Hugging Face's logo

umd-vt-nyu
/
ar1e4_cos_llama1b_bsz64_16

Safetensors
Model card Files Community
xcpan commited on
Commit
4b86e9b
·
verified ·
1 Parent(s): df713df

Upload folder using huggingface_hub

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. checkpoint-16290/config.json +62 -0
  2. checkpoint-16290/generation_config.json +13 -0
  3. checkpoint-16290/latest +1 -0
  4. checkpoint-16290/model-00001-of-00003.safetensors +3 -0
  5. checkpoint-16290/model-00002-of-00003.safetensors +3 -0
  6. checkpoint-16290/model-00003-of-00003.safetensors +3 -0
  7. checkpoint-16290/model.safetensors.index.json +0 -0
  8. checkpoint-16290/rng_state_0.pth +3 -0
  9. checkpoint-16290/rng_state_1.pth +3 -0
  10. checkpoint-16290/rng_state_10.pth +3 -0
  11. checkpoint-16290/rng_state_11.pth +3 -0
  12. checkpoint-16290/rng_state_12.pth +3 -0
  13. checkpoint-16290/rng_state_13.pth +3 -0
  14. checkpoint-16290/rng_state_14.pth +3 -0
  15. checkpoint-16290/rng_state_15.pth +3 -0
  16. checkpoint-16290/rng_state_16.pth +3 -0
  17. checkpoint-16290/rng_state_17.pth +3 -0
  18. checkpoint-16290/rng_state_18.pth +3 -0
  19. checkpoint-16290/rng_state_19.pth +3 -0
  20. checkpoint-16290/rng_state_2.pth +3 -0
  21. checkpoint-16290/rng_state_20.pth +3 -0
  22. checkpoint-16290/rng_state_21.pth +3 -0
  23. checkpoint-16290/rng_state_22.pth +3 -0
  24. checkpoint-16290/rng_state_23.pth +3 -0
  25. checkpoint-16290/rng_state_24.pth +3 -0
  26. checkpoint-16290/rng_state_25.pth +3 -0
  27. checkpoint-16290/rng_state_26.pth +3 -0
  28. checkpoint-16290/rng_state_27.pth +3 -0
  29. checkpoint-16290/rng_state_28.pth +3 -0
  30. checkpoint-16290/rng_state_29.pth +3 -0
  31. checkpoint-16290/rng_state_3.pth +3 -0
  32. checkpoint-16290/rng_state_30.pth +3 -0
  33. checkpoint-16290/rng_state_31.pth +3 -0
  34. checkpoint-16290/rng_state_32.pth +3 -0
  35. checkpoint-16290/rng_state_33.pth +3 -0
  36. checkpoint-16290/rng_state_34.pth +3 -0
  37. checkpoint-16290/rng_state_35.pth +3 -0
  38. checkpoint-16290/rng_state_36.pth +3 -0
  39. checkpoint-16290/rng_state_37.pth +3 -0
  40. checkpoint-16290/rng_state_38.pth +3 -0
  41. checkpoint-16290/rng_state_39.pth +3 -0
  42. checkpoint-16290/rng_state_4.pth +3 -0
  43. checkpoint-16290/rng_state_40.pth +3 -0
  44. checkpoint-16290/rng_state_41.pth +3 -0
  45. checkpoint-16290/rng_state_42.pth +3 -0
  46. checkpoint-16290/rng_state_43.pth +3 -0
  47. checkpoint-16290/rng_state_44.pth +3 -0
  48. checkpoint-16290/rng_state_45.pth +3 -0
  49. checkpoint-16290/rng_state_46.pth +3 -0
  50. checkpoint-16290/rng_state_47.pth +3 -0
checkpoint-16290/config.json ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "meta-llama/Llama-3.2-1B-Instruct",
3
+ "architectures": [
4
+ "LlavaLlamaForCausalLM"
5
+ ],
6
+ "attention_bias": false,
7
+ "attention_dropout": 0.0,
8
+ "bos_token_id": 128000,
9
+ "eos_token_id": [
10
+ 128001,
11
+ 128008,
12
+ 128009
13
+ ],
14
+ "freeze_mm_mlp_adapter": false,
15
+ "gen_hidden_size": 1792,
16
+ "gen_pooling": "early_pool2d_4",
17
+ "gen_projector_type": "mlp2x_gelu",
18
+ "gen_vision_tower": "eva-clip-E-14-plus",
19
+ "head_dim": 64,
20
+ "hidden_act": "silu",
21
+ "hidden_size": 2048,
22
+ "image_aspect_ratio": "square",
23
+ "initializer_range": 0.02,
24
+ "intermediate_size": 8192,
25
+ "max_position_embeddings": 131072,
26
+ "mlp_bias": false,
27
+ "mm_hidden_size": 1152,
28
+ "mm_patch_merge_type": "flat",
29
+ "mm_projector_lr": null,
30
+ "mm_projector_type": "mlp2x_gelu",
31
+ "mm_use_im_patch_token": false,
32
+ "mm_use_im_start_end": false,
33
+ "mm_vision_select_feature": "patch",
34
+ "mm_vision_select_layer": -2,
35
+ "mm_vision_tower": "google/siglip-so400m-patch14-384",
36
+ "model_type": "llava_llama",
37
+ "n_query": 64,
38
+ "num_attention_heads": 32,
39
+ "num_hidden_layers": 16,
40
+ "num_key_value_heads": 8,
41
+ "pad_token_id": 128256,
42
+ "pretraining_tp": 1,
43
+ "rms_norm_eps": 1e-05,
44
+ "rope_scaling": {
45
+ "factor": 32.0,
46
+ "high_freq_factor": 4.0,
47
+ "low_freq_factor": 1.0,
48
+ "original_max_position_embeddings": 8192,
49
+ "rope_type": "llama3"
50
+ },
51
+ "rope_theta": 500000.0,
52
+ "tie_word_embeddings": true,
53
+ "tokenizer_model_max_length": 256,
54
+ "tokenizer_padding_side": "right",
55
+ "torch_dtype": "bfloat16",
56
+ "transformers_version": "4.43.1",
57
+ "tune_mm_mlp_adapter": false,
58
+ "use_cache": false,
59
+ "use_mm_proj": true,
60
+ "vision_tower_pretrained": null,
61
+ "vocab_size": 128260
62
+ }
checkpoint-16290/generation_config.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "attn_implementation": "flash_attention_2",
3
+ "bos_token_id": 128000,
4
+ "do_sample": true,
5
+ "eos_token_id": [
6
+ 128001,
7
+ 128008,
8
+ 128009
9
+ ],
10
+ "temperature": 0.6,
11
+ "top_p": 0.9,
12
+ "transformers_version": "4.43.1"
13
+ }
checkpoint-16290/latest ADDED
@@ -0,0 +1 @@
 
