Training in progress, step 50

axolotl_config.yaml

@@ -1,5 +1,5 @@
-base_model: trl-internal-testing/tiny-random-LlamaForCausalLM
-batch_size: 256
+base_model: unsloth/OpenHermes-2.5-Mistral-7B
+batch_size: 32
 bf16: true
 chat_template: tokenizer_default_fallback_alpaca
 datasets:
@@ -15,7 +15,7 @@ datasets:
     system_prompt: ''
 device_map: auto
 eval_sample_packing: false
-eval_steps: 200
+eval_steps: 100
 flash_attention: true
 gradient_checkpointing: true
 group_by_length: true
@@ -24,8 +24,8 @@ hub_strategy: checkpoint
 learning_rate: 0.0002
 logging_steps: 10
 lr_scheduler: cosine
-max_steps: 2500
-micro_batch_size: 32
+max_steps: 10000
+micro_batch_size: 2
 model_type: AutoModelForCausalLM
 num_epochs: 100
 optimizer: adamw_bnb_8bit
@@ -33,7 +33,7 @@ output_dir: /root/.sn56/axolotl/tmp/test-repo
 pad_to_sequence_len: true
 resize_token_embeddings_to_32x: false
 sample_packing: true
-save_steps: 40
+save_steps: 50
 save_total_limit: 2
 sequence_len: 2048
 tokenizer_type: LlamaTokenizerFast
@@ -42,7 +42,7 @@ trust_remote_code: true
 val_set_size: 0.1
 wandb_entity: ''
 wandb_mode: online
-wandb_name: trl-internal-testing/tiny-random-LlamaForCausalLM-argilla/databricks-dolly-15k-curated-en
+wandb_name: unsloth/OpenHermes-2.5-Mistral-7B-argilla/databricks-dolly-15k-curated-en
 wandb_project: Gradients-On-Demand
 wandb_run: your_name
 wandb_runid: default

config.json

@@ -1,31 +1,28 @@
 {
-  "_name_or_path": "trl-internal-testing/tiny-random-LlamaForCausalLM",
+  "_name_or_path": "unsloth/OpenHermes-2.5-Mistral-7B",
   "architectures": [
-    "LlamaForCausalLM"
+    "MistralForCausalLM"
   ],
-  "attention_bias": false,
   "attention_dropout": 0.0,
-  "bos_token_id": 0,
-  "eos_token_id": 2,
-  "head_dim": 4,
+  "bos_token_id": 1,
+  "eos_token_id": 32000,
+  "head_dim": 128,
   "hidden_act": "silu",
-  "hidden_size": 16,
+  "hidden_size": 4096,
   "initializer_range": 0.02,
-  "intermediate_size": 64,
-  "max_position_embeddings": 2048,
-  "mlp_bias": false,
-  "model_type": "llama",
-  "num_attention_heads": 4,
-  "num_hidden_layers": 2,
-  "num_key_value_heads": 4,
-  "pad_token_id": -1,
-  "pretraining_tp": 1,
-  "rms_norm_eps": 1e-06,
-  "rope_scaling": null,
+  "intermediate_size": 14336,
+  "max_position_embeddings": 32768,
+  "model_type": "mistral",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "num_key_value_heads": 8,
+  "pad_token_id": 0,
+  "rms_norm_eps": 1e-05,
   "rope_theta": 10000.0,
+  "sliding_window": 4096,
   "tie_word_embeddings": false,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.48.1",
   "use_cache": false,
-  "vocab_size": 32000
+  "vocab_size": 32002
 }

model-00001-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:720105d2b88abf15d23994fec61854b0a1497a10d1d208715bd59904ab5936ff
+oid sha256:d9e33d86b7643ca1dcb0ff586538eca0faa1c5ceb646f55ab91b4c3f84b9791f
 size 4943178720

model-00002-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:22f56733ac237ee5fada526e2a06e89d311e7ce1cb761f9b8f6986910b678378
+oid sha256:09671ff7f4231f95a58974cee08101fdbd61e0e44a2da6afaa3f08be535bc56c
 size 4999819336

model-00003-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:31711beccc0b973ef7bf69d348d844233b185f6888f0520ae136ada35d74f983
+oid sha256:867e8292011b488643369bc98aece5a8d9682c4d0825739096988e70f10694e1
 size 4540532728

special_tokens_map.json

@@ -7,7 +7,7 @@
     "single_word": false
   },
   "eos_token": {
-    "content": "</s>",
+    "content": "<|im_end|>",
     "lstrip": false,
     "normalized": false,
     "rstrip": false,

tokenizer.json

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d606f46a8aa6f29f0a0abdec7c3ffddefc9f9bfe26919532d209a0a850e25029
-size 3619013
+oid sha256:04222cd76979c181cd3f72c3bf6982fe2a09d9f4b8f23d82902efde18f1d0668
+size 3506125

tokenizer.model

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
-size 499723
+oid sha256:dadfd56d766715c61d2ef780a525ab43b8e6da4de6865bda3d95fdef5e134055
+size 493443

tokenizer_config.json

@@ -1,7 +1,7 @@
 {
   "add_bos_token": true,
   "add_eos_token": false,
-  "add_prefix_space": null,
+  "add_prefix_space": true,
   "added_tokens_decoder": {
     "0": {
       "content": "<unk>",
@@ -26,17 +26,38 @@
       "rstrip": false,
       "single_word": false,
       "special": true
+    },
+    "32000": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "32001": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
     }
   },
+  "additional_special_tokens": [],
   "bos_token": "<s>",
-  "chat_template": "{% for message in messages %}{% if message['role'] == 'user' %}{{ '### Instruction: ' + message['content'] + '\n\n' }}{% elif message['role'] == 'assistant' %}{{ '### Response: ' + message['content'] + eos_token}}{% endif %}{% endfor %}",
+  "chat_template": "{% 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": false,
-  "eos_token": "</s>",
+  "eos_token": "<|im_end|>",
   "extra_special_tokens": {},
   "legacy": true,
-  "model_max_length": 2048,
+  "model_max_length": 32768,
   "pad_token": "<unk>",
+  "padding_side": "right",
+  "sp_model_kwargs": {},
+  "spaces_between_special_tokens": false,
   "tokenizer_class": "LlamaTokenizer",
+  "trust_remote_code": false,
   "unk_token": "<unk>",
   "use_default_system_prompt": true,
   "use_fast": true

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c5805d33db3bd7bbd9181584d8158e2b8616e3f18d56ec44b12ac78d4f859a79
-size 6840
+oid sha256:c7626f1764e5176ea7bd748263d627f8cb6423b299d8a596d8f39c9c0a75e460
+size 6776