Hugging Face's logo

odats
/
nmt_21

Text Generation
Transformers
Safetensors
gemma3_text
Generated from Trainer
grpo
trl
conversational
text-generation-inference
Model card Files
xet
Community
odats commited on
Commit
38d0952
·
verified ·
1 Parent(s): 945b2ce

Training in progress, step 5800

Browse files
Files changed (5) hide show
  1. README.md +3 -3
  2. config.json +85 -54
  3. model-00001-of-00002.safetensors +1 -1
  4. model-00002-of-00002.safetensors +1 -1
  5. training_args.bin +1 -1
README.md CHANGED
@@ -1,5 +1,5 @@
1
  ---
2
- base_model: google/gemma-3-1b-it
3
  library_name: transformers
4
  model_name: nmt_21
5
  tags:
@@ -11,7 +11,7 @@ licence: license
11
 
12
  # Model Card for nmt_21
13
 
14
- This model is a fine-tuned version of [google/gemma-3-1b-it](https://huggingface.co/google/gemma-3-1b-it).
15
  It has been trained using [TRL](https://github.com/huggingface/trl).
16
 
17
  ## Quick start
@@ -27,7 +27,7 @@ print(output["generated_text"])
27
 
28
  ## Training procedure
29
 
30
- [<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/oleg-dats/nmt/runs/1oefvyo4)
31
 
32
 
33
  This model was trained with GRPO, a method introduced in [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://huggingface.co/papers/2402.03300).
 
1
  ---
2
+ base_model: google/gemma-3-4b-it
3
  library_name: transformers
4
  model_name: nmt_21
5
  tags:
 
11
 
12
  # Model Card for nmt_21
13
 
14
+ This model is a fine-tuned version of [google/gemma-3-4b-it](https://huggingface.co/google/gemma-3-4b-it).
15
  It has been trained using [TRL](https://github.com/huggingface/trl).
16
 
17
  ## Quick start
 
27
 
28
  ## Training procedure
29
 
30
+ [<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/oleg-dats/nmt/runs/yn1sslos)
31
 
32
 
33
  This model was trained with GRPO, a method introduced in [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://huggingface.co/papers/2402.03300).
config.json CHANGED
@@ -1,64 +1,95 @@
1
  {
2
- "_sliding_window_pattern": 6,
3
  "architectures": [
4
- "Gemma3ForCausalLM"
5
  ],
6
- "attention_bias": false,
7
- "attention_dropout": 0.0,
8
- "attn_logit_softcapping": null,
9
  "bos_token_id": 2,
10
- "cache_implementation": "hybrid",
11
  "dtype": "bfloat16",
 
12
  "eos_token_id": 1,
13
- "final_logit_softcapping": null,
14
- "head_dim": 256,
15
- "hidden_activation": "gelu_pytorch_tanh",
16
- "hidden_size": 1152,
17
  "initializer_range": 0.02,
18
- "intermediate_size": 6912,
19
- "layer_types": [
20
- "sliding_attention",
21
- "sliding_attention",
22
- "sliding_attention",
23
- "sliding_attention",
24
- "sliding_attention",
25
- "full_attention",
26
- "sliding_attention",
27
- "sliding_attention",
28
- "sliding_attention",
29
- "sliding_attention",
30
- "sliding_attention",
31
- "full_attention",
32
- "sliding_attention",
33
- "sliding_attention",
34
- "sliding_attention",
35
- "sliding_attention",
36
- "sliding_attention",
37
- "full_attention",
38
- "sliding_attention",
39
- "sliding_attention",
40
- "sliding_attention",
41
- "sliding_attention",
42
- "sliding_attention",
43
- "full_attention",
44
- "sliding_attention",
45
- "sliding_attention"
46
- ],
47
- "max_position_embeddings": 32768,
48
- "model_type": "gemma3_text",
49
- "num_attention_heads": 4,
50
- "num_hidden_layers": 26,
51
- "num_key_value_heads": 1,
52
  "pad_token_id": 0,
53
- "query_pre_attn_scalar": 256,
54
- "rms_norm_eps": 1e-06,
55
- "rope_local_base_freq": 10000,
56
- "rope_scaling": null,
57
- "rope_theta": 1000000,
58
- "sliding_window": 512,
59
- "sliding_window_pattern": 6,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
60
  "transformers_version": "4.57.0",
61
- "use_bidirectional_attention": false,
62
- "use_cache": true,
63
- "vocab_size": 262144
 
 
 
 
 
