TriLMs unpacked to FP16 - compatible with any implementation supporting LLaMa architecture in huggingface's transformers format.
AI & ML interests
None defined yet.
QuantLMs, unpacked to FP16 format - compatible with FP16 GEMMs. After unpacking, QuantLMs have the same architecture as LLaMa.
-
SpectraSuite/QuantLM_3.9B_8bit_Unpacked
Text Generation • 4B • Updated • 6 -
SpectraSuite/QuantLM_2.3B_8bit_Unpacked
Text Generation • 2B • Updated • 17 -
SpectraSuite/QuantLM_1.5B_8bit_Unpacked
Text Generation • 2B • Updated • 2 -
SpectraSuite/QuantLM_1.1B_8bit_Unpacked
Text Generation • 1B • Updated • 5
QuantLMs, unpacked to FP16 format - compatible with FP16 GEMMs. After unpacking, QuantLMs have the same architecture as LLaMa.
FP16 LLMs with LLaMa architecture, trained on same 300B tokens as TriLMs.
QuantLMs, unpacked to FP16 format - compatible with FP16 GEMMs. After unpacking, QuantLMs have the same architecture as LLaMa.
QuantLMs, unpacked to FP16 format - compatible with FP16 GEMMs. After unpacking, QuantLMs have the same architecture as LLaMa.
TriLMs unpacked to FP16 - compatible with any implementation supporting LLaMa architecture in huggingface's transformers format.
FP16 LLMs with LLaMa architecture, trained on same 300B tokens as TriLMs.
QuantLMs, unpacked to FP16 format - compatible with FP16 GEMMs. After unpacking, QuantLMs have the same architecture as LLaMa.
-
SpectraSuite/QuantLM_3.9B_8bit_Unpacked
Text Generation • 4B • Updated • 6 -
SpectraSuite/QuantLM_2.3B_8bit_Unpacked
Text Generation • 2B • Updated • 17 -
SpectraSuite/QuantLM_1.5B_8bit_Unpacked
Text Generation • 2B • Updated • 2 -
SpectraSuite/QuantLM_1.1B_8bit_Unpacked
Text Generation • 1B • Updated • 5
QuantLMs, unpacked to FP16 format - compatible with FP16 GEMMs. After unpacking, QuantLMs have the same architecture as LLaMa.
QuantLMs, unpacked to FP16 format - compatible with FP16 GEMMs. After unpacking, QuantLMs have the same architecture as LLaMa.
QuantLMs, unpacked to FP16 format - compatible with FP16 GEMMs. After unpacking, QuantLMs have the same architecture as LLaMa.