 
1
+ global_step16290
checkpoint-16290/model-00001-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1bfafa18917f828c0b880a85200023eb4ca04cf1e334b55c9343f45f54f4d32e
3
+ size 4959476728
checkpoint-16290/model-00002-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:49290ef1d8933723e8b40ceeeb16a9117706a0073b7e88d8c0d56695b987a6c0
3
+ size 4945870360
checkpoint-16290/model-00003-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:04a2e8ebeffee8ec3e9214098c355ba296d24b72b73d9f87dae040db096458e8
3
+ size 2662825752
checkpoint-16290/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-16290/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5a0ef6f96a48e59aa52c4b471312c2a62378c19acc7ebbae839612b03a7d775a
3
+ size 15984
checkpoint-16290/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ab11d533c0fdad46ea8b8e295ba5fdb705e078eeb88cc28f37d82913508766e9
3
+ size 15984
checkpoint-16290/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9caa774d8768c038cd2939bf152fd06f67ec519371049f26f136110c4be6ab01
3
+ size 15997
checkpoint-16290/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:46e9db21361e903250f6d586c34c7d2a56fff72546ae48f50225bdce436c4219
3
+ size 15997
checkpoint-16290/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:602064667364a43234b9a45207a984492971522520dc806d75dff82f0cb9b17a
3
+ size 15997
checkpoint-16290/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f51f92af7e27985b6718b134d2d768c996204c15b1140f4d68c08aa4ebbdb132
3
+ size 15997
checkpoint-16290/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9701e5b58910efd4c6c0f70cd3b236f68c0596b578b98f4fd50bcd887d1db5bf
3
+ size 15997
checkpoint-16290/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d9a4d923b8738731745a5648e5403982e9354030a18b3cae82eb83d7efe2df36
3
+ size 15997
checkpoint-16290/rng_state_16.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e8b6b08b1e7aea716c2555a4cddd3364141b01fc76d316d4ec8d1818ea63d899
3
+ size 15997
checkpoint-16290/rng_state_17.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a61ab4170297b675a29b06c45fe759cf116492c4f84cc5c0ae12e8ec9952c500
3
+ size 15997
checkpoint-16290/rng_state_18.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c7f2a0cf3781aa55cf49ba437ae7ea7e8825383310efc2c2b67805c3753875de
3
+ size 15997
checkpoint-16290/rng_state_19.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dffd79573620ab6193b93350aa50412e9f485509e6422f2c179f77e45cb19d89
3
+ size 15997
checkpoint-16290/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:615c168147e3465ce5bfab6da2ff4afc68566ce00ec0f0c6c9fc988038a58d0a
3
+ size 15984
checkpoint-16290/rng_state_20.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e043da3051988e8b3046c7af8577da887019673f53e03a3b4d2c7b836c0b8ede
3
+ size 15997
checkpoint-16290/rng_state_21.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:829a4d53a8b01a1efee9fd4aff5eca4d049f66908a21785bb1dbfb223e3a5500
3
+ size 15997
checkpoint-16290/rng_state_22.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cb9857bfc4b4bd393b245536787fcc7c4385b2aa18a3c8fcac5e109eecb96dff
3
+ size 15997
checkpoint-16290/rng_state_23.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8a14affa54d74fe781c2643fef4b7c37998164a98cd90311cece93a7c4916954
3
+ size 15997
checkpoint-16290/rng_state_24.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b0e7f3847336458d35c84aefb88b7100275ac2619d6657afdf7645d43634ad46
3
+ size 15997
checkpoint-16290/rng_state_25.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5769fbf3965991e500d858b50307668ec889073469779912eb1f450c7848205d
3
+ size 15997
checkpoint-16290/rng_state_26.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:94a6eadb0c5204a1ea1480aa4ba919833a344b0c976ae58e8026ebca95ff9bca
3
+ size 15997
