tmfi-us/Legion-V2.1-LLaMa-70B-FP8-Dynamic

About

FP8 w8a8 dynamic quants of Tarek07/Legion-V2.1-LLaMa-70B.

Used following Python script with llmcompressor to generate:

from transformers import AutoTokenizer, AutoModelForCausalLM
from llmcompressor.transformers import oneshot
from llmcompressor.modifiers.quantization import QuantizationModifier

MODEL_ID = 'Tarek07/Legion-V2.1-LLaMa-70B'

model = AutoModelForCausalLM.from_pretrained(
  MODEL_ID, device_map="auto", torch_dtype="auto",
)
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)

# Configure the simple PTQ quantization
recipe = QuantizationModifier(
  targets="Linear", scheme="FP8_DYNAMIC", ignore=["lm_head"])

# Apply the quantization algorithm.
oneshot(model=model, recipe=recipe)

# Save the model.
SAVE_DIR = MODEL_ID.split("/")[1] + "-FP8-Dynamic"
model.save_pretrained(SAVE_DIR)
tokenizer.save_pretrained(SAVE_DIR)

Quantization recipe can be found in recipe.yaml