Update configs.py
Browse files- configs.py +7 -1
configs.py
CHANGED
|
@@ -27,12 +27,18 @@ model_info = {
|
|
| 27 |
original_prompt_template='<s>{prompt}',
|
| 28 |
interpretation_prompt_template='<s>[INST] [X] [/INST] {prompt}',
|
| 29 |
layers_format=llama_layers_format),
|
| 30 |
-
'Mixtral 8x7B Instruct': dict(model_path='TheBloke/Mixtral-8x7B-Instruct-v0.1-AWQ',
|
| 31 |
token=os.environ['hf_token'], wait_with_hidden_states=True,
|
| 32 |
original_prompt_template='<s>{prompt}',
|
| 33 |
interpretation_prompt_template='<s>[INST] [X] [/INST] {prompt}',
|
| 34 |
layers_format=llama_layers_format
|
| 35 |
),
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 36 |
'GPT-2 Small': dict(model_path='gpt2', original_prompt_template='{prompt}',
|
| 37 |
interpretation_prompt_template='User: [X]\n\nAnswer: {prompt}',
|
| 38 |
layers_format=gpt_layers_format),
|
|
|
|
| 27 |
original_prompt_template='<s>{prompt}',
|
| 28 |
interpretation_prompt_template='<s>[INST] [X] [/INST] {prompt}',
|
| 29 |
layers_format=llama_layers_format),
|
| 30 |
+
'Mixtral 8x7B Instruct (Experimental)': dict(model_path='TheBloke/Mixtral-8x7B-Instruct-v0.1-AWQ',
|
| 31 |
token=os.environ['hf_token'], wait_with_hidden_states=True,
|
| 32 |
original_prompt_template='<s>{prompt}',
|
| 33 |
interpretation_prompt_template='<s>[INST] [X] [/INST] {prompt}',
|
| 34 |
layers_format=llama_layers_format
|
| 35 |
),
|
| 36 |
+
'CodeLLAMA 70B Instruct': dict(model_path='TheBloke/CodeLlama-70B-Instruct-GPTQ',
|
| 37 |
+
token=os.environ['hf_token'], wait_with_hidden_states=True,
|
| 38 |
+
original_prompt_template='<s>{prompt}',
|
| 39 |
+
interpretation_prompt_template='<s>[INST] [X] [/INST] {prompt}',
|
| 40 |
+
layers_format=llama_layers_format
|
| 41 |
+
),
|
| 42 |
'GPT-2 Small': dict(model_path='gpt2', original_prompt_template='{prompt}',
|
| 43 |
interpretation_prompt_template='User: [X]\n\nAnswer: {prompt}',
|
| 44 |
layers_format=gpt_layers_format),
|