v4: ChatML -> LLama3 template

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:23d40b2624d779257ec389b441c65d7ba5365bb86c0b6d5d3341cbf5dbc2cb25
+oid sha256:d8c5b217a8ccef05a066f359ae73341bcff727022c428578e8ef7f85f9c56f7d
 size 4976698672

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5d773178b71d3c0e544ad4a971e754a5a9eb6f884e882e80cdf9497287fc4ff8
+oid sha256:4115f9788f398b4ab09e463c9265add7f37ce8fd985fc809280dad614b71ae2e
 size 4999802720

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:47c2fbf539155e5796e525249ed99744f8f1453970313e6078b6becf20aa085a
+oid sha256:b2e4d243f89d797559130d0a377bfa97ea278711074e5113044a8f8da8f2503e
 size 4915916176

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:47d7b7ef31a88c44cc3fd39c43f91b4fe87a7a492f1e3ecee72dbefa4e479f49
+oid sha256:6c91b732fd6453ccc08c7de471705fabbfbb0b69801bff9b2d96979a44db89a1
 size 1168138808

special_tokens_map.json

@@ -1,24 +1,20 @@
 {
-  "additional_special_tokens": [
-    "<|im_start|>",
-    "<|im_end|>"
-  ],
   "bos_token": {
-    "content": "<|im_start|>",
+    "content": "<|begin_of_text|>",
     "lstrip": false,
     "normalized": false,
     "rstrip": false,
     "single_word": false
   },
   "eos_token": {
-    "content": "<|im_end|>",
+    "content": "<|eot_id|>",
     "lstrip": false,
     "normalized": false,
     "rstrip": false,
     "single_word": false
   },
   "pad_token": {
-    "content": "<|im_start|>",
+    "content": "<|begin_of_text|>",
     "lstrip": false,
     "normalized": false,
     "rstrip": false,

tokenizer.json

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8e7cad82642921f006bd0a63b22c613ccbfcf663e81a42da76a476e69bafc8df
-size 9084486
+oid sha256:e134af98b985517b4f068e3755ae90d4e9cd2d45d328325dc503f1c6b2d06cc7
+size 9085698

tokenizer_config.json

@@ -1,7 +1,7 @@
 {
   "added_tokens_decoder": {
     "128000": {
-      "content": "<|im_start|>",
+      "content": "<|begin_of_text|>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
@@ -73,7 +73,7 @@
       "special": true
     },
     "128009": {
-      "content": "<|im_end|>",
+      "content": "<|eot_id|>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
@@ -2049,20 +2049,16 @@
       "special": true
     }
   },
-  "additional_special_tokens": [
-    "<|im_start|>",
-    "<|im_end|>"
-  ],
-  "bos_token": "<|im_start|>",
-  "chat_template": "{% if messages[0]['role'] == 'system' %}{% set loop_messages = messages[1:] %}{% set system_message = messages[0]['content'] %}{% else %}{% set loop_messages = messages %}{% set system_message = 'Ты — Сайга, русскоязычный автоматический ассистент. Ты разговариваешь с людьми и помогаешь им.' %}{% endif %}{% if not add_generation_prompt is defined %}{% set add_generation_prompt = false %}{% endif %}{% for message in loop_messages %}{% if loop.index0 == 0 %}{{'<|im_start|>system\n' + system_message + '<|im_end|>\n'}}{% endif %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\n' }}{% endif %}",
+  "bos_token": "<|begin_of_text|>",
+  "chat_template": "{% set loop_messages = messages %}{% for message in loop_messages %}{% set content = '<|start_header_id|>' + message['role'] + '<|end_header_id|>\n\n'+ message['content'] | trim + '<|eot_id|>' %}{% if loop.index0 == 0 %}{% set content = bos_token + content %}{% endif %}{{ content }}{% endfor %}{% if add_generation_prompt %}{{ '<|start_header_id|>assistant<|end_header_id|>\n\n' }}{% endif %}",
   "clean_up_tokenization_spaces": true,
-  "eos_token": "<|im_end|>",
+  "eos_token": "<|eot_id|>",
   "model_input_names": [
     "input_ids",
     "attention_mask"
   ],
   "model_max_length": 8192,
-  "pad_token": "<|im_start|>",
+  "pad_token": "<|begin_of_text|>",
   "padding_side": "left",
   "tokenizer_class": "PreTrainedTokenizerFast"
 }