kimgreg commited on
Commit
75b8e50
·
verified ·
1 Parent(s): 8e5d352

checkpoints of ctl-k2-sft

Browse files
ctl-k2-sft/e0.25/config.json ADDED
@@ -0,0 +1,65 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Lfm2ForCausalLM"
4
+ ],
5
+ "block__name_mlp": "parallel_mlp_merged",
6
+ "block_auto_adjust_ff_dim": false,
7
+ "block_dim": 1024,
8
+ "block_ffn_dim_multiplier": 1.0,
9
+ "block_ffn_te_autocast": false,
10
+ "block_ffn_use_quantized_params": false,
11
+ "block_mlp_init_scale": 1.0,
12
+ "block_multiple_of": 256,
13
+ "block_norm_eps": 1e-05,
14
+ "block_out_init_scale": 1.0,
15
+ "block_sequence_parallel_norm_across_tp": false,
16
+ "block_use_swiglu": true,
17
+ "block_use_xavier_init": true,
18
+ "bos_token_id": 1,
19
+ "conv_L_cache": 3,
20
+ "conv_bias": false,
21
+ "conv_dim": 1024,
22
+ "conv_use_xavier_init": true,
23
+ "dtype": "float32",
24
+ "eos_token_id": 7,
25
+ "ffn_te_autocast": false,
26
+ "ffn_use_quantized_params": false,
27
+ "full_attn_idxs": null,
28
+ "hidden_size": 1024,
29
+ "initializer_range": 0.02,
30
+ "intermediate_size": 2560,
31
+ "layer_types": [
32
+ "conv",
33
+ "conv",
34
+ "full_attention",
35
+ "conv",
36
+ "full_attention",
37
+ "conv",
38
+ "full_attention",
39
+ "conv",
40
+ "full_attention",
41
+ "conv",
42
+ "full_attention",
43
+ "conv",
44
+ "full_attention",
45
+ "conv"
46
+ ],
47
+ "max_position_embeddings": 128000,
48
+ "model_type": "lfm2",
49
+ "norm_eps": 1e-05,
50
+ "num_attention_heads": 16,
51
+ "num_heads": 16,
52
+ "num_hidden_layers": 14,
53
+ "num_key_value_heads": 8,
54
+ "pad_token_id": 0,
55
+ "rope_parameters": {
56
+ "rope_theta": 1000000.0,
57
+ "rope_type": "default"
58
+ },
59
+ "sequence_parallel_norm_across_tp": false,
60
+ "tie_word_embeddings": true,
61
+ "transformers_version": "5.22.2",
62
+ "use_cache": true,
63
+ "use_pos_enc": true,
64
+ "vocab_size": 65536
65
+ }
ctl-k2-sft/e0.25/generation_config.json ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 1,
4
+ "do_sample": true,
5
+ "eos_token_id": 7,
6
+ "output_attentions": false,
7
+ "output_hidden_states": false,
8
+ "pad_token_id": 0,
9
+ "repetition_penalty": 1.05,
10
+ "temperature": 0.1,
11
+ "top_k": 50,
12
+ "transformers_version": "5.22.2",
13
+ "use_cache": true
14
+ }
ctl-k2-sft/e0.25/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:54b8ca4b933ff434b06718df36d3b4e544bc5390e20ef07b69b626eccac21da0
3
+ size 593618936
ctl-k2-sft/e0.75/config.json ADDED
@@ -0,0 +1,65 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Lfm2ForCausalLM"
4
+ ],
5
+ "block__name_mlp": "parallel_mlp_merged",
6
+ "block_auto_adjust_ff_dim": false,
7
+ "block_dim": 1024,
8
+ "block_ffn_dim_multiplier": 1.0,
9
+ "block_ffn_te_autocast": false,
10
+ "block_ffn_use_quantized_params": false,
11
+ "block_mlp_init_scale": 1.0,
12
+ "block_multiple_of": 256,
13
+ "block_norm_eps": 1e-05,
14
+ "block_out_init_scale": 1.0,
