elif message['role'] == 'system' %}\n{{ '<|system|>\n' + message['content'] + eos_token }}\n{% elif message['role'] == 'assistant' %}\n{{ '<|assistant|>\n' + message['content'] + eos_token }}\n{% endif %}\n{% if loop.last and add_generation_prompt %}\n{{ '<|assistant|>' }}\n{% endif %}\n{% endfor %}",
|
30 |
+
"clean_up_tokenization_spaces": false,
|
31 |
+
"eos_token": "</s>",
|
32 |
+
"legacy": false,
|
33 |
+
"model_max_length": 2048,
|
34 |
+
"pad_token": "</s>",
|
35 |
+
"padding_side": "right",
|
36 |
+
"sp_model_kwargs": {},
|
37 |
+
"tokenizer_class": "LlamaTokenizer",
|
38 |
+
"unk_token": "<unk>",
|
39 |
+
"use_default_system_prompt": false
|
40 |
+
}
|
checkpoint-2000/trainer_state.json
ADDED
@@ -0,0 +1,43 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": null,
|
3 |
+
"best_model_checkpoint": null,
|
4 |
+
"epoch": 1.2836970474967908,
|
5 |
+
"eval_steps": 500,
|
6 |
+
"global_step": 2000,
|
7 |
+
"is_hyper_param_search": false,
|
8 |
+
"is_local_process_zero": true,
|
9 |
+
"is_world_process_zero": true,
|
10 |
+
"log_history": [
|
11 |
+
{
|
12 |
+
"epoch": 0.32,
|
13 |
+
"learning_rate": 5e-05,
|
14 |
+
"loss": 31251.434,
|
15 |
+
"step": 500
|
16 |
+
},
|
17 |
+
{
|
18 |
+
"epoch": 0.64,
|
19 |
+
"learning_rate": 5e-05,
|
20 |
+
"loss": 31304.524,
|
21 |
+
"step": 1000
|
22 |
+
},
|
23 |
+
{
|
24 |
+
"epoch": 0.96,
|
25 |
+
"learning_rate": 5e-05,
|
26 |
+
"loss": 31288.94,
|
27 |
+
"step": 1500
|
28 |
+
},
|
29 |
+
{
|
30 |
+
"epoch": 1.28,
|
31 |
+
"learning_rate": 5e-05,
|
32 |
+
"loss": 31300.232,
|
33 |
+
"step": 2000
|
34 |
+
}
|
35 |
+
],
|
36 |
+
"logging_steps": 500,
|
37 |
+
"max_steps": 3116,
|
38 |
+
"num_train_epochs": 2,
|
39 |
+
"save_steps": 500,
|
40 |
+
"total_flos": 3.813388100724326e+16,
|
41 |
+
"trial_name": null,
|
42 |
+
"trial_params": null
|
43 |
+
}
|
checkpoint-2000/training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d259bb3b980ce004ad43f80c2d7f127a6d085a805ca671c63ea8e38660b643bb
|
3 |
+
size 4728
|
checkpoint-2500/config.json
ADDED
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_name_or_path": "./obcq_deployment",
|
3 |
+
"architectures": [
|
4 |
+
"LlamaForCausalLM"
|
5 |
+
],
|
6 |
+
"attention_bias": false,
|
7 |
+
"bos_token_id": 1,
|
8 |
+
"eos_token_id": 2,
|
9 |
+
"hidden_act": "silu",
|
10 |
+
"hidden_size": 2048,
|
11 |
+
"initializer_range": 0.02,
|
12 |
+
"intermediate_size": 5632,
|
13 |
+
"is_decoder": true,
|
14 |
+
"max_position_embeddings": 2048,
|
15 |
+
"model_type": "llama",
|
16 |
+
"num_attention_heads": 32,
|
17 |
+
"num_hidden_layers": 22,
|
18 |
+
"num_key_value_heads": 4,
|
19 |
+
"pretraining_tp": 1,
|
20 |
+
"rms_norm_eps": 1e-05,
|
21 |
+
"rope_scaling": null,
|
22 |
+
"rope_theta": 10000.0,
|
23 |
+
"tie_word_embeddings": false,
|
24 |
+
"torch_dtype": "float16",
|
25 |
+
"transformers_version": "1.7.0.20240122",
|
26 |
+
"use_cache": false,
|
27 |
+
"use_past": false,
|
28 |
+
"vocab_size": 32000
|
29 |
+
}
|
checkpoint-2500/generation_config.json
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"bos_token_id": 1,
|
3 |
+
"eos_token_id": 2,
|
4 |
+
"max_length": 2048,
|
5 |
+
"pad_token_id": 0,
|
6 |
+
"transformers_version": "1.7.0.20240122"
|
7 |
+
}
|
checkpoint-2500/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:39a5872508c57494ed2b601b032a556cc05da89f856f6628cd7c2196efd7dc4c
|
3 |
+
size 4400363238
|
checkpoint-2500/pytorch_model.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:ed6df076e5b420888f4eb6c7bfd39c01e7ca41f0404f945ff69d86071774fffd
|
3 |
+
size 2207873726
|
checkpoint-2500/recipe.yaml
ADDED
@@ -0,0 +1,104 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
test_stage_0:
|
2 |
+
obcq_modifiers:
|
3 |
+
LogarithmicEqualizationModifier:
|
4 |
+
mappings:
|
5 |
+
- - - re:.*q_proj
|
6 |
+
- re:.*k_proj
|
7 |
+
- re:.*v_proj
|
8 |
+
- re:.*input_layernorm
|
9 |
+
- - - re:.*gate_proj
|
10 |
+
- re:.*up_proj
|
