Upload folder using huggingface_hub

config.json

@@ -1,29 +1,24 @@
 {
   "_attn_implementation_autoset": true,
-  "_name_or_path": "nguyenthanhthuan/Llama_3.2_1B_Intruct_Tool_Calling_V2",
+  "_name_or_path": "NousResearch/Hermes-3-Llama-3.2-3B",
   "architectures": [
     "LlamaForCausalLM"
   ],
   "attention_bias": false,
   "attention_dropout": 0.0,
   "bos_token_id": 128000,
-  "eos_token_id": [
-    128001,
-    128008,
-    128009
-  ],
-  "head_dim": 64,
+  "eos_token_id": 128039,
+  "head_dim": 128,
   "hidden_act": "silu",
-  "hidden_size": 2048,
+  "hidden_size": 3072,
   "initializer_range": 0.02,
   "intermediate_size": 8192,
   "max_position_embeddings": 131072,
   "mlp_bias": false,
   "model_type": "llama",
-  "num_attention_heads": 32,
-  "num_hidden_layers": 16,
+  "num_attention_heads": 24,
+  "num_hidden_layers": 28,
   "num_key_value_heads": 8,
-  "pad_token_id": 128004,
   "pretraining_tp": 1,
   "rms_norm_eps": 1e-05,
   "rope_scaling": {
@@ -36,7 +31,6 @@
   "rope_theta": 500000.0,
   "tie_word_embeddings": true,
   "transformers_version": "4.46.3",
-  "unsloth_version": "2024.10.7",
   "use_cache": true,
   "vocab_size": 128256
 }

generation_config.json

@@ -1,4 +1,5 @@
 {
+  "_from_model_config": true,
   "bos_token_id": 128000,
   "do_sample": true,
   "eos_token_id": [
@@ -6,8 +7,6 @@
     128008,
     128009
   ],
-  "max_length": 131072,
-  "pad_token_id": 128004,
   "temperature": 0.6,
   "top_p": 0.9,
   "transformers_version": "4.46.3"

onnx/model.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4be0a1c8367f0f08fa30f806a6703e403d88a8f864dbbb24adf920799a132dcf
-size 448765
+oid sha256:a5d41a9ea95192f86edd67138737ce7ad0fc0d86853a6d5facce1f0ed10041c6
+size 776375

onnx/model.onnx_data

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c4f269b9750ade29e17b6daec21c5d76bf2f0c1206b9e8e1090a2ad2d3b1ef95
-size 4943257600
+oid sha256:5e32108d9e31e081988686862455045e14bb58e868ec76e552a9edcd8367996b
+size 12850999296

onnx/model_bnb4.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:973e5064cbd1cffa31cd30c9a1371cc648af2831391fdbf7768cd9419e7d1c00
-size 1598757277
+oid sha256:3028fefc768fbea550846e7e26e7a9d2f07801158f0e7c43d4a5591bb63b6347
+size 822344

onnx/model_bnb4.onnx_data

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fd39711b03ac0b3460bbb30b0f0cec6e8ef82acf58f88bb2110c851fb4c8ecd8
-size 6323781632
+oid sha256:7ae8301ff4d0d8ca41d75596769753f408add40db29558f0c4f87024b0f75d3e
+size 3162157056

onnx/model_fp16.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7a2f337289d0805d9807d78c4194f2dbc1d3dfb4d38e1447f8134d76284b3711
-size 457645
+oid sha256:602f3d0ed7296cfc7c3ea4e1a0e3e3ba1eb0b8cbada03e1519aecdaaa1f324d3
+size 791765

onnx/model_fp16.onnx_data

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c9733aab730ec38ed1ed2b8e85ba96773627c212e41c4faec2771f3163968ad1
-size 2471628800
+oid sha256:e5b07b6bdd0967f15120ecb05ecb75ce1b7ef7bb9f39b66f0c315688340810e9
+size 6425499648

onnx/model_int8.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a392606574450b062054554547f0b8d9c8b77b1678417107e37d007e31acc5b3
-size 1236578925
+oid sha256:c1a0ce71408e06e3d2395ad382b45a961508a99fcb0e4d6f4af910c0e1ef975d
+size 991648

onnx/model_int8.onnx_data

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:89ee296e92fdd6fb0cdcc3d13f43a24d7a1a28c37d546a7c98392c013b4e186b
-size 8031059968
+oid sha256:657f76d3174d98bdec5b911858a5cade2863a345eb391bd97b79dd6e8deb2e84
+size 3213275136