15
+ "block_sequence_parallel_norm_across_tp": false,
16
+ "block_use_swiglu": true,
17
+ "block_use_xavier_init": true,
18
+ "bos_token_id": 1,
19
+ "conv_L_cache": 3,
20
+ "conv_bias": false,
21
+ "conv_dim": 1024,
22
+ "conv_use_xavier_init": true,
23
+ "dtype": "float32",
24
+ "eos_token_id": 7,
25
+ "ffn_te_autocast": false,
26
+ "ffn_use_quantized_params": false,
27
+ "full_attn_idxs": null,
28
+ "hidden_size": 1024,
29
+ "initializer_range": 0.02,
30
+ "intermediate_size": 2560,
31
+ "layer_types": [
32
+ "conv",
33
+ "conv",
34
+ "full_attention",
35
+ "conv",
36
+ "full_attention",
37
+ "conv",
38
+ "full_attention",
39
+ "conv",
40
+ "full_attention",
41
+ "conv",
42
+ "full_attention",
43
+ "conv",
44
+ "full_attention",
45
+ "conv"
46
+ ],
47
+ "max_position_embeddings": 128000,
48
+ "model_type": "lfm2",
49
+ "norm_eps": 1e-05,
50
+ "num_attention_heads": 16,
51
+ "num_heads": 16,
52
+ "num_hidden_layers": 14,
53
+ "num_key_value_heads": 8,
54
+ "pad_token_id": 0,
55
+ "rope_parameters": {
56
+ "rope_theta": 1000000.0,
57
+ "rope_type": "default"
58
+ },
59
+ "sequence_parallel_norm_across_tp": false,
60
+ "tie_word_embeddings": true,
61
+ "transformers_version": "5.22.2",
62
+ "use_cache": true,
63
+ "use_pos_enc": true,
64
+ "vocab_size": 65536
65
+ }
ctl-k2-sft/e0.75/generation_config.json ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 1,
4
+ "do_sample": true,
5
+ "eos_token_id": 7,
6
+ "output_attentions": false,
7
+ "output_hidden_states": false,
8
+ "pad_token_id": 0,
9
+ "repetition_penalty": 1.05,
10
+ "temperature": 0.1,
11
+ "top_k": 50,
12
+ "transformers_version": "5.22.2",
13
+ "use_cache": true
14
+ }
ctl-k2-sft/e0.75/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:01afb850bb75ab28f919a85c619f0019bd597291a630b82dc5f7b0fe74d67130
3
+ size 593618936
ctl-k2-sft/e1/config.json ADDED
@@ -0,0 +1,65 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Lfm2ForCausalLM"
4
+ ],
5
+ "block__name_mlp": "parallel_mlp_merged",
6
+ "block_auto_adjust_ff_dim": false,
7
+ "block_dim": 1024,
8
+ "block_ffn_dim_multiplier": 1.0,
9
+ "block_ffn_te_autocast": false,
10
+ "block_ffn_use_quantized_params": false,
11
+ "block_mlp_init_scale": 1.0,
12
+ "block_multiple_of": 256,
13
+ "block_norm_eps": 1e-05,
14
+ "block_out_init_scale": 1.0,
15
+ "block_sequence_parallel_norm_across_tp": false,
16
+ "block_use_swiglu": true,
17
+ "block_use_xavier_init": true,
18
+ "bos_token_id": 1,
19
+ "conv_L_cache": 3,
20
+ "conv_bias": false,
21
+ "conv_dim": 1024,
22
+ "conv_use_xavier_init": true,
23
+ "dtype": "float32",
24
+ "eos_token_id": 7,
25
+ "ffn_te_autocast": false,
26
+ "ffn_use_quantized_params": false,
27
+ "full_attn_idxs": null,
28
+ "hidden_size": 1024,
29
+ "initializer_range": 0.02,
30
+ "intermediate_size": 2560,
31
+ "layer_types": [
32
+ "conv",
33
+ "conv",
34
+ "full_attention",
35
+ "conv",
36
+ "full_attention",
