NickyNicky/Mix_TinyLlama-3x1B_oasst2_chatML_Cluster_3_2_1_V1

Model Card for Model MixLlama

experts:
  - source_model: NickyNicky/TinyDolphin-2.8-1.1b_oasst2_chatML_Cluster_1_V1
    positive_prompts:
      - ""

  - source_model: NickyNicky/TinyDolphin-2.8-1.1b_oasst2_chatML_Cluster_2_V1
    positive_prompts:
      - ""

  - source_model: NickyNicky/TinyDolphin-2.8-1.1b_oasst2_chatML_Cluster_3_V1
    positive_prompts:
      - ""

base_model: NickyNicky/TinyDolphin-2.8-1.1b_oasst2_chatML_Cluster_1_V1
gate_mode: random # one of "hidden", "cheap_embed", or "random"
dtype: bfloat16 # output dtype (float32, float16, or bfloat16)

from transformers import (
    AutoModelForCausalLM,
    AutoTokenizer,
    BitsAndBytesConfig,
    HfArgumentParser,
    TrainingArguments,
    pipeline,
    logging,
    GenerationConfig,
    TextIteratorStreamer,
)
import torch

new_model= "NickyNicky/Mix_TinyLlama-3x1B_oasst2_chatML_Cluster_3_2_1_V1"
model = AutoModelForCausalLM.from_pretrained(#f'NickyNicky/{new_model}',
                                             new_model,
                                             device_map="auto",
                                             trust_remote_code=True,
                                             torch_dtype=torch.bfloat16,

                                             low_cpu_mem_usage= True,
                                            #  use_flash_attention_2=False,

                                             )


tokenizer = AutoTokenizer.from_pretrained(new_model,
                                          max_length=2048,
                                          trust_remote_code=True,
                                          use_fast = True,
                                          )

tokenizer.pad_token = tokenizer.eos_token
# tokenizer.padding_side = 'left'
tokenizer.padding_side = 'right'


prompt= """<|im_start|>system
You are a helpful AI assistant.<|im_end|>
<|im_start|>user
escribe una historia de amor.<|im_end|>
<|im_start|>assistant
"""

inputs = tokenizer.encode(prompt,
                          return_tensors="pt",
                          add_special_tokens=False).cuda()#.to("cuda") # False # True


generation_config = GenerationConfig(
              max_new_tokens=700,
              temperature=0.5,
              top_p=0.9,
              top_k=40,
              repetition_penalty=1.1, #1.1, # 1.0 means no penalty, > 1.0 means penalty, 1.2 from CTRL paper
              do_sample=True,
              pad_token_id=tokenizer.eos_token_id,
              eos_token_id=tokenizer.eos_token_id,
          )
outputs = model.generate(
                         generation_config=generation_config,
                         input_ids=inputs,)
# tokenizer.decode(outputs[0], skip_special_tokens=False) #True
print(tokenizer.decode(outputs[0], skip_special_tokens=False))