onnx/model_q4.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5aa75b94996669549ce465cbe97688b269e438d4986e665695a7e98a89927a74
-size 1659573821
+oid sha256:80b807bdfa953bb48484bdcd13fd917f5ef9fbdb4b511c3037c18aac71edd5f8
+size 819792

onnx/model_q4.onnx_data

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:460a97e29c395098e40630ae977e9a8f1158e1b474c9209e4ad0e46334afa2a4
-size 6792822784
+oid sha256:96e6b616d56af52e3014c1bf52b870b5fd5fb32f6df66614eb1eb4a0401eb949
+size 3338317824

onnx/model_q4f16.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2f678c74e7eab4dac069eb38d0575cef3571205c12531fee6cd3d30341735a54
-size 1073292786
+oid sha256:9653d2bb65a19201c6b3aa760cc26cd748179d641d95b08bde9817876ac204d7
+size 835365

onnx/model_q4f16.onnx_data

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:73694682378ff20d4d65c680ee98b5415a0f74d96d3437db644b702ff05833c9
-size 5272576000
+oid sha256:55cb782dce22fcccbd96d2dfa2905246f79412edfc99ebd9d774ac442e0c4f29
+size 2373801984

onnx/model_quantized.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a392606574450b062054554547f0b8d9c8b77b1678417107e37d007e31acc5b3
-size 1236578925
+oid sha256:9cc3735938054c404db34b6281455df06be1eef9574a3abf8093a3402dd2cd73
+size 992946

onnx/model_quantized.onnx_data

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:89ee296e92fdd6fb0cdcc3d13f43a24d7a1a28c37d546a7c98392c013b4e186b
-size 8031059968
+oid sha256:657f76d3174d98bdec5b911858a5cade2863a345eb391bd97b79dd6e8deb2e84
+size 3213275136

onnx/model_uint8.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:607c40015dceb3d5970af129484bce2f8dba78f3ad10881ea4586d20a850cfc6
-size 1236578987
+oid sha256:811d5dd6980f7fef1f903a9ba8eff4459a10e2608fc3f36824bea6debf1d91f5
+size 991999

onnx/model_uint8.onnx_data

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d881b50aa33ca4d9b531031029fc3b40c677033299addbf58f5ee1c468c85e43
-size 8031059968
+oid sha256:4e7b640fee0a83fb4cc57cd0a3c6908fc934e336d1f5f8b5fe6dea70da7a0c15
+size 3213275136

special_tokens_map.json

@@ -7,14 +7,14 @@
     "single_word": false
   },
   "eos_token": {
-    "content": "<|eot_id|>",
+    "content": "<|im_end|>",
     "lstrip": false,
     "normalized": false,
     "rstrip": false,
     "single_word": false
   },
   "pad_token": {
-    "content": "<|finetune_right_pad_id|>",
+    "content": "<|end_of_text|>",
     "lstrip": false,
     "normalized": false,
     "rstrip": false,

tokenizer.json

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6b9e4e7fb171f92fd137b777cc2714bf87d11576700a1dcd7a399e7bbe39537b
-size 17209920
+oid sha256:9f908f9b84390fd12c6d0c356765257846c53f60bf472ff4996a440a1e230373
+size 17209403