37
+ "conv",
38
+ "full_attention",
39
+ "conv",
40
+ "full_attention",
41
+ "conv",
42
+ "full_attention",
43
+ "conv",
44
+ "full_attention",
45
+ "conv"
46
+ ],
47
+ "max_position_embeddings": 128000,
48
+ "model_type": "lfm2",
49
+ "norm_eps": 1e-05,
50
+ "num_attention_heads": 16,
51
+ "num_heads": 16,
52
+ "num_hidden_layers": 14,
53
+ "num_key_value_heads": 8,
54
+ "pad_token_id": 0,
55
+ "rope_parameters": {
56
+ "rope_theta": 1000000.0,
57
+ "rope_type": "default"
58
+ },
59
+ "sequence_parallel_norm_across_tp": false,
60
+ "tie_word_embeddings": true,
61
+ "transformers_version": "5.22.2",
62
+ "use_cache": true,
63
+ "use_pos_enc": true,
64
+ "vocab_size": 65536
65
+ }
ctl-k2-sft/e1/generation_config.json ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 1,
4
+ "do_sample": true,
5
+ "eos_token_id": 7,
6
+ "output_attentions": false,
7
+ "output_hidden_states": false,
8
+ "pad_token_id": 0,
9
+ "repetition_penalty": 1.05,
10
+ "temperature": 0.1,
11
+ "top_k": 50,
12
+ "transformers_version": "5.22.2",
13
+ "use_cache": true
14
+ }
ctl-k2-sft/e1/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3b0ba8f4b54c64ceb5e3c0e0f5a8765d63506b5894c73cb6758c283b5c51f72f
3
+ size 593618936
ctl-k2-sft/e20/config.json ADDED
@@ -0,0 +1,65 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Lfm2ForCausalLM"
4
+ ],
5
+ "block__name_mlp": "parallel_mlp_merged",
6
+ "block_auto_adjust_ff_dim": false,
7
+ "block_dim": 1024,
8
+ "block_ffn_dim_multiplier": 1.0,
9
+ "block_ffn_te_autocast": false,
10
+ "block_ffn_use_quantized_params": false,
11
+ "block_mlp_init_scale": 1.0,
12
+ "block_multiple_of": 256,
13
+ "block_norm_eps": 1e-05,
14
+ "block_out_init_scale": 1.0,
15
+ "block_sequence_parallel_norm_across_tp": false,
16
+ "block_use_swiglu": true,
17
+ "block_use_xavier_init": true,
18
+ "bos_token_id": 1,
19
+ "conv_L_cache": 3,
20
+ "conv_bias": false,
21
+ "conv_dim": 1024,
22
+ "conv_use_xavier_init": true,
23
+ "dtype": "float32",
24
+ "eos_token_id": 7,
25
+ "ffn_te_autocast": false,
26
+ "ffn_use_quantized_params": false,
27
+ "full_attn_idxs": null,
28
+ "hidden_size": 1024,
29
+ "initializer_range": 0.02,
30
+ "intermediate_size": 2560,
31
+ "layer_types": [
32
+ "conv",
33
+ "conv",
34
+ "full_attention",
35
+ "conv",
36
+ "full_attention",
37
+ "conv",
38
+ "full_attention",
39
+ "conv",
40
+ "full_attention",
41
+ "conv",
42
+ "full_attention",
43
+ "conv",
44
+ "full_attention",
45
+ "conv"
46
+ ],
47
+ "max_position_embeddings": 128000,
48
+ "model_type": "lfm2",
49
+ "norm_eps": 1e-05,
50
+ "num_attention_heads": 16,
51
+ "num_heads": 16,
52
+ "num_hidden_layers": 14,
53
+ "num_key_value_heads": 8,
54
+ "pad_token_id": 0,
55
+ "rope_parameters": {
56
+ "rope_theta": 1000000.0,
57
+ "rope_type": "default"
58
+ },
59
+ "sequence_parallel_norm_across_tp": false,
60
+ "tie_word_embeddings": true,
61
+ "transformers_version": "5.22.2",