11 |
+
- re:.*post_attention_layernorm
|
12 |
+
QuantizationModifier:
|
13 |
+
ignore:
|
14 |
+
- LlamaRotaryEmbedding
|
15 |
+
- LlamaRMSNorm
|
16 |
+
- SiLUActivation
|
17 |
+
- MatMulOutput_QK
|
18 |
+
- MatMulOutput_PV
|
19 |
+
- model.layers.21.mlp.down_proj
|
20 |
+
- model.layers.7.mlp.down_proj
|
21 |
+
- model.layers.2.mlp.down_proj
|
22 |
+
- model.layers.8.self_attn.q_proj
|
23 |
+
- model.layers.8.self_attn.k_proj
|
24 |
+
post_oneshot_calibration: true
|
25 |
+
scheme_overrides:
|
26 |
+
Linear:
|
27 |
+
weights:
|
28 |
+
num_bits: 8
|
29 |
+
symmetric: true
|
30 |
+
strategy: channel
|
31 |
+
MatMulLeftInput_QK:
|
32 |
+
input_activations:
|
33 |
+
num_bits: 8
|
34 |
+
symmetric: true
|
35 |
+
MatMulLeftInput_PV:
|
36 |
+
input_activations:
|
37 |
+
num_bits: 8
|
38 |
+
symmetric: true
|
39 |
+
Embedding:
|
40 |
+
input_activations: null
|
41 |
+
weights:
|
42 |
+
num_bits: 8
|
43 |
+
symmetric: false
|
44 |
+
SparseGPTModifier:
|
45 |
+
sparsity: 0.5
|
46 |
+
block_size: 128
|
47 |
+
sequential_update: true
|
48 |
+
quantize: true
|
49 |
+
percdamp: 0.01
|
50 |
+
mask_structure: 0:0
|
51 |
+
targets:
|
52 |
+
- re:model.layers.\d*$
|
53 |
+
test_stage_1:
|
54 |
+
obcq_modifiers:
|
55 |
+
LogarithmicEqualizationModifier:
|
56 |
+
mappings:
|
57 |
+
- - - re:.*q_proj
|
58 |
+
- re:.*k_proj
|
59 |
+
- re:.*v_proj
|
60 |
+
- re:.*input_layernorm
|
61 |
+
- - - re:.*gate_proj
|
62 |
+
- re:.*up_proj
|
63 |
+
- re:.*post_attention_layernorm
|
64 |
+
QuantizationModifier:
|
65 |
+
ignore:
|
66 |
+
- LlamaRotaryEmbedding
|
67 |
+
- LlamaRMSNorm
|
68 |
+
- SiLUActivation
|
69 |
+
- MatMulOutput_QK
|
70 |
+
- MatMulOutput_PV
|
71 |
+
- model.layers.21.mlp.down_proj
|
72 |
+
- model.layers.7.mlp.down_proj
|
73 |
+
- model.layers.2.mlp.down_proj
|
74 |
+
- model.layers.8.self_attn.q_proj
|
75 |
+
- model.layers.8.self_attn.k_proj
|
76 |
+
post_oneshot_calibration: true
|
77 |
+
scheme_overrides:
|
78 |
+
Linear:
|
79 |
+
weights:
|
80 |
+
num_bits: 8
|
81 |
+
symmetric: true
|
82 |
+
strategy: channel
|
83 |
+
MatMulLeftInput_QK:
|
84 |
+
input_activations:
|
85 |
+
num_bits: 8
|
86 |
+
symmetric: true
|
87 |
+
MatMulLeftInput_PV:
|
88 |
+
input_activations:
|
89 |
+
num_bits: 8
|
90 |
+
symmetric: true
|
91 |
+
Embedding:
|
92 |
+
input_activations: null
|
93 |
+
weights:
|
94 |
+
num_bits: 8
|
95 |
+
symmetric: false
|
96 |
+
SparseGPTModifier:
|
97 |
+
sparsity: 0.5
|
98 |
+
block_size: 128
|
99 |
+
sequential_update: true
|
100 |
+
quantize: true
|
101 |
+
percdamp: 0.01
|
102 |
+
mask_structure: 0:0
|
103 |
+
targets:
|
104 |
+
- re:model.layers.\d*$
|
checkpoint-2500/rng_state.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:1ff264f99d31b522cc7e2a4eac9d38606d0c58a34c0adc74d71e0ca8b371dc36
|
3 |
+
size 14244
|
checkpoint-2500/scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d9de07d7293968b9d2922a7f8c0991f17ddef6d3caf4a35f0fc3606f616dc368
|
3 |
+
size 1064
|
checkpoint-2500/special_tokens_map.json
ADDED
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"bos_token": {
|
3 |
+
"content": "<s>",
|
4 |
+
"lstrip": false,
|
5 |
+
"normalized": false,
|
6 |
+
"rstrip": false,
|
7 |
+
"single_word": false
|
8 |
+
},
|
9 |
+
"eos_token": {
|
10 |
+
"content": "</s>",
|
11 |
+
"lstrip": false,
|
12 |
+
"normalized": false,
|
13 |
+
"rstrip": false,
|
14 |
+
"single_word": false
|
15 |
+
},
|
16 |
+
"pad_token": {
|
17 |
+
"content": "</s>",
|
18 |
+
"lstrip": false,
|
19 |
+
"normalized": false,
|
20 |
+
"rstrip": false,
|
21 |
+
"single_word": false
|
22 |
+
},
|
23 |
+
"unk_token": {
|
24 |
+
"content": "<unk>",
|
25 |
+
"lstrip": false,
|
26 |
+
"normalized": false,
|
27 |
+
"rstrip": false,
|
28 |
+
"single_word": false
|
29 |
+
}
|
30 |
+
}
|
checkpoint-2500/tokenizer.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-2500/tokenizer.model
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
|
3 |
+
size 499723
|