Upload 7 files

README.md

@@ -29,8 +29,17 @@ model_type: MistralForCausalLM
 tokenizer_type: LlamaTokenizer
 is_mistral_derived_model: true
 
+# git clone https://github.com/OpenAccess-AI-Collective/axolotl
+# cd axolotl
+
+# pip3 install packaging
+# pip3 install -e '.[flash-attn,deepspeed]'
+
 # accelerate launch -m axolotl.cli.train ./llama_7b_config.yaml
 
+# accelerate launch -m axolotl.cli.inference ./llama_7b_config.yaml \
+#     --lora_model_dir="dohonba/mistral_7b_fingpt"
+
 load_in_8bit: true
 load_in_4bit: false
 strict: false
@@ -49,9 +58,6 @@ dataset_prepared_path:
 val_set_size: 0.05
 output_dir: ./lora-out
 
-chat_template: chatml
-default_system_message: You are a helpful assistant, specialising in financial text sentiment and emotional analysis. # Currently only supports chatml.
-
 sequence_len: 512
 sample_packing: true
 pad_to_sequence_len: true
@@ -71,7 +77,7 @@ wandb_name:
 wandb_log_model:
 
 gradient_accumulation_steps: 1
-micro_batch_size: 16
+micro_batch_size: 14
 # max_steps: 1000
 num_epochs: 2
 optimizer: adamw_bnb_8bit
@@ -94,7 +100,7 @@ flash_attention: true
 s2_attention:
 
 warmup_steps: 50
-evals_per_epoch: 2
+evals_per_epoch: 0
 eval_table_size:
 eval_table_max_new_tokens: 128
 saves_per_epoch: 1
@@ -115,7 +121,7 @@ special_tokens:
 
 This model is a fine-tuned version of [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) on the None dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.0942
+- Loss: 0.0917
 
 ## Model description
 
@@ -135,8 +141,8 @@ More information needed
 
 The following hyperparameters were used during training:
 - learning_rate: 0.0002
-- train_batch_size: 16
-- eval_batch_size: 16
+- train_batch_size: 14
+- eval_batch_size: 14
 - seed: 42
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: cosine
@@ -147,11 +153,8 @@ The following hyperparameters were used during training:
 
 | Training Loss | Epoch | Step | Validation Loss |
 |:-------------:|:-----:|:----:|:---------------:|
-| 3.8894        | 0.0   | 1    | 3.9316          |
-| 0.1106        | 0.5   | 247  | 0.1150          |
-| 0.0956        | 1.0   | 494  | 0.0998          |
-| 0.0705        | 1.48  | 741  | 0.0990          |
-| 0.0858        | 1.98  | 988  | 0.0942          |
+| 0.08          | 1.02  | 566  | 0.0986          |
+| 0.0919        | 1.98  | 1110 | 0.0917          |
 
 
 ### Framework versions

adapter_config.json

@@ -19,13 +19,13 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
+    "v_proj",
     "up_proj",
-    "q_proj",
     "down_proj",
-    "v_proj",
-    "o_proj",
     "gate_proj",
-    "k_proj"
+    "k_proj",
+    "o_proj",
+    "q_proj"
   ],
   "task_type": "CAUSAL_LM"
 }

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fc7b054e6913ab8281151c4ba3ee1016c6930c443a8ea9aae2f778ae60f47ea9
+oid sha256:ed44fe38ab406e5b8a358047c9754533cfe79a547f5ae7d159b73dc5b5ba2319
 size 335705741

tokenizer_config.json

@@ -29,7 +29,6 @@
   },
   "additional_special_tokens": [],
   "bos_token": "<s>",
-  "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 = 'You are a helpful assistant, specialising in financial text sentiment and emotional analysis.' %}{% 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 %}",
   "clean_up_tokenization_spaces": false,
   "eos_token": "</s>",
   "legacy": true,