62
+ "use_cache": true,
63
+ "use_pos_enc": true,
64
+ "vocab_size": 65536
65
+ }
ctl-k2-sft/e20/generation_config.json ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 1,
4
+ "do_sample": true,
5
+ "eos_token_id": 7,
6
+ "output_attentions": false,
7
+ "output_hidden_states": false,
8
+ "pad_token_id": 0,
9
+ "repetition_penalty": 1.05,
10
+ "temperature": 0.1,
11
+ "top_k": 50,
12
+ "transformers_version": "5.22.2",
13
+ "use_cache": true
14
+ }
ctl-k2-sft/e20/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:61a330954dc113d25db54bdf792a8b9312c4236e57210e6b9ee2f61c15ff1053
3
+ size 593618936
ctl-k2-sft/e3/config.json ADDED
@@ -0,0 +1,65 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Lfm2ForCausalLM"
4
+ ],
5
+ "block__name_mlp": "parallel_mlp_merged",
6
+ "block_auto_adjust_ff_dim": false,
7
+ "block_dim": 1024,
8
+ "block_ffn_dim_multiplier": 1.0,
9
+ "block_ffn_te_autocast": false,
10
+ "block_ffn_use_quantized_params": false,
11
+ "block_mlp_init_scale": 1.0,
12
+ "block_multiple_of": 256,
13
+ "block_norm_eps": 1e-05,
14
+ "block_out_init_scale": 1.0,
15
+ "block_sequence_parallel_norm_across_tp": false,
16
+ "block_use_swiglu": true,
17
+ "block_use_xavier_init": true,
18
+ "bos_token_id": 1,
19
+ "conv_L_cache": 3,
20
+ "conv_bias": false,
21
+ "conv_dim": 1024,
22
+ "conv_use_xavier_init": true,
23
+ "dtype": "float32",
24
+ "eos_token_id": 7,
25
+ "ffn_te_autocast": false,
26
+ "ffn_use_quantized_params": false,
27
+ "full_attn_idxs": null,
28
+ "hidden_size": 1024,
29
+ "initializer_range": 0.02,
30
+ "intermediate_size": 2560,
31
+ "layer_types": [
32
+ "conv",
33
+ "conv",
34
+ "full_attention",
35
+ "conv",
36
+ "full_attention",
37
+ "conv",
38
+ "full_attention",
39
+ "conv",
40
+ "full_attention",
41
+ "conv",
42
+ "full_attention",
43
+ "conv",
44
+ "full_attention",
45
+ "conv"
46
+ ],
47
+ "max_position_embeddings": 128000,
48
+ "model_type": "lfm2",
49
+ "norm_eps": 1e-05,
50
+ "num_attention_heads": 16,
51
+ "num_heads": 16,
52
+ "num_hidden_layers": 14,
53
+ "num_key_value_heads": 8,
54
+ "pad_token_id": 0,
55
+ "rope_parameters": {
56
+ "rope_theta": 1000000.0,
57
+ "rope_type": "default"
58
+ },
59
+ "sequence_parallel_norm_across_tp": false,
60
+ "tie_word_embeddings": true,
61
+ "transformers_version": "5.22.2",
62
+ "use_cache": true,
63
+ "use_pos_enc": true,
64
+ "vocab_size": 65536
65
+ }
ctl-k2-sft/e3/generation_config.json ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 1,
4
+ "do_sample": true,
5
+ "eos_token_id": 7,
6
+ "output_attentions": false,
7
+ "output_hidden_states": false,
8
+ "pad_token_id": 0,
9
+ "repetition_penalty": 1.05,
10
+ "temperature": 0.1,
11
+ "top_k": 50,
12
+ "transformers_version": "5.22.2",
13
+ "use_cache": true
14
+ }
ctl-k2-sft/e3/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f02a75ed45f994b163ee8ede130ee840fe0fadb3455cd88571363b8c32cbe61c