 
 
 
 
 
 
64
  }
 
1
  {
 
2
  "architectures": [
3
+ "Gemma3ForConditionalGeneration"
4
  ],
5
+ "boi_token_index": 255999,
 
 
6
  "bos_token_id": 2,
 
7
  "dtype": "bfloat16",
8
+ "eoi_token_index": 256000,
9
  "eos_token_id": 1,
10
+ "image_token_index": 262144,
 
 
 
11
  "initializer_range": 0.02,
12
+ "mm_tokens_per_image": 256,
13
+ "model_type": "gemma3",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
14
  "pad_token_id": 0,
15
+ "text_config": {
16
+ "_sliding_window_pattern": 6,
17
+ "attention_bias": false,
18
+ "attention_dropout": 0.0,
19
+ "attn_logit_softcapping": null,
20
+ "final_logit_softcapping": null,
21
+ "head_dim": 256,
22
+ "hidden_activation": "gelu_pytorch_tanh",
23
+ "hidden_size": 2560,
24
+ "initializer_range": 0.02,
25
+ "intermediate_size": 10240,
26
+ "layer_types": [
27
+ "sliding_attention",
28
+ "sliding_attention",
29
+ "sliding_attention",
30
+ "sliding_attention",
31
+ "sliding_attention",
32
+ "full_attention",
33
+ "sliding_attention",
34
+ "sliding_attention",
35
+ "sliding_attention",
36
+ "sliding_attention",
37
+ "sliding_attention",
38
+ "full_attention",
39
+ "sliding_attention",
40
+ "sliding_attention",
41
+ "sliding_attention",
42
+ "sliding_attention",
43
+ "sliding_attention",
44
+ "full_attention",
45
+ "sliding_attention",
46
+ "sliding_attention",
47
+ "sliding_attention",
48
+ "sliding_attention",
49
+ "sliding_attention",
50
+ "full_attention",
51
+ "sliding_attention",
52
+ "sliding_attention",
53
+ "sliding_attention",
54
+ "sliding_attention",
55
+ "sliding_attention",
56
+ "full_attention",
57
+ "sliding_attention",
58
+ "sliding_attention",
59
+ "sliding_attention",
60
+ "sliding_attention"
61
+ ],
62
+ "max_position_embeddings": 131072,
63
+ "model_type": "gemma3_text",
64
+ "num_attention_heads": 8,
65
+ "num_hidden_layers": 34,
66
+ "num_key_value_heads": 4,
67
+ "query_pre_attn_scalar": 256,
68
+ "rms_norm_eps": 1e-06,
69
+ "rope_local_base_freq": 10000.0,
70
+ "rope_scaling": {
71
+ "factor": 8.0,
72
+ "rope_type": "linear"
73
+ },
74
+ "rope_theta": 1000000.0,
75
+ "sliding_window": 1024,
76
+ "use_bidirectional_attention": false,
77
+ "use_cache": true,
78
+ "vocab_size": 262208
79
+ },
80
  "transformers_version": "4.57.0",
81
+ "vision_config": {
82
+ "attention_dropout": 0.0,
83
+ "hidden_act": "gelu_pytorch_tanh",
84
+ "hidden_size": 1152,
85
+ "image_size": 896,
86
+ "intermediate_size": 4304,
87
+ "layer_norm_eps": 1e-06,
88
+ "model_type": "siglip_vision_model",
89
+ "num_attention_heads": 16,
90
+ "num_channels": 3,
91
+ "num_hidden_layers": 27,
92
+ "patch_size": 14,
93
+ "vision_use_head": false
94
+ }
95
  }
model-00001-of-00002.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:e904c326c0afb55cce8a33df889a4ceab24ea9c9a586a0c3feb7febf834b2a4f
3
  size 4961251752
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6a653ce67ccca35b508e079bb5328530c791b3e71f267b172865041b6b04e88e
3
  size 4961251752
model-00002-of-00002.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:ce36aaeda348a869c066fbeb88dd662af65a0cef6445c6a10e6fca78bc27f7b5
3
  size 3639026128
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:24d395bd995ee0eb269aab4db2c4197c2515af165dc9f1c94bbc3314f2dace51
3
  size 3639026128
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:5dcc949cc10c8ffc858cb199cdb4ab6d028edbb9888ede489a27d7d06d3733dd
3
  size 6840
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:54eac288f57ef58be29ddb0e8adea9bc712650afeaf8d7d4f42255c44a942cbb
3
  size 6840