checkpoint-16290/rng_state_27.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:14b47f44655c527cc086d50908e4cd5125ec2df319786603499a0146847d4165
3
+ size 15997
checkpoint-16290/rng_state_28.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6fff46f787b4037d38bf09e5691a2101fac14bf13d40d349191dad3395be5b99
3
+ size 15997
checkpoint-16290/rng_state_29.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a6534a796189d9730e984bc43d45bc60aec5134a22696befd655312f79252959
3
+ size 15997
checkpoint-16290/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:79f71e8f8674ecaef9f8cdcbf7ac457a8b8ff15b12694ba2a2fffcb4b43f0f08
3
+ size 15984
checkpoint-16290/rng_state_30.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6a049bfcbe82f24c9cedc9279ab10620451ad6e052ead0360721e7ff8201223e
3
+ size 15997
checkpoint-16290/rng_state_31.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b3c21e2820256f7cf3034d5fdeaa02c6a516ea271c435278b4841e67a73ecf0e
3
+ size 15997
checkpoint-16290/rng_state_32.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dc1aa7fd4d9a19331909887d64465390c7ac8a8f91fadb89097681b989e82ee8
3
+ size 15997
checkpoint-16290/rng_state_33.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fa2f5338d608013932b2d5add89190ba4204c14b9b5a10baef0dbeec743ba4e0
3
+ size 15997
checkpoint-16290/rng_state_34.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:788a7c2ef18dc41377b5c3c917e2b69cb13c9dd1e653da51c5568671a6ae5c14
3
+ size 15997
checkpoint-16290/rng_state_35.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:144d78ca5a8b14119401cb9bcfb0faaaf3d91a2ef366c17e13e5e2402d2853d5
3
+ size 15997
checkpoint-16290/rng_state_36.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6d6fd596c2b53a54a1cd68ddbba780784791a99620d6a97fca72d3113b97cf5d
3
+ size 15997
checkpoint-16290/rng_state_37.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:89b8f62a00d4656dd87ea6159d719dc8d9cfb1e4e4f1dd69db905b2bfcd07093
3
+ size 15997
checkpoint-16290/rng_state_38.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:91a1dcc842ce6e269ce0e3c6ee7a4073bd6e00f9baadd58b716bb550b69d2691
3
+ size 15997
checkpoint-16290/rng_state_39.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ec3e19d750b7e2b8c03275440d6a4c7955b2941388923e35b4bf843e8987b6b6
3
+ size 15997
checkpoint-16290/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:88cf6d674dab5545c300a55135f08ca935730a3d35e2c419fb0b333f19482c19
3
+ size 15984
checkpoint-16290/rng_state_40.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5fd25c0b55c1f70994aefad6a1f65a13a189b964654444a2603e184e689620b5
3
+ size 15997
checkpoint-16290/rng_state_41.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7466699c1f0ebd5c55c1dcbab8f5da373ff4845cf6ed9250b3484ffb53945d63
3
+ size 15997
checkpoint-16290/rng_state_42.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2815b10d9414010f95880938ec06c60db688aad64da7384d646e3e4d3d909634
3
+ size 15997
checkpoint-16290/rng_state_43.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:82632e20f6f15c4e00a1d0a092ec5621c42e74f841a863bc42ffc6c5aa87db69
3
+ size 15997
checkpoint-16290/rng_state_44.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f65e7529d4d5a49a11a98eb931c28f370a87daf67300fd5f6564330ba193a8d9
3
+ size 15997
checkpoint-16290/rng_state_45.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b8ffabc904744cf73161feb472ff6d17b89f03b6c1443a75bd0d1349c08657b7
3
+ size 15997
checkpoint-16290/rng_state_46.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:05c1100879102745b9d87b746e54dd21b4b58798535d3fedce8565b8fa466287
3
+ size 15997
checkpoint-16290/rng_state_47.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c27665a3b0bf8c1cf4ef9272df93d9b7c2907c6cb1c4facff61424d01179387a
3
+ size 15997