tokenizer_config.json

@@ -17,20 +17,20 @@
       "special": true
     },
     "128002": {
-      "content": "<|reserved_special_token_0|>",
+      "content": "<tool_call>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128003": {
-      "content": "<|reserved_special_token_1|>",
+      "content": "<tool_response>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128004": {
       "content": "<|finetune_right_pad_id|>",
@@ -89,231 +89,231 @@
       "special": true
     },
     "128011": {
-      "content": "<|reserved_special_token_3|>",
+      "content": "<tools>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128012": {
-      "content": "<|reserved_special_token_4|>",
+      "content": "</tools>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128013": {
-      "content": "<|reserved_special_token_5|>",
+      "content": "</tool_call>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128014": {
-      "content": "<|reserved_special_token_6|>",
+      "content": "</tool_response>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128015": {
-      "content": "<|reserved_special_token_7|>",
+      "content": "<schema>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128016": {
-      "content": "<|reserved_special_token_8|>",
+      "content": "</schema>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128017": {
-      "content": "<|reserved_special_token_9|>",
+      "content": "<scratch_pad>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128018": {
-      "content": "<|reserved_special_token_10|>",
+      "content": "</scratch_pad>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128019": {
-      "content": "<|reserved_special_token_11|>",
+      "content": "<SCRATCHPAD>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128020": {
-      "content": "<|reserved_special_token_12|>",
+      "content": "</SCRATCHPAD>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128021": {
-      "content": "<|reserved_special_token_13|>",
+      "content": "<REASONING>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128022": {
-      "content": "<|reserved_special_token_14|>",
+      "content": "</REASONING>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128023": {
-      "content": "<|reserved_special_token_15|>",
+      "content": "<INNER_MONOLOGUE>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128024": {
-      "content": "<|reserved_special_token_16|>",
+      "content": "</INNER_MONOLOGUE>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128025": {
-      "content": "<|reserved_special_token_17|>",
+      "content": "<PLAN>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128026": {
-      "content": "<|reserved_special_token_18|>",
+      "content": "</PLAN>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128027": {
-      "content": "<|reserved_special_token_19|>",
+      "content": "<EXECUTION>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128028": {
-      "content": "<|reserved_special_token_20|>",
+      "content": "</EXECUTION>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128029": {
-      "content": "<|reserved_special_token_21|>",
+      "content": "<REFLECTION>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128030": {
-      "content": "<|reserved_special_token_22|>",
+      "content": "</REFLECTION>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128031": {
-      "content": "<|reserved_special_token_23|>",
+      "content": "<THINKING>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128032": {
-      "content": "<|reserved_special_token_24|>",
+      "content": "</THINKING>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128033": {
-      "content": "<|reserved_special_token_25|>",
+      "content": "<SOLUTION>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128034": {
-      "content": "<|reserved_special_token_26|>",
+      "content": "</SOLUTION>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128035": {
-      "content": "<|reserved_special_token_27|>",
+      "content": "<EXPLANATION>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128036": {
-      "content": "<|reserved_special_token_28|>",
+      "content": "</EXPLANATION>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128037": {
-      "content": "<|reserved_special_token_29|>",
+      "content": "<UNIT_TEST>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128038": {
-      "content": "<|reserved_special_token_30|>",
+      "content": "</UNIT_TEST>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128039": {
-      "content": "<|reserved_special_token_31|>",
+      "content": "<|im_end|>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
@@ -321,12 +321,12 @@
       "special": true
     },
     "128040": {
-      "content": "<|reserved_special_token_32|>",
+      "content": "<|im_start|>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
-      "special": true
+      "special": false
     },
     "128041": {
       "content": "<|reserved_special_token_33|>",
@@ -2050,15 +2050,14 @@
     }
   },
   "bos_token": "<|begin_of_text|>",
-  "chat_template": "\n{%- if messages[0]['role'] == 'system' -%}\n<|start_header_id|>system<|end_header_id|>\n\n{{ messages[0]['content'] }}\n{%- endif -%}\n{%- for message in messages -%}\n    {%- if message['role'] == 'user' -%}\n        <|start_header_id|>user<|end_header_id|>\n        {{ message['content'] }}\n    {%- elif message['role'] in ['assistant'] -%}\n        <|start_header_id|>assistant<|end_header_id|>\n        {{ message['content'] }}<|eot_id|>\n    {%- elif message['role'] in ['tool_call'] -%}\n        <|start_header_id|>tool_call<|end_header_id|>\n        {{ message['content'] }}<|eot_id|>\n    {%- elif message['role'] == 'tool_response' -%}\n        <|start_header_id|>tool_respone<|end_header_id|>\n        {{ message['content'] }}\n    {%- endif -%}\n{%- endfor -%}\n{%- if add_generation_prompt -%} <|start_header_id|>assistant<|end_header_id|>\n{%- endif -%}\n",
+  "chat_template": "{% if not add_generation_prompt is defined %}{% set add_generation_prompt = false %}{% endif %}{% for message in messages %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\n' }}{% endif %}",
   "clean_up_tokenization_spaces": true,
-  "eos_token": "<|eot_id|>",
+  "eos_token": "<|im_end|>",
   "model_input_names": [
     "input_ids",
     "attention_mask"
   ],
   "model_max_length": 131072,
-  "pad_token": "<|finetune_right_pad_id|>",
-  "padding_side": "left",
+  "pad_token": "<|end_of_text|>",
   "tokenizer_class": "PreTrainedTokenizerFast"
 }