3
+ size 593618936
ctl-k2-sft/e8/config.json ADDED
@@ -0,0 +1,65 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Lfm2ForCausalLM"
4
+ ],
5
+ "block__name_mlp": "parallel_mlp_merged",
6
+ "block_auto_adjust_ff_dim": false,
7
+ "block_dim": 1024,
8
+ "block_ffn_dim_multiplier": 1.0,
9
+ "block_ffn_te_autocast": false,
10
+ "block_ffn_use_quantized_params": false,
11
+ "block_mlp_init_scale": 1.0,
12
+ "block_multiple_of": 256,
13
+ "block_norm_eps": 1e-05,
14
+ "block_out_init_scale": 1.0,
15
+ "block_sequence_parallel_norm_across_tp": false,
16
+ "block_use_swiglu": true,
17
+ "block_use_xavier_init": true,
18
+ "bos_token_id": 1,
19
+ "conv_L_cache": 3,
20
+ "conv_bias": false,
21
+ "conv_dim": 1024,
22
+ "conv_use_xavier_init": true,
23
+ "dtype": "float32",
24
+ "eos_token_id": 7,
25
+ "ffn_te_autocast": false,
26
+ "ffn_use_quantized_params": false,
27
+ "full_attn_idxs": null,
28
+ "hidden_size": 1024,
29
+ "initializer_range": 0.02,
30
+ "intermediate_size": 2560,
31
+ "layer_types": [
32
+ "conv",
33
+ "conv",
34
+ "full_attention",
35
+ "conv",
36
+ "full_attention",
37
+ "conv",
38
+ "full_attention",
39
+ "conv",
40
+ "full_attention",
41
+ "conv",
42
+ "full_attention",
43
+ "conv",
44
+ "full_attention",
45
+ "conv"
46
+ ],
47
+ "max_position_embeddings": 128000,
48
+ "model_type": "lfm2",
49
+ "norm_eps": 1e-05,
50
+ "num_attention_heads": 16,
51
+ "num_heads": 16,
52
+ "num_hidden_layers": 14,
53
+ "num_key_value_heads": 8,
54
+ "pad_token_id": 0,
55
+ "rope_parameters": {
56
+ "rope_theta": 1000000.0,
57
+ "rope_type": "default"
58
+ },
59
+ "sequence_parallel_norm_across_tp": false,
60
+ "tie_word_embeddings": true,
61
+ "transformers_version": "5.22.2",
62
+ "use_cache": true,
63
+ "use_pos_enc": true,
64
+ "vocab_size": 65536
65
+ }
ctl-k2-sft/e8/generation_config.json ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 1,
4
+ "do_sample": true,
5
+ "eos_token_id": 7,
6
+ "output_attentions": false,
7
+ "output_hidden_states": false,
8
+ "pad_token_id": 0,
9
+ "repetition_penalty": 1.05,
10
+ "temperature": 0.1,
11
+ "top_k": 50,
12
+ "transformers_version": "5.22.2",
13
+ "use_cache": true
14
+ }
ctl-k2-sft/e8/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4bbfd1639c7afe57a1c0cb93a09144c4640c388709e035474151851f035e7205
3
+ size 593618936
ctl-k2-sft/run-config.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": ".local/models/lfm2.5-230m",
3
+ "kb": "data/kb/kb-v1",
4
+ "data": "data/synth/kb-v1-ctl-k2",
5
+ "out": "results/runs/ctl-k2-sft",
6
+ "epochs": 20,
7
+ "save_at": "0.05,0.1,0.25,0.5,0.75,1,1.5,2,3,5,8,12,20",
8
+ "objective": "sft",
9
+ "objective_args": "{}",
10
+ "replay": null,
11
+ "replay_ratio": 0.0,
12
+ "lr": 5e-05,
13
+ "batch": 16,
14
+ "seed": 1234,
15
+ "data_manifest": {
16
+ "train/pairs.jsonl": "9590c2ac8d0e8b531f5fe4bd5062b4e027dce3276bc92679ea96790ed0be08f1"
17
+ }
18
+ }