Update README.md

README.md

@@ -6,15 +6,17 @@ license_link: LICENSE
 
 ## Model description
 
-Yi-34B model fine-tuned on AEZAKMI v1 dataset that is derived from airoboros 2.2.1 and airoboros 2.2. Finetuned with axolotl, using qlora and nf4 double quant, 1 epoch, batch size 1, lr 0.00007, lr scheduler constant, sequence length 1200. Training took around 33 hours on single local RTX 3090 Ti.
-I had power target set to 320W for the GPU, and while I didn't measure power at the wall, it was probably something around 500W. Given the average electricity price in my region, this training run cost me around $3. This was my first attempt at training Yi-34B with this dataset.
-Main feature of this model is that it's output should be free of refusals and it feels somehow more natural than airoboros. Prompt format is standard chatml. Don't expect it to be good at math, riddles or be crazy smart. My end goal with AEZAKMI is to create a cozy free chatbot.
+Yi-34B base model fine-tuned on AEZAKMI v1 dataset. Training took around 33 hours on single local RTX 3090 Ti.
+It's like airoboros but with less gptslop, no refusals and less typical language used by RLHFed OpenAI models.
+Say goodbye to  "It's important to remember"! \
+Prompt format is standard chatml. Don't expect it to be good at math, riddles or be crazy smart. My end goal with AEZAKMI is to create a cozy free chatbot.
+Cost of this fine-tune is about $3 in electricity. This was my first attempt at training Yi-34B with this dataset.
+Base model used for fine-tuning was 4k context Yi-34B-Llama model shared by chargoddard.
 
-I used 4096 ctx Yi-34B-Llama uploaded by chargoddard as a base for this training.
 
 ## Prompt Format
 
-I recommend using ChatML format, as this was used during fine-tune
+I recommend using ChatML format, as this was used during fine-tune. \
 Here's a prompt format you should use, you can set a different system message, model seems to respect that fine, so it wasn't overfitted.
 
 ```
@@ -27,15 +29,52 @@ A chat.<|im_end|>
 
 ## Intended uses & limitations
 
-Use is limited by Yi license
+Use is limited by Yi license.
 
 ## Known Issues
 
 I recommend to set repetition penalty to something around 1.05 to avoid repetition. So far I had good experience running this model with temperature 1.2.
-Multi-turn conversations could be a bit better, if you ask it to re-write something with some fixes it will have a tendency to just repeat the previous response verbatim without any improvements - this is especially noticeable with repp 1.0
+Multi-turn conversations could be a bit better, if you ask it to re-write something with some fixes it will have a tendency to just repeat the previous response verbatim without any improvements - this is especially noticeable with repp 1.0.
 There is still some gptslop left - some responses will have last paragraph with text "Remember that bla bla bla", I will try to get rid of it in the next version of the dataset.
 Stories have ChatGPT like paragraph spacing, I will try to introduce a bit more stories that have long paragraphs in the next dataset version.
 
+## Axolotl training parameters
+
+- bnb_4bit_use_double_quant: true
+- bnb_4bit_compute_dtype: torch.bfloat16
+- is_llama_derived_model: true
+- load_in_4bit: true
+- adapter: qlora
+- sequence_len: 1200
+- sample_packing: false
+- lora_r: 16
+- lora_alpha: 32
+- lora_target_modules:
+  - q_proj
+  - v_proj
+  - k_proj
+  - o_proj
+  - gate_proj
+  - down_proj
+  - up_proj
+ - lora_target_linear: true
+ - pad_to_sequence_len: true
+ - micro_batch_size: 1
+ - gradient_accumulation_steps: 1
+ - num_epochs: 1
+ - optimizer: adamw_bnb_8bit
+ - lr_scheduler: constant
+ - learning_rate: 0.00007
+ - train_on_inputs: false
+ - group_by_length: false
+ - bf16: true
+ - bfloat16: true
+ - flash_optimum: false
+ - gradient_checkpointing: true
+ - flash_attention: true
+ - seed: 42
+
+
 ## Upcoming
 
 I will release adapter files and maybe exllama v2 quant shortly.